Improvement of 2D thread-partitioned GEMM for M << N case#5276
Merged
martin-frbg merged 1 commit intoOpenMathLib:developfrom May 21, 2025
Merged
Conversation
Collaborator
|
Thank you |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #5270
The 2D thread partitioning in GEMM (PR#4655) requires nthreads_m % 2 == 0. This can prevent optimal nthreads_m and nthreads_n combinations on architectures like A64FX (48 cores) or Grace (144 cores) when M<<N, due to core counts having divisors other than 2.
Specifically, when matrix size N is significantly larger than M, the number of threads for N direction should be increased.
However, if nthreads_m includes divisors other than 2, such as 3, the increase of nthreads_n is prevented by ' nthreads_m % 2 == 0 '.
This modification removes the nthreads_m % 2 == 0 restriction and selects the combination that minimizes the following objective function 'n * nthreads_m + m * nthreads_n'.
This change improves the performance of multi-threaded GEMM for M << N cases.