Construct shapes in memory and support more shapes #1436

gunney1 · 2024-10-02T14:39:25Z

Extend interface for intersection shaping

This PR is an extension of the support for intersection shaping
It does the following:
- Adds a number of shapes to intersection shaping, such as analytical spheres, cones, cylinders. Currently, intersection shaping supports only shapes described by tet meshes and C2C surfaces of revolution.
- Adds interfaces for specifying the shapes in memory (instead of reading them from files)
- Re-factors some code to reduce dependence on the MFEM mesh interface, as an initial step toward supporting Blueprint meshes. The full Blueprint support is in a companion PR.

This PR is the first of 3 related PRs. They are separated to keep the code changes to a manageable size.

This PR is a first pass at supporting more shapes generating those shapes in memory. We would like to get something working asap for an application requirement. We will revisit the interface and performance by Spring 2025.
The second PR is to extend support to Blueprint meshes. Support Blueprint mesh in shaping #1455. It should be reviewed AFTER this one is approved.
In the first 2 PRs (including this one), I tried to minimize re-factoring and moving codes around. The third PR will be to re-factor and clean up. See issue Revisit interface design and performance in IntersectionShaper #1445.

This commit adds test code for testing in-memory shape construction and shaping with those objects.

…hape.

Fix by changing the pointer references to shared_ptr references. This bug was recently created when factoring out the generation of the mesh so it can be saved outside the scope it was generated in.

…ct-shapes-in-memory

to ensure the user provided a blueprint mesh.

Want to avoid having small ones that aren't well resolved with a coarse mesh.

…ct-shapes-in-memory

The check is valid but it causes a mysterious double-free error with gcc-13 on docker tests when shared libs are used, even if the checking code is never executed.

…ct-shapes-in-memory

bmhan12

LGTM!

I appreciate the "todo" issue tracking in #1445.

kennyweiss

Thanks @gunney1!

Please ensure that API changes are documented in the RELEASE_NOTES.

I have a concern about whether the proposed change to the inout octree API will work with our Fortran interface. Please test this with shroud's make generate command followed by running the Fortran tests.

src/axom/klee/Geometry.hpp

src/axom/klee/GeometryOperatorsIO.hpp

src/axom/quest/DiscreteShape.cpp

src/axom/quest/examples/quest_shape_in_memory.cpp

gunney1 · 2025-01-11T00:01:51Z

Thanks @gunney1!

Please ensure that API changes are documented in the RELEASE_NOTES.

I have a concern about whether the proposed change to the inout octree API will work with our Fortran interface. Please test this with shroud's make generate command followed by running the Fortran tests.

All the tests passed after make generate. Does that include the Fortran tests? Or is there a separate make target for Fortran tests? I can't tell from the make help output.

kennyweiss · 2025-01-11T01:05:37Z

I have a concern about whether the proposed change to the inout octree API will work with our Fortran interface. Please test this with shroud's make generate command followed by running the Fortran tests.

All the tests passed after make generate. Does that include the Fortran tests? Or is there a separate make target for Fortran tests? I can't tell from the make help output.

Thanks @gunney1. I looked a bit closer, and we don't expose the inout_init() overload with the mint::Mesh, so I think it's ok:

axom/src/axom/quest/interface/quest_shroud.yaml

Lines 40 to 48 in 8f678a9

    
           ## InOut query methods 
        
           - decl: int inout_init( const std::string& fileName, MPI_Comm comm) 
        
             format: 
        
               function_suffix: _mpi 
        
             cpp_if: ifdef AXOM_USE_MPI 
        
           - decl: int inout_init( const std::string& fileName) 
        
             format: 
        
               function_suffix: _serial 
        
             cpp_if: ifndef AXOM_USE_MPI

@ltaylor16 -- could you please look at the changes to inout.*pp and quest_shroud.yaml and confirm?

Arlie-Capps · 2025-01-13T18:25:36Z

Brian, I appreciate that you split this into three PRs. I'll be reviewing this one today.

src/axom/primal/operators/split.hpp

src/axom/quest/DiscreteShape.cpp

src/axom/quest/MeshViewUtil.hpp

Arlie-Capps · 2025-01-13T23:50:18Z

src/axom/quest/Shaper.hpp

+   * \brief Loads the shape from file into m_surfaceMesh and, if its a C2C
+   *        contour, computes a revolvedVolume for the shape.


