Skip to content

Latest commit

 

History

History
909 lines (836 loc) · 56.8 KB

RELEASE-NOTES.md

File metadata and controls

909 lines (836 loc) · 56.8 KB
Raw

Axom Software Release Notes

Notes describing significant changes in each Axom release are documented in this file.

The format of this file is based on Keep a Changelog.

The Axom project release numbers follow Semantic Versioning.

Unreleased - Release date yyyy-mm-dd

Added

  • Adds the following methods to axom::Array to conform more closely with the std::vector interface:
    • Array::front(): returns a reference to the first element
    • Array::back(): returns a reference to the last element
    • Array::resize(size, T value): resizes the array, and sets any new elements to value.
  • Adds an ArrayView::empty() method to return whether the view is empty or not.
  • Adds an area() function to primal::Polygon
  • Adds initial support for using Slam types on the GPU
  • Adds support for using ArrayViewIndirection indirection policy with slam::Map and slam::BivariateMap
  • Adds const_iterator support to slam::BivariateMap and slam::SubMap
  • Primal: Adds a Hexahedron primitive
  • Primal: Adds a clip() operator for computing the intersection of a Hexahedron and another Tetrahedron as a Polyhedron
  • Primal: Adds an intersection_volume() operator for computing the volume of intersection between a primitive and a Tetrahedron
  • Primal: Adds a primal::Polyhedron::from_primitive() operator that returns a Polyhedron object from a given primitive.
  • Adds DataStore::getBufferInfo() and Group::getDataInfo methods that insert information into a Conduit Node about buffers in a DataStore object or data in a Group subtree. The information can be accessed from the Node by the caller from specifically named fields in the Node.

Changed

  • IntersectionShaper now implements material replacement rules.
  • axom::Array move constructors are now noexcept.
  • Exported CMake targets, cli11, fmt, sol, and sparsehash, have been prefixed with axom:: to guard against conflicts.
  • DistributedClosestPoint query now supports any blueprint-valid mesh format, including multidomain. Domain underloading and overloading can be expressed using multidomain format. Closest points are identified by cp_rank, cp_domain_index, and cp_index. The new cp_domain_index specifies the domain containing the closest point.
  • DistributedClosestPoint interfacing variable names closest_point and min_distance have been changed to cp_coords and cp_distance, respectively, to match the naming convention of other interfacing variables.
  • Updates mfem dependency to v4.5
  • Updates to fmt version 9.1.0
  • Updates uberenv submodule to HEAD of main on 28Dec2022
  • Updates blt submodule to HEAD of develop on 28Dec2022
  • Adds vcpkg ports for RAJA, Umpire with optional OpenMP feature for automated Windows build
  • Reduce size of ArrayView::subspan to prevent accessing invalid memory.
  • Updates conduit dependency to v0.8.6
  • Adds vcpkg port for lua as optional dependency on Windows
  • Adds additional parameters to quest's PointInCell query to control the Newton solve from physical to reference space for a given element
  • Remove function pointer call in IteratorBase::advance()
  • Slam: IndirectionPolicy::data() now returns a reference to the underlying buffer Rebinding an indirection to a new buffer is now achieved through IndirectionPolicy::ptr(), which returns a mutable pointer to the buffer.

Fixed

  • Fixed issues with CUDA build in CMake versions 3.14.5 and above. Now require CMake 3.18+ for CUDA/non-gpu builds.
  • Checks validity of bounding boxes in primal's intersection operators against planes and triangles before using the geometry.
  • Improves import logic for lua dependency
  • Improves import logic for mfem dependency in device builds when mfem is configured with caliper

Version 0.7.0 - Release date 2022-08-30

Added

  • Adds a view() method to axom::Array class to simplify creation of a corresponding axom::ArrayView
  • Adds GPU/OpenMP support to spin::ImplicitGrid. The following functions run with the user-specified execution space (specified as a template argument on ImplicitGrid):
    • ImplicitGrid::insert(nelems, bboxes): insert a batch of bounding boxes into the implicit grid
    • ImplicitGrid::getCandidatesAsArray(nquery, queryObjs, ...): query the implicit grid for a batch of query objects, and generate a CSR-format array for the candidates. In addition, ImplicitGrid::getQueryObject() returns an object that may be used within a GPU kernel to query the implicit grid.
  • Added initial implementation of GPU/OpenMP-accelerated point-in-cell queries
  • Added an alternative surface mesh tester function to Quest, based on ImplicitGrid
  • Add const versions of begin() and end() for Array and ArrayView
  • Add support for passing compatible custom allocator IDs to axom::Array with explicitly specified memory space
  • Adds constructor overloads to axom::Array that uses both uninitialized data (ArrayOptions::Uninitialized) and custom allocators
  • Adds a comparison operator (operator<()) to StackArray. This allows it to be used as a key for std::map
  • Use compiler intrinsics for axom's bit utility functions: popCount(), trailingZeros() and leadingZeros()
  • Adds a random-access iterator to slam::DynamicSet
  • Adds an overload to ImpicitGrid::getCandidates() from a grid cell of the lattice. This makes it easier to iterate over the bins of the spatial index.
  • Defines iterator traits on axom::Array<T>/ArrayView<T> iterators, to allow passing iterator pairs to standard library functions
  • Adds full support for calling methods on axom::Array<T> allocated in device-only memory.
  • Adds ability to index into subarrays of a multidimensional axom::Array<T> using operator[] and operator()
  • Adds ability to build axom using the hip compiler. Support for running device kernels with hip will be added in the future.
  • Adds new host-configs for HIP on LLNL platforms
  • Adds GPU/OpenMP support to spin::UniformGrid. The following functions run with a user-specified execution space (specified as a template argument on UniformGrid):
    • UniformGrid::initialize(): creates/re-creates a uniform grid with an array of objects and their corresponding bounding boxes
    • UniformGrid::getCandidatesAsArray(): query the uniform grid for objects that share a grid cell with the query bounding box In addition, UniformGrid::getQueryObject() returns an object that may be used within a GPU kernel to query the uniform grid.
  • Adds ability to specify a storage policy in UniformGrid. Two policies are provided:
    • DynamicGridStorage stores the bins as an array of arrays (default)
    • FlatGridStorage stores the bins as a flat array of elements, where each bin is a slice of the array
  • Adds a templated uniform grid-based surface mesh tester function to Quest
  • Adds an initializer list constructor and assignment operator to axom::Array
  • Adds an overload of axom::Array::resize(ArrayOptions::Uninitialized, dims) to support resizes without constructing or initializing new elements
  • Adds examples and tests for using Slic interface in Fortran
  • Adds examples for using the BVH device traversal API
  • Adds a ScatteredInterpolation query to quest, which enables interpolating scalar fields at arbitrary points from a given 2D or 3D point mesh in the Mesh Blueprint format. The current implementation generates a Deluanay complex over the points and performs linear interpolation over the triangle/tetrahedron at the query points.
  • Adds a HIP execution policy for device kernels to run on AMD GPU hardware
  • Adds new Slic macros that allow you to selectively print messages only on root ranks. For example, SLIC_ERROR_ROOT(msg) and SLIC_ERROR_ROOT_IF(EXP, msg). This can be set via slic::initialize(bool is_root = true) or slic::setIsRoot().
  • Adds forward iterators to the Views and Groups of a sidre::Group. These can be accessed via the range-for syntax as for(auto& view: grp.views()){...}, Or using the iterator syntax as for(auto& it = grp.views().begin(), itEnd = grp.views().end(); it ! itEnd ; ++it) {...}, and similarly for the groups of a group.
  • Adds forward iterators to the Attributes and Bufferss of a sidre::DataStore, with a similar syntax, e.g. for(auto& buf : datastore.buffers()){...}.
  • Adds an overload of ImplicitGrid::getCandidatesAsArray() to accept query points/bounding boxes as an axom::ArrayView.
  • Adds a primal::closest_point(point,sphere) overload to find the closest point on a sphere to a given point
  • Adds an overload to quest's SignedDistance query to return the closest point on the surface to the query point and the surface normal at that point. Also exposes this functionality in quest's signed_distance C API.
  • Adds utility function for linear interpolation (lerp) of two numbers
  • Adds utility function to compute binomial coefficients
  • Adds a CurvedPolygon class to primal representing a polygon with BezierCurves as edges
  • Adds functions to compute the moments (area, centroid) of a CurvedPolygon
  • Adds functions to evaluate integrals over BezierCurve and CurvedPolygon objects
  • Adds a ArrayViewIndirection storage policy to Slam
  • Adds set accessor methods to slam::DynamicVariableRelation
  • Adds a new component to Axom, multimat, to simplify the handing of multi-material meshes and fields.
  • Adds functions to compute winding numbers and in/out queries for Polygon and CurvedPolygon objects.
  • Adds constants.hpp to primal to track geometric constants. Initially includes a value for PRIMAL_TINY, a small constant that can be added to denominators to avoid division by zero.
  • DistributedClosestPoint query now supports "domain underloading" -- ranks that are passed in can have empty object meshes and/or empty query meshes
  • 'BezierCurve' objects now support Rational Bezier curve functionality
  • Primal: Adds a clip() operator for computing the intersection of a Tetrahedron and another Tetrahedron as a Polyhedron
  • Added slic::outputLocalMessages() to output messages from the current rank to the console for MPI-enabled LogStreams.

Changed

  • Axom now requires C++14 and will default to that if not specified via BLT_CXX_STD.
  • Moved bit-twiddling functions to core component
  • axom::Array now default-initializes its data by default. To create an Array with uninitialized elements, pass an axom::ArrayOptions::Uninitialized as the first constructor argument.
  • axom::ArrayView<const T> can now be created from a const Array<T>
  • Added new ExecSpace template parameter to spin::ImplicitGrid. ExecSpace is now the second template parameter (out of three) and defaults to axom::SEQ_EXEC.
  • Instead of saving the entire DataStore, MFEMSidreDataCollection will now save only its domain and global groups
  • When an inlet::Field fails a range or valid value constraint, the provided value and corresponding range/set of valid values are now included in the error message
  • IOManager::write now allows the calling code to pass in the full name of the root file that it will produce
  • Improved consistency of orientation operations in primal::Plane, primal::Sphere, primal::orientation() and primal::in_sphere()
  • Improved efficiency for primal::in_sphere() -- the computations are now based on (D+1)x(D+1) determinants instead of (D+2)x(D+2) determinants for D-dimensional spheres
  • Improved efficiency for (signed) area/volume functions of primal::Segment, primal::Triangle and primal::Tetrahedron
  • Updates interface for primal::Sphere to use more of primal, e.g. uses primal::Point instead of T* to represent points
  • Adds circumsphere() functions to primal::Triangle and primal::Tetrahedron to return the Sphere that circumscribes the triangle/tetrahedron's vertices
  • Adds operator overloads to subtract a primal::Vector from a primal::Point to yield a new primal::Point
  • Consolidates quest::findTriMeshIntersections*() implementations for BVH and ImplicitGrid
  • BVH::find*() batch functions now return the total number of candidate intersections found
  • Enables empty axom::Array<T> to be iterated over with begin()/end()
  • Removed AXOM_VERSION_EXTRA in favor of axom::gitSHA() and adding the SHA to axom::getVersion() and axom::about()
  • Use more specific type trait checks in ArrayOps, to avoid generating unnecessary copies in fill/destroy operations on otherwise trivially-copyable/destructible types.
  • axom::Array now consistently propagates the allocator ID on copy, move, and swap operations when possible. This is a breaking change; copy-construction of a dynamic array from a device array will no longer automatically move the array to host memory, and will instead maintain the same allocator ID as the source array.
  • The device traversal method BVH::TraverserType::traverse_tree() now supports passing in arbitrary query objects for BVH traversal.
  • Moved inlet::LuaReader::solState() to be a protected function that now returns a std::shared_ptr<axom::sol::state>. This is an advanced feature that could cause users to break an input file state after verification. This also allows us to not expose axom/sol.hpp to all users of Inlet. This greatly reduces compile times. Using this feature requires both a derived class and including axom/sol.hpp in the user code.
  • Renamed some overloads of function createView of axom::sidre::Group which accept int ndims, IndexType *shape arguments to be createViewWithShape or createViewWithShapeAndAllocate.
  • Replaced an unused older incarnation of iterators in sidre with a new std-compliant implementation
  • Removed an out-of-date manually-generated header file in sidre sidre/core/sidre.hpp. We recommend using the automatically generated header file axom/sidre.hpp to include sidre functionality.
  • Removed functions from sidre::ItemCollection base class that were not common to all derived classes and added a new derived class sidre::IndexedCollection
  • Spin: BVH::findPoints/Rays/BoundingBoxes() candidate search methods now accept an axom::ArrayView<IndexType> for the offsets and counts output arrays, and return candidates as an axom::Array<IndexType>.
  • Renamed primal::Polygon::centroid() to primal::Polygon::vertexMean() because it was not actually computing the centroid.
  • axom:sidre:IndexType is now the same type as axom:IndexType. Before, Sidre always used int64_t. Now it respects the define AXOM_USE_64BIT_INDEXTYPE.
  • Mint now depends on the Slam component.
  • Renames indirection policies in slam: The c-array indirection policy was renamed from ArrayIndirection to CArrayIndirection and the axom::Array-based indirection policy was renamed from CoreArrayIndirection to ArrayIndirection.
  • Mfem dependency updated to 4.4
  • primal::detail::intersect_ray now correctly identifies intersections between collinear Segment and Ray objects.
  • Improved efficiency and robustness of barycentric coordinate and circumsphere computation for Triangles and Tetrahedra.

