Fix compact process binding for OpenMPI mpirun #137
Merged
Conversation
This file contains 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
… compact process binding, since for e.g. 2 nodes 4 tasks it will lead to task 0 on node 0, task 1 on node 1, task 2 on node 0, task 3 on node 1. Compact would have been task 0 and 1 on node 0, and task 2 and 3 on node 1. Thus, this was a bug. Mapping by slot (combined with specifying PE=n) IS correct. The PE=n will set cpus-per-rank to n (see https://www.open-mpi.org/doc/current/man1/mpirun.1.php), while mapping by slot will mean that each rank is mapped to a consecutive slot. The only scenario in which mapping by slot will result in non-compact mapping is if oversubscription is enabled - then the oversubscribed ranks will be assigned round robin to the slots.
This was referenced May 3, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Currently, the mapping was set by node. That is NOT the same thing as compact process binding, since for e.g. 2 nodes 4 tasks it will lead to task 0 on node 0, task 1 on node 1, task 2 on node 0, task 3 on node 1. Compact would have been task 0 and 1 on node 0, and task 2 and 3 on node 1. Thus, this was a bug.
Mapping by slot (combined with specifying PE=n) IS correct. The PE=n will set cpus-per-rank to n (see https://www.open-mpi.org/doc/current/man1/mpirun.1.php), while mapping by slot will mean that each rank is mapped to a consecutive slot. The only scenario in which mapping by slot will result in non-compact mapping is if oversubscription is enabled - then the oversubscribed ranks will be assigned round robin to the slots. This shouldn't be an issue for the test suite, we don't do oversubscription here: our hooks always ask the resource scheduler for an amount of tasks equal to the amount of parallel processes we will launch. Thus, the number of slots should match the number of processes, and oversubscription isn't needed.