"if it's a C2C contour" or "if the file is a C2C contour"

I re-worked the comment, which was written when this step was undergoing significant changes. Now, there are non-file shapes. This function now ensures that the discrete shape representation is ready for use, computing it if needed. Computing the revolved volume for C2C shapes is there for backward compatibility.

Arlie-Capps

Thanks, Brian.

…ct-shapes-in-memory

gunney1 added 26 commits July 19, 2024 08:58

Add const to input arguments to assure they are not changed.

04c3971

Preliminary ability to generate shape in memory.

7bd74e5

This commit adds test code for testing in-memory shape construction and shaping with those objects.

Add 2 more checks and exit with number of failures found.

88e8222

Fix bug that fails to apply GeometryOperators that are not composites.

126934b

Add optional scaling to test the use of geometry operator.

03b3f07

Add analytical sphere test and factor out code to generate discrete s…

2cbfb3c

…hape.

Remove obsolete code from Shaper.

758b7ab

Some name changes to fit accepted terminology.

711b29a

Add SOR, cylinder and cone to in-memory shaping.

63358e9

Small changes to comments.

cf58388

Implement arbitrary locations for VOR (not always on x-axis).

e3e8974

Add hexahedron shaping (in-memory only for now, not through klee).

381a75a

Set DiscreteShape default behavior for adaptive refinement of SOR.

931d1ba

Fix mesh pointers that get passed into classes then deleted.

4bee170

Fix by changing the pointer references to shared_ptr references. This bug was recently created when factoring out the generation of the mesh so it can be saved outside the scope it was generated in.

Get in-memory shaping working on rzansel host execution.

b409d47

Enable multi-policy testing for in-memory shaping.

e5ac6bd

Merge remote-tracking branch 'gh/develop' into feature/gunney/constru…

df21a07

…ct-shapes-in-memory

Support single tet shape in in-memory shaping.

6815047

Support plane shape.

ae8bfb2

Merge remote-tracking branch 'gh/develop' into feature/gunney/constru…

d51a85f

…ct-shapes-in-memory

Make some groups and views const where they are not modified.

382313e

More robust way to get temporary unique Group name.

d73df72

Change parameter ordering.

933e4e3

Rename memory-blueprint format to blueprint-tet and add checks

42bf676

to ensure the user provided a blueprint mesh.

Factor out code to shorten a method for readability.

6585fa2

Reformat.

408b427

gunney1 self-assigned this Oct 2, 2024

gunney1 added Quest Issues related to Axom's 'quest' component Klee Related to the Klee component labels Oct 2, 2024

Check for valid sidre Group in cases where it's required.

39232fd

gunney1 added 9 commits November 25, 2024 16:37

Remove need to enter explicit mesh dimension.

72e8769

Add TODO note about problems with using huge tets to represent plane.

7cb17c6

Autoformat.

6a497aa

Tweak shapes to make them about the same size.

1b87370

Want to avoid having small ones that aren't well resolved with a coarse mesh.

Change fake material name to match shape name.

a2e185c

Merge remote-tracking branch 'gh/develop' into feature/gunney/constru…

b3db9b4

…ct-shapes-in-memory

Remove verifying user mesh is a blueprint tet mesh.

fbb5a9e

The check is valid but it causes a mysterious double-free error with gcc-13 on docker tests when shared libs are used, even if the checking code is never executed.

Autoformat.

9ec6b08

Merge remote-tracking branch 'gh/develop' into feature/gunney/constru…

582a459

…ct-shapes-in-memory

gunney1 requested a review from Arlie-Capps January 6, 2025 13:21

gunney1 added 2 commits January 8, 2025 09:00

Update RELEASE-NOTES.

eb1cf6f

Merge remote-tracking branch 'gh/develop' into feature/gunney/constru…

eb226ee

…ct-shapes-in-memory

bmhan12 approved these changes Jan 9, 2025

View reviewed changes

kennyweiss requested changes Jan 10, 2025

View reviewed changes

kennyweiss approved these changes Jan 11, 2025

View reviewed changes

rhornung67 approved these changes Jan 13, 2025

View reviewed changes

Changes from code review.

69b57cc