Fixed

  • Fixed a bug relating to swap and assignment operations for multidimensional axom::Arrays
  • Fixed over-eager caching of restored mfem::FiniteElementSpaces in sidre::MFEMSidreDataCollection
  • Fixed a bug in which Inlet verification bails out on the first failure, which resulted in incomplete error lists
  • Fixed a bug in quest::PointInCell when not using RAJA
  • Fixed a potential memory leak in axom::Array<T> for non-trivial types T which allocate memory
  • Added a guard in axom::ArrayList for axom configurations without Umpire to fix a compiler error (XL compiler)
  • Inlined some fully specialized functions in quest::Delaunay to avoid "multiply-defined" linker errors
  • Fixed axom::Array<T> fill operations on uninitialized memory
  • Fixed behavior of axom::Array<T>::resize(new_size) with new_size < curr_size
  • Fixed computation of signs in quest::SignedDistance when closest point is along an edge with a sharp dihedral angle and the adjacent triangles have significantly different areas
  • Fixed bug in axom::Path that ignored the leading delimiter character if one was present
  • Fixed gcc compiler errors in configurations without RAJA or Umpire
  • Fixed axom::Array<T> behavior on copy-construction when T is a non-trivial type
  • Replaced using statement in SidreDataTypesIds.h with typedef since C syntax is required in this file.
  • Fixed bug on two-dimensional sidre::Array<T> construction where the size is set to the underlying buffer capacity, instead of the actual number of elements
  • Fixed axom::Array<T>::insert behavior with non-trivial types.
  • Fixed bug in Slic macros for MPI-based LogStreams not aborting when using collective Error or Warning macros

Version 0.6.1 - Release date 2021-11-17

Added

  • Added a config variable, AXOM_DEBUG_DEFINE to control whether the AXOM_DEBUG compiler define is enabled. By DEFAULT, it is enabled for Debug and RelWithDebInfo configurations, but this can be overriden by setting AXOM_DEBUG_DEFINE to ON or OFF.
  • axom::Array is now GPU-compatible, in particular via a memory space template parameter and via extensions to axom::ArrayView that allow for copying into kernels and transfers between memory spaces.
  • Adds some utility arithmetic operators for adding and subracting primal::Points and primal::Vectors

Changed

  • Renamed AXOM_NOT_USED macro to AXOM_UNUSED_PARAM for better consistency with other Axom macros
  • Added explicit to axom::Inlet::InletVector constructors and added a constructor that accepts a double*
  • AXOM_ENABLE_MFEM_SIDRE_DATACOLLECTION configuration option is now ON by default (rather than OFF). This option should be disabled if mfem was configured with MFEM_USE_SIDRE.

Fixed

  • The AXOM_DEBUG compiler define is now properly exported via the axom CMake target when it is enabled
  • Added tolerance parameter EPS to primal::closest_point() operator. This effectively snaps closest points to the triangle boundaries vertices and edges when they are within EPS, improving consistency when, e.g., querying multiple triangles from the same mesh.
  • Fixed regression in SignedDistance queries for query points closest to edges or vertices of the input triangle mesh
  • Guard fmt's compiler defines from clashing with downstream fmt's.

Version 0.6.0 - Release date 2021-11-04

Added

  • Added new CMake option to allow users to turn off Axom created tools: AXOM_ENABLE_TOOLS
  • Inlet can now log verification errors to a user-processable list instead of using Slic
  • Sidre parallel I/O: Added new mapping arrays to the automatically-generated Blueprint index to support new schema for multi-domain parallel meshes.
  • Added support for optional third-party c2c ("contours to codes") library for parsing 2D spline data. c2c is currently only available for Axom configurations on LLNL platforms.
  • Primal's intersect(Ray, Segment) can now return the parametric coordinates of the intersection along both the ray and the segment (when the intersection exists)
  • Primal's intersect(Segment, BoundingBox) can now return the parametric coordinates bounding the portion of the segment contained within the BoundingBox (when the intersection exists)
  • Generalizes Quest's InOutOctree class to work with 2D line segment meshes. Previously, it only worked with 3D triangle meshes
  • Added support for reading c2c ".contour" files in Quest. Contours that enclose a 2D region can be linearized into line segment meshes and loaded into Quest's InOutOctree for in/out queries.
  • Updated the Quest inout C API to support 2D queries using the c2c library, when Axom is configured with c2c
  • Updated the C++ Quest "containment" example to support 2D in/out queries (in addition to the already supported 3D queries)
  • Added axom::Array modeled after std::vector. Previous axom::Array renamed to axom::MCArray. Future changes to both arrays are expected.
  • Added a data_collection_util tool to generate Mesh Blueprint compliant high order distributed meshes from an mfem mesh or over a Cartesian domain
  • Added utility functions axom::utilities::getHostName() and axom::utilities::getUserName().
  • Added new axom::primal::ZipIterable<T> type to convert structure-of-arrays data to a given Primal geometric primitive.
  • Quest: Added a computeDistances() function to SignedDistance class for batched signed-distance queries.
  • Spin: Added a getTraverser() function to BVH, enabling the customized traversal of a BVH from within a device kernel.
  • Primal: Adds an Octahedron primitive
  • Primal: Adds a Polyhedron primitive for representing convex polyhedra bounded by planar polygons in 3D
  • Primal: Adds a clip() operator for computing the intersection of a Tetrahedron and an Octahedron as a Polyhedron
  • Klee: Adds a new component, klee, for specifying non-conformal shape overlays for materials onto simulation meshes. This component defines a schema for defining, transforming and overlaying 2D and 3D shapes and validates klee input files. See the klee documentation for more information.
  • Quest: Adds a new query for sampling-based "shaping" onto low- or high-order computational meshes
  • Quest: Adds a new query for intersection-based "shaping" of revolved contours onto 3D hexahedral meshes. This capability uses a RAJA policy operate on various execution spaces (host, openmp, device).
  • Quest: Adds a "shaping" example for embedding a klee specification onto an MFEM mesh
  • Added Sidre function View::clear().
  • Core now provides an axom::ArrayView that provides view/indexing semantics over a raw pointer. This replaces the external buffer logic previously provided by axom::Array.

