Add optional single-thread cutoff for small level-3 problems#942
Draft
jeffhammond wants to merge 1 commit into
Draft
Add optional single-thread cutoff for small level-3 problems#942jeffhammond wants to merge 1 commit into
jeffhammond wants to merge 1 commit into
Conversation
Below a work threshold (m*n*k), thread-spawn and barrier overhead dominates tiny GEMMs -- e.g. 16x16x16 runs ~10x slower with many threads than with one. Add a size-gated cutoff to the native and SUP level-3 thread decorators that forces a single thread when m*n*k < BLIS_SMALL_MT_THRESHOLD. The threshold is precision-aware: single-precision real GEMM uses a larger register tile (8x12 vs 6x8) so its small matrices have too few micro-tiles to parallelize, and its crossover is ~8x higher than double's. Opt-in: define BLIS_SMALL_MT_THRESHOLD (e.g. 200000) to enable; undefined leaves behavior unchanged. Threading policy only -- results are identical. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
jeffhammond
marked this pull request as draft
July 16, 2026 15:23
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Below a work threshold (mnk), thread-spawn and barrier overhead dominates tiny GEMMs -- e.g. 16x16x16 runs ~10x slower with many threads than with one. Add a size-gated cutoff to the native and SUP level-3 thread decorators that forces a single thread when mnk < BLIS_SMALL_MT_THRESHOLD. The threshold is precision-aware: single-precision real GEMM uses a larger register tile (8x12 vs 6x8) so its small matrices have too few micro-tiles to parallelize, and its crossover is ~8x higher than double's.
Opt-in: define BLIS_SMALL_MT_THRESHOLD (e.g. 200000) to enable; undefined leaves behavior unchanged. Threading policy only -- results are identical.