common/ucx: call opal_progress when waiting for pmix_fence #11143
Conversation
|
Hello! The Git Commit Checker CI bot found a few problems with this PR: 31fb932: common/ucx: call opal_progress when waiting for pm...
Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks! |
| MCA_COMMON_UCX_PROGRESS_LOOP(worker) { | ||
| if(fenced) { | ||
| break; | ||
| } |
There was a problem hiding this comment.
seems the difference if whether progress is called if fenced==true before line 419?
but if fence is already completed why still need to call progress?
There was a problem hiding this comment.
no, the fix here is that MCA_COMMON_UCX_PROGRESS_LOOP also invokes opal_progress which progresses hcoll
|
Hello! The Git Commit Checker CI bot found a few problems with this PR: 31fb932: common/ucx: call opal_progress when waiting for pm...
Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks! |
Signed-off-by: Devendar Bureddy <[email protected]> (cherry picked from commit ed18cd3)
5 participants
Why
Fix hang during exit with TCP transport
How
Call opal_progress during pmix fence. Otherwise a ucx/tcp transport used by coll/hcoll component would not make progress an cause a hang.
Signed-off-by: Devendar Bureddy [email protected]