Skip to content

Conversation

michal-shalev
Copy link
Contributor

@michal-shalev michal-shalev commented Aug 20, 2025

What?

Add GPU device function stubs for UCX backend with template-based coordination levels (thread, warp, block, grid) for memory transfers, signal transfers, and status checking.

Why?

To establish foundation for GPU-initiated transfers.
See also: #705

How?

Added 5 templated device functions in nixl_device.cuh that return NIXL_ERR_NOT_SUPPORTED as placeholders:
nixlGpuPostSingleWriteXferReq - memory transfers
nixlGpuPostSignalXferReq - signal transfers
nixlGpuPostPartialWriteXferReq - partial transfers
nixlGpuPostWriteXferReq - full transfers
nixlGpuGetXferStatus - status checking

@michal-shalev michal-shalev requested a review from a team as a code owner August 20, 2025 14:27
Copy link

copy-pr-bot bot commented Aug 20, 2025

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

Copy link

👋 Hi michal-shalev! Thank you for contributing to ai-dynamo/nixl.

Your PR reviewers will review your contribution then trigger the CI to test your changes.

🚀

@yosefe
Copy link
Contributor

yosefe commented Aug 25, 2025

/build

@michal-shalev michal-shalev changed the title API/CUDA/UCX: Create Device API API/GPU/UCX: Create Device API Aug 25, 2025
@yosefe
Copy link
Contributor

yosefe commented Aug 28, 2025

/build

@michal-shalev
Copy link
Contributor Author

/build

Signed-off-by: Michal Shalev <[email protected]>
@michal-shalev
Copy link
Contributor Author

/build

@yosefe yosefe merged commit add1824 into ai-dynamo:main Aug 28, 2025
16 of 17 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants