Add wasm32 SIMD128 kernel for sgemm#99
Merged
Merged
Conversation
Contributor
Author
|
@bluss Could you take a look? This helps my personal wasm use case. |
Owner
|
The PR i missing tests, nothing in CI is running tests (from tests/) directory for your new code. |
8x8 microkernel ported from the AArch64 NEON path. Gated on target_arch="wasm32" + target_feature="simd128"; falls through to the scalar kernel otherwise.
Run tests/ and lib unit tests on wasm32-wasip1 via wasmtime, with matrix entries for both the new simd128 kernel and the scalar fallback path.
Contributor
Author
|
@bluss Thanks for the comment. I added CI for wasm using wasmtime |
Owner
|
Thanks, that's accepted. wasm simd support for sgemm (f32) only in this case, which is I assume the most used kernel. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
8×8 microkernel (ported from the AArch64 NEON path).
Gated on
cfg(all(target_arch="wasm32", target_feature="simd128")).Since wasm has no standardized runtime feature detection for SIMD,
dispatch is compile-time only and the scalar fallback stays in place
otherwise.
In wasmtime 36 on Apple M2 Pro, sgemm runs 3.86×–4.11× faster than the scalar fallback across 64x64x64–512x512x512.