Repository navigation
CUDA: Add Support to Creating StructModel Array - #7772
Conversation
|
gpuci run tests |
|
Question: I think you meant to use the registered type to construct an array? x = cuda.local.array(3, dtype=types.Float3) |
|
Oops, apologies for the typo. |
There isn't a |
|
gpuci run tests |
1 similar comment
|
gpuci run tests |
This reverts commit b7b5281.
|
/azp run |
gmarkall
left a comment
There was a problem hiding this comment.
I think this is looking good and almost ready to merge:
- I think the fail is because #7817 needs merging to resolve an issue with the docs build (unrelated to this PR, so shouldn't be an issue)
- Could you rename
utils.pytoextension_usecases.pyin line with the naming of other files containing test use cases please? - I was going to suggest editing the docs to mention that CUDA extensions can be used in shared and local memory now, but I note there's absolutely no documentation on extending the CUDA target that's part of the official documentation at the moment. This is something that we ought to fix, but not for this PR.
|
gpuci run tests |
gmarkall
left a comment
There was a problem hiding this comment.
Many thanks for the latest round of changes - this now looks good to me!
|
@esc Could this have a CUDA smoketest buildfarm run please? |
|
|
Was 💚 |
|
I think this can be marked as ready to merge! |
This PR adds support to creating a cuda array of user defined
StructModeltype. An example usage is like: