Skip to content

Limit the CUDA cmake architectures to 86 for CI builds - #2161

Merged
kunal-vaishnavi merged 2 commits into
mainfrom
baijumeswani/build-fewer-sms
May 14, 2026
Merged

Limit the CUDA cmake architectures to 86 for CI builds#2161
kunal-vaishnavi merged 2 commits into
mainfrom
baijumeswani/build-fewer-sms

Conversation

@baijumeswani

Copy link
Copy Markdown
Collaborator

No description provided.

Copilot AI review requested due to automatic review settings May 13, 2026 17:49

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Pull request overview

This PR narrows the CUDA architectures built by Windows GPU CI workflows to SM 86, matching the A10 runner hardware and reducing CI build time/memory versus the broader release-oriented preset defaults.

Changes:

  • Adds explanatory comments in the Windows CUDA and WinML workflows.
  • Overrides CMAKE_CUDA_ARCHITECTURES to 86-real during CMake configure for those CI jobs.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File Description
.github/workflows/win-winml-x64-build.yml Limits WinML x64 CI CUDA compilation to SM 86.
.github/workflows/win-cuda-x64-build.yml Limits CUDA x64 CI CUDA compilation to SM 86 while preserving sccache compiler launcher settings.

Comment thread test/python/requirements.txt
@baijumeswani

Copy link
Copy Markdown
Collaborator Author

Closing-reopening to trigger all CIs again.

@kunal-vaishnavi
kunal-vaishnavi merged commit 15febe2 into main May 14, 2026
28 of 32 checks passed
@kunal-vaishnavi
kunal-vaishnavi deleted the baijumeswani/build-fewer-sms branch May 14, 2026 16:47
@baijumeswani

Copy link
Copy Markdown
Collaborator Author

Thanks for the review. :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants