Add GPU support #75
Merged
Add GPU support #75
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
We don't have CUDA testing through Github actions but on HiPerGator with CUDA we see that all CUDA tests pass. Not sure why the ExteriorCalculus tests take so long here but I don't see this behavior on my own machine. I think this is ready to merge. |
lukem12345
commented
May 2, 2024
lukem12345
commented
May 2, 2024
lukem12345
commented
May 2, 2024
This was referenced May 2, 2024
lukem12345
commented
May 3, 2024
| @test all(isapprox.(dec_inv_hodge_star(Val{i}, sd, GeometricHodge())(V_1), Array(dec_inv_hodge_star(Val{i}, sd, GeometricHodge(), Val{:CUDA})(CuArray(V_1))); atol = 1e-10)) | ||
| end | ||
| for i in 1:1 | ||
| for sd in dual_meshes_2D[1:end-1] |
There was a problem hiding this comment.
The sphere is skipped here because you need to add support for meshes embedded in 3D here before we merge right
There was a problem hiding this comment.
No all meshes work fine. The matrix operators are generated on the CPU, as before, and just ported to the GPU. I skip the sphere here because it's inverse calculation is very unstable for some reason. To make the tests pass I'd need to drop down atol significantly.
Also added some higher level calls to wedge_product
Now need to change functions to be overloading
The method of handling CUDA functions from the extension should change. For now we have placeholder functions in the main module that are solely meant to be overloaded. This should probably changed to have CUDA overload the regular CPU variants. Also for some reason being in the test environment does not let the extension load, meaning the CUDA tests aren't running.
This seems to be needed since CUDA is really iffy with Int8s.
Also added DEC CUDA matrix checks
Tests run if ran manually. will not work with the ] test command. Should consider test-specific dependencies in the top-level project.toml: https://pkgdocs.julialang.org/v1/creating-packages/#Test-specific-dependencies
Also removed the test/Project.toml and updated the main Project.toml to include extra test dependencies.
This has been replaced to work with the multiple-dispatch method used by other CUDA operators.
No history here
Also only supports Julia 1.10
GeorgeR227
force-pushed
the
CUDA-Extensions
branch
from
34acd97 to
4105b3f
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Before we merge, test that all DEC CUDA matrices load properly