rocfft: consider current rank when setting/checking callbacks #3663
+111
−42
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Outdated suggestions cannot be applied.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.
Motivation
Fix callback tests on MPI transforms, when not all ranks need the same number of callback pointers.
Technical Details
Tests try to assert that the correct number of callback pointers are passed to the execution info. With MPI transforms, this assertion was counting bricks on all ranks rather than just the current rank.
When the tests set callbacks on an MPI transform, they also need to only look at the current rank when deciding how many callback pointers to add to the execution info.
Similarly, when rocFFT builds up its own internal map of device IDs to callback pointers, it also needs to look at the current rank.
Test Plan
Ran MPI tests with callbacks enabled, both without
--ngpusand with--ngpus 1. Also ran single-proc multi-device tests with callbacks enabled.Test Result
Tests pass.
Submission Checklist