Skip to content

test/mpi: Add non-collective sessions init test #7429

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

test/mpi: Add non-collective sessions init test #7429

wants to merge 3 commits into from
+60 −7

Conversation

@raffenet
Copy link
Contributor

@raffenet raffenet commented May 21, 2025 •
edited
Loading

Pull Request Description

Add a test to verify the non-collective behavior of
MPI_Session_init. First, run a collective init/finalize sequence over
all launched processes. Second, try to re-initialize and finalize a new
session from just a single process and ensure it does not hang.

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.

@raffenet
Copy link
Contributor Author

raffenet commented May 21, 2025

test:mpich/ch4/most

@raffenet raffenet force-pushed the session-non-coll branch from 8bd3d15 to 05bfcfb Compare May 21, 2025 19:43
@raffenet
Copy link
Contributor Author

raffenet commented May 21, 2025

test:mpich/ch4/most

hzhou
hzhou reviewed May 22, 2025
View reviewed changes
test/mpi/session/session_non_coll.c Outdated

MPI_Session_init(MPI_INFO_NULL, MPI_ERRORS_ABORT, &session);
MPI_Group_from_session_pset(session, "mpi://world", &group);
MPI_Group_rank(group, &rank);
Copy link
Contributor

@hzhou hzhou May 22, 2025

Choose a reason for hiding this comment

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

Also check size and the tests require size > 1

Copy link
Contributor Author

@raffenet raffenet May 22, 2025

Choose a reason for hiding this comment

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

new version does this

@raffenet raffenet force-pushed the session-non-coll branch from 05bfcfb to 694d395 Compare May 22, 2025 18:34
raffenet
raffenet commented May 22, 2025
View reviewed changes
@raffenet raffenet requested a review from hzhou May 28, 2025 15:17
hzhou
hzhou requested changes Jun 5, 2025
View reviewed changes
@raffenet
Copy link
Contributor Author

raffenet commented Jun 20, 2025

test:mpich/ch4/most

@raffenet raffenet requested a review from hzhou June 20, 2025 19:40
raffenet
raffenet commented Jun 20, 2025
View reviewed changes
@raffenet
Copy link
Contributor Author

raffenet commented Jun 20, 2025

test:mpich/ch4/most

@raffenet
Copy link
Contributor Author

raffenet commented Jun 20, 2025

test:mpich/ch4/most

@raffenet
Copy link
Contributor Author

raffenet commented Jun 23, 2025

The test fails with ch4:ofi when the PSM3 provider is selected. FI_PROVIDER=^psm3 allows it to pass.
ch4:ucx fails because of a PMI world barrier in MPIDI_UCX_mpi_finalize_hook.

Will address these issues in a follow-up patch.

@raffenet
Copy link
Contributor Author

raffenet commented Jun 23, 2025

The test fails with ch4:ofi when the PSM3 provider is selected. FI_PROVIDER=^psm3 allows it to pass.

This is actually the same issue. When psm3 is in use, the ch4:ofi netmod will do a PMI barrier in finalize.

@raffenet
Copy link
Contributor Author

raffenet commented Jun 23, 2025

test:mpich/ch4/most
test:mpich/ch3/most

@raffenet
Copy link
Contributor Author

raffenet commented Aug 11, 2025

test:mpich/ch4/most
test:mpich/ch3/most

@raffenet
Copy link
Contributor Author

raffenet commented Aug 11, 2025

test:mpich/ch4/most
test:mpich/ch3/most

@raffenet
Copy link
Contributor Author

raffenet commented Aug 19, 2025

test:mpich/ch4/most
test:mpich/ch3/most

@raffenet
test/mpi: Fix .gitignore for sessions tests
e22288f
@raffenet
test/mpi: Add local only sessions init test
ec6b39f 
Add a test to verify the local only behavior of MPI_Session_init. To do
so, we launch an extra process via mpiexec that will never call
MPI_Session_init.
@raffenet
ch4/{ofi,ucx}: Skip collective finalize in sessions model
4b8a027 
If there is no MPI_COMM_WORLD, we should skip calling a global PMI
barrier during finalization. It is not guaranteed that world processes
have initialized MPI (thus PMI), so a barrier could hang.
@raffenet
Copy link
Contributor Author

raffenet commented Sep 30, 2025

test:mpich/ch4/most
test:mpich/ch3/most

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

Reviewers

@hzhou hzhou Awaiting requested review from hzhou

Requested changes must be addressed to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@raffenet @hzhou