feature: detect MPI desynchronizations#4091
Draft
kdrienCG wants to merge 13 commits into
Draft
Conversation
kdrienCG
requested review from
MelReyCG,
OmarDuran,
bd713,
castelletto1,
corbett5,
cssherman,
dkachuma,
herve-gross,
jhuang2601,
joshua-white,
rrsettgast and
tjb-ltk
as code owners
kdrienCG
added
type: feature
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The goal of this PR is to add a way to detect MPI desynchronizations.
MPI desynchronizations occur when an MPI operation is not called on all ranks (for example, due to an
ifcondition or aforloop). When this happens, subsequent MPI operations may become non-blocking, potentially leading to incorrect results.Identifying whether a desynchronization has occurred, and locating it, can be long.
This PR proposes mechanism to detect such desynchronizations. It also outputs the current stack trace when such desynchronization occurs.
Implementation
The current approach adds a
barrier(which should remain blocking even if a desynchronization occurs) after every collective MPI call, combined with atimeout()to detect infinite hangs. If the timeout is exceeded (there is an infinite hang), it indicates that not every ranks of the current communicator participated in the collective MPI operation we are monitoring, indicating a desynchronization.Compile option
This feature is currently behind a compile option (
-DGEOS_ENABLE_MPI_DESYNC_DETECTION=ON).All code related to the detection is behind guards, and doesn't get compiled if GEOS is not built with the compile option.
I am not anchored on this choice, neither for the implementation. I'm open to any proposition.