Skip to content

Conversation

@lowpolyneko
Copy link
Collaborator

@lowpolyneko lowpolyneko commented Jul 24, 2025

Pull Request Description

This PR adds a performance testlist to tests/mpi/bench for a future ch4:ofi gpu performance build job. To be merged in once the test is stabilized.

Author Checklist

  • Provide Description
    Particularly focus on why, not what. Reference background, issues, test failures, xfail entries, etc.
  • Commits Follow Good Practice
    Commits are self-contained and do not do two things at once.
    Commit message is of the form: module: short description
    Commit message explains what's in the commit.
  • Passes All Tests
    Whitespace checker. Warnings test. Additional tests via comments.
  • Contribution Agreement
    For non-Argonne authors, check contribution agreement.
    If necessary, request an explicit comment from your companies PR approval manager.

@lowpolyneko lowpolyneko changed the title test: add ch4:ofi --with-cuda peformance testfile test: add ch4:ofi --with-cuda performance testfile Jul 24, 2025
@lowpolyneko lowpolyneko force-pushed the ewong/bench-tests branch 2 times, most recently from 9ab0ecb to ee5b6c7 Compare July 28, 2025 17:25
@lowpolyneko lowpolyneko requested a review from hzhou July 28, 2025 18:15
@lowpolyneko
Copy link
Collaborator Author

Might be worth integrating #7536 into this depending on need.

@hzhou
Copy link
Contributor

hzhou commented Aug 13, 2025

Looks good. How about squash some commits? Each commit should be self-contained and do not do two things at once. But we don't have to keep multiple commits history showing how a single thing (e.g. adding the -gpu option to tests) was developed. That latter has little value for maintenance.

@lowpolyneko
Copy link
Collaborator Author

Agreed, squashed the PR to three commits. Feel free to squash and merge to reduce it further if needed.

Introduces a testlist running latency and bandwidth tests for senddev
and recvdev (host/device) combinations. The result of this testlist is
fed into gnuplot by the test suite to generate performance graph
artifacts.
Reintroduces the `gpu` testlist flag removed in PR pmodels#7506 as `hasgpu=?`.
However, it is now implemented as a GPU library dependency requirement
which skips tests if they are not detected by Automake. Also adds a
corresponding `MPITEST_HASGPU` environment variable for manual testing.
Copy link
Contributor

@hzhou hzhou left a comment

Choose a reason for hiding this comment

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

LGTM

@hzhou hzhou merged commit 7fcdc20 into pmodels:main Aug 14, 2025
4 checks passed
@hzhou hzhou changed the title test: add ch4:ofi --with-cuda performance testfile test: add ch4:ofi --with-gpu performance testfile Aug 21, 2025
@hzhou hzhou changed the title test: add ch4:ofi --with-gpu performance testfile test: add ch4:ofi --hasgpu performance testfile Aug 21, 2025
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.

2 participants