Changed

  • MFEMSidreDataCollection now reuses FESpace/QSpace objects with the same basis

  • Harden configuration options for BLT tools (style, code quality, etc.) against accidentally being enabled for users. Developers will always give a full path (e.g. CLANGFORMAT_EXECUTABLE)

  • Inlet: Writers are passed directly to Inlet::write instead of being registered

  • Inlet objects can now be constructed without a user-provided sidre::DataStore

  • Conduit version changed to v. 0.7.2

  • Renames AXOM_DEBUG_VAR macro to AXOM_UNUSED_VAR since there were many cases where the latter was not the appropriate name. This macro elides warnings about unused variables

  • Inlet's isUserProvided can now be used to query the status of subobjects of a Container via a name parameter

  • Upgrades our vcpkg usage for automated Windows builds of our TPLs to its 2021.05.12 release

  • Upgrades built-in cli11 library to its v1.9.1 release

  • Quest's inout C API has two new functions: inout_set_dimension() and inout_set_segments_per_knot_span(). The latter is only applicable for 2D queries on c2c contours

  • Spin: Refactored BVH public API based on user suggestions BVH constructor only handles setting up default values, while the actual building of the BVH is now done in a BVH::initialize(primal::BoundingBox*, int) method. Alternate Umpire allocator IDs are supplied via BVH::setAllocatorID(int). Other BVH methods have been modified to accept or return Primal primitives.

  • Spin: Removed hard dependency on RAJA and Umpire from BVH.

  • Moved slam::IteratorBase to axom::IteratorBase.

  • sidre::Array now derives from axom::MCArray.

  • axom::Array is now multidimensional; it intends to behave like std::vector in the 1D case and numpy.ndarray in the multidimensional case

  • Quest: SignedDistance has been modified to use spin::BVH instead of BVHTree. This enables signed-distance queries to run on the GPU, as specified via a new template parameter.

  • Spin: Removed BVHTree class in favor of BVH.

  • Quest's signed_distance C API: Removed functions related to old BVHTree class and added functions related to BVH class

    • Removed: void signed_distance_set_max_levels( int maxLevels )
    • Removed: void signed_distance_set_max_occupancy( int maxOccupancy )
    • Added: void signed_distance_set_allocator( int allocatorID )
    • Added: void signed_distance_set_execution_space( SignedDistExec execSpace )

  • All built-in third-party libraries (fmt, cli11, sol, and sparsehash) have been guarded to allow downstream users to have their own versions. This includes moving their headers under include/axom instead of include/ and moving their C++ namespace under axom (eg. fmt to axom::fmt). If you don't use our built-n TPLs this has no affect on you, but if you do these are some the changes you will need to make:

    Library Namespace changes Header include changes
    fmt fmt:: → axom::fmt:: #include "fmt/fmt.hpp" → #include "axom/fmt.hpp"
    sol sol:: → axom::sol:: #include "sol/sol.hpp" → #include "axom/sol.hpp"
    sparsehash google:: → axom::google:: #include "sparsehash → #include "axom/sparsehash
    cli11 CLI:: → axom::CLI:: #include "CLI11/CLI11.hpp" → #include "axom/CLI11.hpp"

  • Moved axom::MCArray and the sidre::Array it was based on into mint as axom::deprecated::MCArray and sidre::deprecated::MCArray, respectively. sidre::Array is now based on axom::Array.

  • utilities::string::split now returns a vector instead of using an out-parameter, Inlet's string utilities were moved to Core, and splitLastNTokens was renamed to rsplitN

  • axom::Array-related classes have been moved into individual files.

  • RAJA dependency updated to 0.14.0

  • Umpire dependency updated to 0.6.0. Support for versions prior to v2.1.0 was removed.

  • Conduit dependency updated to 0.7.2+ (develop as of Sept 13, 2021). This was required because Spack is now using HDF5's CMake build system.

  • Internal BLT dependency updated to 0.4.1

Fixed

  • Fixed Primal's intersect(Ray, Segment) calculation for Segments that do not have unit length
  • Fixed problem with Cray Fortran compiler not recognizing MSVC pragmas in axom/config.hpp. The latter are now only added in MSVC configurations.
  • Fixed bug in Mint's VTK output for fields of type int64 and float
  • Improved loading of data collections in MFEMSidreDataCollection
  • Added workaround to MFEMSidreDataCollection for C++14 standard library feature that was not available in [email protected]
  • Delayed finalizing reloaded mesh in MFEMSidreDataCollection until after setting the nodal GridFunction (when applicable)
  • Transposed R and Z coordinates when linearizing NURBS curves in c2c reader
  • Fixed user-reported in/out ambiguity within some InOutOctree cases with grazing triangles

Version 0.5.0 - Release date 2021-05-14

Added

  • Added the MFEMSidreDataCollection class for describing MFEM meshes and associated fields. This class was adapted from MFEM's SidreDataCollection and is enabled when Axom is built with MFEM and the AXOM_ENABLE_MFEM_SIDRE_DATACOLLECTION CMake option is enabled.
  • Added slic::setAbortFunction to configure a custom callback when Slic aborts.
  • Added a batched option to quest's InOutOctree containment query example application. This uses a kernel to test for containment on an array of points. The query uses OpenMP threading, when available.
  • Inlet: Added support for user-defined conversions from Inlet tables to user-defined types, and support for arrays of user-defined types
  • Added compiler define NOMINMAX to axom/config.hpp.in to avoid problems with the Windows min and max macros.
  • Added cpp14 variant to Spack package to allow Inlet::LuaReader to be used more easily.
  • Inlet: Added support for string-keyed associative arrays (dictionaries)
  • Inlet: Added support for defining and retrieving functions from the input file
  • Inlet: Added support for YAML and JSON input files
  • Inlet: Added support for mixed-key (integer and string) associative arrays
  • Inlet: Added support for deeply nested containers of structs
  • Inlet: Added support for void and strings in Lua-defined functions
  • Inlet: Added get<std::vector<T>> for retrieving arrays without index information
  • Inlet: Added a new Writer for generating JSON schemas which can be used by text editors for autocompletion
  • Inlet: SphinxWriter will now document the signature of function callbacks added to a schema
  • Axom::Path - New class for performing basic path operations with user-selectable delimiter characters
  • Inlet: Added a method to inlet::Inlet that retrieves the set of unexpected names in the input file
  • Inlet: Added an option to mark Containers as strict, which fail verification when unexpected entries are present
  • Added support in MFEMSidreDataCollection for registering QFunctions (data associated with quadrature points on a mesh)
  • Inlet: The internal hierarchy of an Inlet object can be reconstructed from a Sidre group, excluding callback functions
  • Added new overloaded version of method sidre::DataStore::generateBlueprintIndex to incorporate new MPI features in conduit and allow for generation of a blueprint index on an under-decomposed parallel mesh
  • Added new method sidre::View::importArrayNode to import a conduit::Node holding array data directly into a sidre::View
  • Added support for registering material and species sets in MFEMSidreDataCollection. These correspond to matsets and specsets in the Mesh Blueprint

Changed

  • Converted Uberenv to a git submodule. We previously vendored a copy of this script.
  • The Sidre Datastore no longer rewires Conduit's error handlers to Slic by default. It can be explicitly rewired using the static DataStore::setConduitSLICMessageHandlers() method.
  • Inlet: Changed SchemaCreator to an abstract class and added missing functions
  • Inlet: Added ability to access the Reader class from Inlet and Sol Lua state from the LuaReader class
  • Inlet: Switched accessor interface to match that of the STL with operator[] and T get<T>()
  • Inlet: std::shared_ptr<T> has been replaced with T& in non-owning contexts and std::unique_ptr<T> in owning contexts - specifically, within Inlet's internal tree structure
  • Unified core and SPIO unit tests into fewer executables to limit size of build directory
  • Renamed axom::slic::UnitTestLogger to axom::slic:SimpleLogger because it's used in more than just unit tests.
  • Inlet: Input file functions can now be of arbitrary signature subject to type and arity restrictions
  • Exported all symbols on Windows by default when compiling a dynamic library
  • Updated TPL conduit to version 0.6.0 released Nov 2, 2020.
  • Updated built-in TPL sparsehash to version 2.0.4 released Aug 11, 2020.
  • Inlet: Exposed primal::Vector in Lua for use in input-file-defined functions
  • The MFEMSidreDataCollection will now reconstruct fields and the mesh when a datastore is Loaded in
  • Inlet: Cleaned up Table interface to eliminate ambiguity and duplicated functionality
  • Inlet: Renamed DocWriter to Writer and refactored its interface
  • Inlet: Renamed Table to Container
  • Inlet collections of mixed or incorrect type will now fail verification, even if they're not marked as required
  • Required Inlet collections no longer fail Inlet verification if they are empty in the input file
  • Inlet: operator bool for Field and Container has been replaced with more precise isUserProvided and exists, which also returns true if a default value was specified.
  • Updated built-in TPL fmt to master branch snapshot, March 26, 2021.
  • Inlet: SphinxWriter will now print only one element schema per container instead of printing the same schema for each element in the container
  • Updated BLT to version 0.4.0 released 9 Apr 2021
  • Updated MFEM to version 4.2 released 30 Oct 2020. Axom no longer requires MFEM to be built serially
  • The macro for exporting symbols is now AXOM_EXPORT instead of AXOM_API
  • Updated Conduit to v0.6.0
  • Updated SCR to compatibility with v3.0rc1

Fixed

  • Updated to new BLT version that does not fail when ClangFormat returns an empty version string. BLT/Axom now issues a warning and disables the style build target if version is unknown or wrong.
  • Inlet: Apply lambda verifiers on generic containers to individual elements for consistency
  • Inlet: Fixed a bug relating to nested table lookups of primitive arrays and functions
  • Fixed a bug relating to deeply nested callback functions in Inlet
  • Inlet: Always ignore primitive array elements that do not match the requested type
  • Inlet: Empty structs/collections of structs with required sub-elements no longer fail verification
  • Quest: Fixed a bug with InOutOctree for triangles that lie on faces of octree blocks
  • Updated to use newer Conduit config directory
  • Add support for legacy hdf5 cmake build system

Version 0.4.0 - Release date 2020-09-22

Added

  • Exposed the tolerance parameter EPS that is used to determine intersections between triangles in primal:intersect() as an optional final parameter.
  • Added BVH spatial index option to the mesh_tester utility for calculating triangle-triangle intersection.
  • Added axom::execution_space< ExecSpace >::onDevice() to check if execution space is on device.
  • Added Axom macro AXOM_SUPPRESS_HD_WARN to silence host device compiler warnings.
  • Added option to quest's SignedDistance class and C API to toggle whether the distance query computes the sign.
  • Added a batched option to quest's signed distance query example application. This computes all distance queries on an array of points using a single call to computeDistance. The query uses OpenMP threading, when available.
  • Added new component, Inlet, to assist in retrieving and storing data from an input deck.
  • Added the ability to specify an Umpire allocator ID to use with the BVH. This allows the application to use a device allocator for the BVH and avoid use of Unified Memory (UM) on the GPU, which can hinder perfomrmance, or use a pool allocator to mitigate the latencies associated with allocation/deallocation. The allocator ID is specified as an optional argument to the BVH constructor.
  • Added new CMake option, AXOM_ENABLE_ANNOTATIONS, to enable/disable code annotations in Axom. Default is OFF.
  • Added Axom annotation macros. The macros can be used to annotate functions, using the AXOM_PERF_MARK_FUNCTION macro, or at a more fine grain level, different sections of code can be annotated by wrapping them within an AXOM_PERF_MARK_SECTION block. As a first cut, this works with NVTX tools. However, the hooks are in place to add support for Caliper in the future.
  • Added a simple interface to NVTX that allows an application to set the color and category for NVTX ranges corresponding to annotated code in Axom. The application can now call axom::nvtx:set_color() and axom::nvtx::set_category() to set the corresponding parameters respectively. This facilitates in the performance evaluation by allowing developers to easily filter out calls by category or visually by setting a different color to use in GUI tools, such as, NVVP and NSight.
  • Added a portable floating_point_limits traits class, to return min(), max(), lowest() and epsilon() of a float or double type. The functionality is equivalent to that provided by std::numeric_limits, but, the code is host/device decorated accordingly such that it can also be called on the device.
  • Added initial support for ray queries using the BVH. The caller may now supply a set of rays to a BVH and the BVH will return a set of candidate BVH bins that intersect each ray.
  • Added initial support for bounding box queries using the BVH. The caller may now supply a set of bounding boxes to a BVH and the BVH will return a set of candidate BVH bins that intersect each bounding box.
  • Added an axom-config.cmake file to axom's installation to streamline incorporating axom into user applications. See <axom-install>/examples/axom for example usages.
  • Added Sol as a built-in TPL for fast and simple C++ and Lua binding. Sol is automatically enabled when LUA_DIR is found. The version of Sol used in this release is v2.20.6, which requires C++14.

Removed

  • Removed the AXOM_ENABLE_CUB option, since CUB is no lonher used directly in Axom code. Instead, we use RAJA::stable_sort with RAJA-v0.12.1 and fallback to std::stable_sort with older versions of RAJA and when the code is built without RAJA.

Changed

  • Updated Axom to support RAJA-v0.12.1 and Umpire-v4.01, but the code remains backwards compatible with previous versions of RAJA and Umpire.
  • Transitioned Axom's code formatting tool from Uncrustify to clang-format. Axom's clang-format rules depend on clang 10.
  • Modified the command line interface for mesh_tester utility. Interface now uses a -m, --method option to select the spatial index, and -p, policy option now accepts a string or integer value.
  • Renamed the AXOM_USE_MPI3option to AXOM_ENABLE_MPI3 for consistency.
  • Modified the API for the BVH to accomodate different query types. The queries are now more explicitly called BVH::findPoints() and BVH::findRays().
  • Modified the API of Axom's memory management routines to not leak usage of Umpire. Instead of passing an umpire::Allocator object to specify an allocator, we now use the corresponding integer ID associated with the allocator.
  • All names in the C API now preserve the case of the C++ function. ex. SIDRE_datastore_new is now SIDRE_DataStore_new.
  • Fortran API in slic module. axom::slic::message Level enums are changed from enum-name_enumerator to namespace_enumerator. ex. level_error is now message_error.
  • Fortran derived-type constructors are now generic functions named afer the derived type. datastore_new is now SidreDataStore iomanager_new is now IOManager

Fixed

  • Fixed a bug in primal::intersect(Segment, BoundingBox) and added regression tests.
  • Spin's octrees can now be used with 64-bit indexes. This allows octrees with up to 64 levels of resolution when using a 64-bit index type.
  • Resolved issue with AXOM_USE_64BIT_INDEXTYPE configurations. Axom can once again be configured with 64-bit index types.
  • Fixed a triangle-triangle intersection case in primal that produced inconsistent results depending on the order of the triangle's vertices.
  • Fixed issue in the parallel construction of the BVH on GPUs, due to incoherent L1 cache that could result in some data corruption in the BVH. The code now calls __threadfence_system() after the parent is computed and stored back to global memory to ensure that the write is visible to all threads.
  • Fixed issue in Mint that would cause the [email protected] compiler to segfault. The mint_cell_types.cpp test was causing a segfault in the compiler. The main issue triggering this compiler bug was the use of constexpr when defining the static cell_info array of structs. The code has been modified to use const instead.
  • Fixed issue in Quest's Signed Distance query that would prevent consecutive calls to Quest when MPI-3 shared memory is enabled due to not properly nullifying internal pointers when finalize is called.
  • Fixed issue where the BVH would dispatch to the CPU sort() routine when the specified execution policy was CUDA_EXEC async. Now, when the execution policy is CUDA_EXEC the code would correctly dispatch to the GPU sort, using CUB (when CUB is enabled), regardless of whether it's synchronous or asynchronous.
  • Fixed issue with missing the bvh_traverse.hpp from the install prefix, which was preventing applications from using the BVH when pointing to an Axom install prefix.
  • Fixed usage of cuda kernel policies in Mint. Raja v0.11.0 changed the way max threads launch bounds is calculated. Consequently, a large number of threads was being launched leading to max registry count violation when linking. We are now using fixed kernel size of 256 threads (16x16 in 2D and 8x8x4 in 3D).
  • Third-party libraries can now build on the Windows platform through uberenv using vcpkg ("zero-to-axom support on Windows")

Known Bugs

  • Encountered a compiler bug on IBM LC platforms when using the IBM XL C/C++ compiler. The issue is manifested in the generate_aabbs_and_centroids method in the spin_bvh.cpp unit test. It seems that the compiler does not handle the lambda capture of the arrays correctly which leads to a segfault. A workaround for the IBM XL compiler is provided.
  • There is a known bug in MVAPICH that prevents consecutive creation/deletion of MPI windows. This was encountered on LC platforms when enabling shared memory in the Signed Distance Query. See the corresponding Github Issue for details.

Version 0.3.3 - Release date 2020-01-31

Added

  • Define different execution spaces. This refines and consolidates the execution policy concepts from mint and spin, which are now defined in Axom core, such that they can be used by other components.
  • Added a generic axom::for_all(), which can be used to write simple parallel loops.
  • Added CLI11 command line parser as a built-in third party library.

Changed

  • Updated Conduit to v0.5.1
  • Updated RAJA to v0.11.0
  • Updated Umpire to v2.1.0, which, natively handles zero byte re-allocations consistently. Workaround for older releases of Umpire is in place for backwards compatibility.
  • Updated BLT to develop (f0ab9a4) as of Jan 15, 2020
  • Set CUDA_SEPARABLE_COMPILATION globally instead of just in a few components.
  • Reduced verbosity of quest's InOutOctree in cases where query point lies on surface.
  • Changed semantics of axom::reallocate(_, 0) to always return a valid pointer. This matches the semantics of Umpire's reallocate(_, 0). Note: Umpire's PR-292 fixed a bug in its handling of this case and Axom includes a workaround to get the new behavior until we upgrade to Umpire v2.0+.

Fixed

  • Fixed a bug in convert_sidre_protocol example. Data truncation functionality now works properly when multiple Views point to the same data.

Version 0.3.2 - Release date 2019-09-22

Added

  • Added support in Mint for reading and writing an unstructured mesh in the SU2 Mesh file format. This includes support for both single and mixed cell type topology unstructured mesh types.
  • Added a new option to enable/disable use of CUB, AXOM_USE_CUB, which is disabled by default. This allows to disable CUB to circumvent issues encountered with the device linker.
  • Added a BezierCurve primitive type to primal. A new intersect operator was also added to compute the intersection points between a pair of Bezier curves of arbitrary order.

Changed

  • Updated Raja TPL to v0.9.0
  • Updated Umpire TPL to v1.0.0
  • AXOM_USE_OPENMP is now being set at configure time accordingly instead of auto-detected based on whether _OPENMP is passed by the compiler. This fixes issues where a host code would compile Axom w/out OpenMP, but, use Axom in parts of the code where OpenMP is enabled.

Fixed

  • Fixed usage of Umpire's MemoryResourceType enum in Axom. Axom was assuming that there was a one-to-one correspondance of the entries in the MemoryResourceType enum and the IDs of the predefined allocators. However, this assumption generally does not hold. This version corrects this by explicitly querying the ID of the predefined allocator for a particular resource and using that subsequently in the code.

Version 0.3.1 - Release date 2019-07-22

Added

  • Added a new implementation of the Bounding Volume Hierarchy(BVH) spatial acceleration data-structure to spin. The new BVH implementation relies on RAJA and allows for constructing and using the BVH sequentially or in parallel on the CPU or in parallel on the GPU. The new implementation is available only when Axom is compiled with RAJA and Umpire support.
  • Added Umpire and RAJA to the spack build.
  • Centralized Axom's memory management functions in a separate header and extended them to use Umpire when enabled.
  • Added the ability to point Axom to an UMPIRE build by specifying UMPIRE_DIR either in a host-config or at the command line. Axom components can link to Umpire, by specifying "umpire" as a dependency. A simple umpire smoke test is also added for regression testing.
  • Added for_all_faces to the mint execution model.
  • Added support for face connectivity in the mint UnstructuredMesh class.
  • Added support for face data and face connectivity in the mint StructuredMesh classes.
  • Added Python module for Quest signed distance interface. See the file src/axom/quest/interface/python/README.md for more information.
  • Added capability in sidre IOManager to read files while running on a number of MPI ranks greater than the number of ranks that were used to create the files.
  • Users can now set the vertex welding threshold parameter in Quest's In/Out query. This was previously not exposed to the user. The default value is 1E-9.
  • Unify all Axom component libraries into one library named axom.
  • The routine that checks if a surface mesh is watertight now marks boundary cells. A cell-centered field, named "boundary", is used to mark boundary cells with a value of "1" and "0" otherwise. This facilitates in visually inspecting the surface mesh and identify the problematic regions for the In/Out and SignedDistance queries.

Removed

  • Moved mint::Array to axom::Array with sidre storage in sidre::Array; also moved mint::IndexType to axom::IndexType.
  • Replaced sidre::SidreLength with sidre::IndexType.
  • Replaced usage of std::size_t in sidre with sidre::IndexType.
  • Added AXOM_ENABLE_EXPORTS which enables CMAKE_ENABLE_EXPORTS to allow demangled axom function names in stack traces. This option is ON by default in debug builds.

Changed

  • Updated conduit TPL to v0.4.0
  • Updated mfem TPL to v4.0
  • Slam's Set and Relation classes are now parameterized by a PositionType and ElementType. Its Map classes are now parametrized by a SetType.
  • Updated the fmt tpl.
  • Replaced old quest C-style interface with a new quest inout API. Functions related to the inout point containment query are prefixed with inout_. The new API has an option to set the verbosity of the inout initialization and query.
  • Changed sidre::IndexType to be a 64bit signed integer.
  • Changed slic::stack_trace to slic::internal::stack_trace which now attempts to output a demangled stack trace.

Fixed

  • Axom can once again be configured with AXOM_ENABLE_EXAMPLES=ON and AXOM_ENABLE_TESTS=OFF.
  • Quest's vertex welding now works with small welding threshold values (e.g. 1E-20). Welding was previously broken when this value was smaller than 1E-8. This fix also resolved an issue with small grid spacing values in primal's RectangularLattice.

Version 0.3.0 - Release date 2018-08-02

Added

  • Added initial implementation of a RAJA-based mesh-aware execution model in Mint. The execution model provides a functional interface for generic traversals and various mesh traversals, e.g., looping over all the nodes or cells of the mesh.
  • Added AXOM macros for lambda expressions and host/device decorators
  • Added the ability to point Axom to a RAJA build by specifying RAJA_DIR at a host config or at the config-build.py. Axom components can now link RAJA by specifying "raja" as a dependency. A simple RAJA smoke test is also added for regression testing.
  • Added isStructured() and isUnstructured() convenience methods to the mint::Mesh object.
  • Added support for using MPI-3 on-node shared memory data-structures to store the input surface mesh in quest. Instead of each rank duplicating the mesh data (e.g. node coordinates and cell connectivity), ranks within the same compute node can now share and utilize the same mesh data buffer. This reduces the memory overhead associated with duplicating the mesh at each rank which can be a limiting factor when reading large meshes. This feature is currently only exposed in Quest's signed distance query and requires Axom to be compiled with MPI-3 support (i.e., setting AXOM_USE_MPI3=ON)
  • Added the ability to call resize() on a Mint UnstructuredMesh. This functionality was not previously exposed in the UnstructuredMesh API.
  • Added support for non-closed surface mesh input to the signed distance query.
  • Added new interface for the signed distance query along with corresponding tests and examples.
  • Updated to fmt version 5.1.0
  • Added AXOM_ENABLE_TESTS which is a CMake dependent option of ENABLE_TESTS
  • Added AXOM_ENABLE_DOCS which is a CMake dependent option of ENABLE_DOCS
  • Added AXOM_ENABLE_EXAMPLES which is a CMake dependent option of ENABLE_EXAMPLES
  • Added jacobi_eigensolve() method for computing the eigenvalues and eigenvectors of real, symmetric matrices.
  • Added matrix_norm() operator for computing matrix norms. The implementations supports the p1-norm, infinity-norm and frobenious matrix norm.
  • Added eigen_sort() routine for sorting the supplied eigenvalues and corresponding eigenvectors in ascending order.
  • Initial integration of Mint and Sidre. Mint can now operate on meshes stored in Sidre and conform to the computational mesh blueprint conventions.
  • Added a sphere-sphere intersection test to Primal.
  • Added a utility function to Quest to weld vertices in a triangle mesh that are within a given tolerance. After welding, all triangles incident in a vertex have the same index for that vertex. This function has been integrated into the mesh_tester utility.
  • Added a bounded All-Nearest-Neighbor query to Quest. This query takes a list of point locations and associated regions, and for each point reports the nearest point in a different region that is no farther than a max search radius.
  • Axom now exports sparsehash version 2.0.3. Previously, it was only used internally.
  • Added a BitSet class to Slam.
  • Added a Tetrahedron primitive to Primal.
  • Added an in_sphere operator to Primal, which is a predicate that is used extensively for Delaunay triangulations.

Removed

  • Axom no longer depends on the Boost library.
  • Removed ENABLE_PYTHON CMake option. Python was only used by Shroud so restricted Python checks to when Shroud generation is enabled
  • Removed Lua as a dependency of Axom.
  • Removed signed distance query functions from the quest.hpp interface. The signed distance query is supported in its own exclusive interface.
  • Removed AXOM_NULLPTR. Use nullptr instead.

Changed

  • Simplified the external constructors for the Mint UnstructuredMesh. Specifically, the caller is no longer required to: (a) specify the dimension, which can be computed internaly based on other input arguments, and (b) specify explicitly the capacity for the node coordinate and connectivity arrays. In the more common case, the capacity is equivalent to the associated size when constructing a mesh by supplied external buffers.
  • Restructured source directory. #includes will now mirror file structure. For example, '#include "sidre/Group.hpp"' is now '#include "axom/sidre/core/Group.hpp"'.
  • The root CMake file for Axom is now located in <axom>/src's root directory, rather than in <axom>
  • Prefixed all Axom CMake options with AXOM_ to avoid conflicts
  • ENABLE_SPARSEHASH -> AXOM_ENABLE_SPARSEHASH
  • ENABLE_ALL_COMPONENTS -> AXOM_ENABLE_COMPONENTS
  • ENABLE_<component name> -> AXOM_ENABLE_<component name>
  • MINT_USE_64BIT_INDEXTYPE -> AXOM_MINT_USE_64BIT_INDEXTYPE
  • MINT_USE_SIDRE -> AXOM_MINT_USE_SIDRE
  • CMake minimum is now 3.8 for non-CUDA builds and 3.9 for CUDA builds
  • Axom now requires a C++11 compiler.
  • Refactored Axom's Matrix/Vector operators and consolidated them in one file.
  • Removed overloaded arithmetic operators from the Matrix class to avoid potential negative performance impacts. Applications should use the new matvecops methods for such operations.
  • Quest STL reader now returns a status code, indicating whether reading the STL file was successful. Also, the reader now leverages the improved Mint API to reserve storage for mesh and avoid re-allocations when reading the STL mesh data into the Mint mesh.
  • Refactored and cleaned up Primal's Sphere class.
  • Refactored Mint and removed all STL usage in preparation for GPUs.

Fixed

  • Fixed minor memory leak in quest fortran example
  • Bugfix for "multiply-defined" linker error in slam::Bitset and quest::PointInCellTraits

Version 0.2.9 - Release date 2018-03-08

Added

  • Updated to conduit version 0.3.1
  • Updated to shroud version 0.8.8
  • Improved platform support for LLNL's blue_os and bg/q systems. Axom now builds with Fortran enabled using the xlc and clang compilers.
  • Improved support for Axom on Windows, including new host-configs for Microsoft's Visual Studio compiler and for the intel compiler on Windows. All Axom components can now be built on Windows, but we do not yet support hdf5 or Fortran on Windows.
  • Added geometric Plane primitive to Primal. The Plane defines an oriented plane in 2D and 3D and provides support for operations such as projection of a point to a plane, signed distance and orientation.
  • Added ability to configure Axom (in particular Sidre and Spio) without hdf5.
  • Improved testing of Scalable Checkpoint Restart (SCR) library in Sidre.
  • Added a Point-In-Cell query to Quest. The Point In Cell query finds the cell in a computational mesh that contains an arbitrary point in space. If such a cell exists, it also finds the isoparametric coordinates of the point with respect to the cell. The query supports higher order mfem meshes.
  • Added cross-product and linspace operators to the vector utilities in numerics

Changed

  • The root cmake file for Axom is now located in <axom>'s root directory, rather than in <axom>/src
  • primal is no longer a header-only library.
  • Modified quest API to allow using a mint mesh that is already resident in memory.

Fixed

  • Fixed a divide-by-zero problem in primal::intersect()
  • Fixed the calculation of the Jacobian in mint::FiniteElement to support elements that are in higher-dimensional ambient space, e.g., surface elements, a Triangle or Quad in 3D.

Legend for sections

Added

  • Use this section for new features

Changed

  • Use this section for changes in existing functionality

Deprecated

  • Use this section for soon-to-be removed features

Removed

  • Use this section for now removed features

Fixed

  • Use this section for any bug fixes

Security

  • Use this section in case of vulnerabilities