Skip to content

[RCCL] Harden toolchain-linux.cmake#4938

Open
nileshnegi wants to merge 7 commits intodevelopfrom
users/nileshnegi/rccl/fix/update-cmake-toolchain
Open

[RCCL] Harden toolchain-linux.cmake#4938
nileshnegi wants to merge 7 commits intodevelopfrom
users/nileshnegi/rccl/fix/update-cmake-toolchain

Conversation

@nileshnegi
Copy link
Copy Markdown
Contributor

@nileshnegi nileshnegi commented Apr 12, 2026

Motivation

Improve toolchain-linux.cmake robustness

Technical Details

  • Improve ROCm detection:
    • Check -DROCM_PATH cmake flag and ROCM_PATH env. flag.
    • Check user's PATH env.
    • Check for /opt/rocm as fallback.
  • Add redundant checks for CXX compiler paths.
  • Differentiate between DebugFast and normal Debug builds.

JIRA ID

Test Plan

Test Result

Submission Checklist

Copilot AI review requested due to automatic review settings April 12, 2026 05:53
@nileshnegi nileshnegi requested a review from a team as a code owner April 12, 2026 05:53
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

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 hardens the RCCL Linux CMake toolchain by making ROCm path and compiler detection more robust and by tweaking default debug flag behavior for “DebugFast” vs regular Debug.

Changes:

  • Add explicit ROCm path detection precedence (-DROCM_PATH > $ROCM_PATH > /opt/rocm) and fail fast if the path doesn’t exist.
  • Add compiler auto-detection fallbacks under ${ROCM_PATH}/bin and ${ROCM_PATH}/llvm/bin, respecting -DCMAKE_{C,CXX}_COMPILER and $CC/$CXX.
  • Adjust default Debug flags to differentiate “DebugFast” from normal Debug.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread projects/rccl/toolchain-linux.cmake Outdated
Comment thread projects/rccl/toolchain-linux.cmake Outdated
Comment thread projects/rccl/toolchain-linux.cmake Outdated
Comment thread projects/rccl/toolchain-linux.cmake Outdated
@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 12, 2026

regression-detection smoke (multi-node) on commit 3ad9128

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 12, 2026

regression-detection smoke (single-node) on commit 3ad9128

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 12, 2026

regression-detection smoke (multi-node) on commit 41c3e27

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 12, 2026

regression-detection smoke (single-node) on commit 41c3e27

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 13, 2026

regression-detection smoke (multi-node) on commit 41c3e27

Artifacts - No regressions preview generated.

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 14, 2026

regression-detection smoke (multi-node) on commit 3a42096

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 14, 2026

regression-detection smoke (single-node) on commit 3a42096

Artifacts - Results

Comment thread projects/rccl/toolchain-linux.cmake Outdated
Comment thread projects/rccl/toolchain-linux.cmake Outdated
@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 15, 2026

regression-detection smoke (multi-node) on commit feda0e3

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 15, 2026

regression-detection smoke (single-node) on commit feda0e3

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 15, 2026

regression-detection smoke (single-node) on commit 447e410

Artifacts - Results

@math-ci
Copy link
Copy Markdown

math-ci bot commented Apr 15, 2026

regression-detection smoke (multi-node) on commit 447e410

Artifacts - Results

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.

3 participants