Skip to content

Update identify_stream_usage CUDA runtime hooks to CUDA 13#19807

Merged
rapids-bot[bot] merged 2 commits intorapidsai:branch-25.10from
robertmaynard:bug/update_identify_stream_usage_to_cuda_13
Aug 27, 2025
Merged

Update identify_stream_usage CUDA runtime hooks to CUDA 13#19807
rapids-bot[bot] merged 2 commits intorapidsai:branch-25.10from
robertmaynard:bug/update_identify_stream_usage_to_cuda_13

Conversation

@robertmaynard
Copy link
Contributor

Description

The kernel launch and memcpyasync API have changed and we need updates to work properly.

Checklist

  • I am familiar with the Contributing Guidelines.
  • New or existing tests cover these changes.
  • The documentation is up to date with these changes.

The kernel launch and memcpyasync API have changed and we need
updates to work properly.
@robertmaynard robertmaynard requested a review from a team as a code owner August 27, 2025 15:48
@robertmaynard robertmaynard requested review from vyasr and wence- August 27, 2025 15:48
@robertmaynard robertmaynard added bug Something isn't working non-breaking Non-breaking change labels Aug 27, 2025
Copy link
Contributor

@davidwendt davidwendt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Verified this works on my local machine.

@davidwendt
Copy link
Contributor

Reference #19768

Comment on lines +214 to +215
// We need to define the __cudaLaunchKernel ABI as
// it isn't part of cuda_runtime.h when compiling as a C++ source
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do we know why this is the case in CUDA 13?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

They changed their API.
My understanding is that the <<< >>> pattern changed to map to a different API which we need to now override.
The internal __cudaLaunchKernel declaration is hidden under ifdef __CUDACC__ in the header files in 13 which makes sense given how it is intended to be used. We just need to hook it here (without calling it) so the forward references are required in this .cpp file.

@davidwendt
Copy link
Contributor

/merge

@rapids-bot rapids-bot bot merged commit 6dc52ce into rapidsai:branch-25.10 Aug 27, 2025
92 checks passed
@jameslamb jameslamb mentioned this pull request Aug 27, 2025
2 tasks
@robertmaynard robertmaynard deleted the bug/update_identify_stream_usage_to_cuda_13 branch August 28, 2025 14:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working non-breaking Non-breaking change

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants