Improve SVE2 optimizations of SimdWinogradKernel3x3Block2x2SetInput - #898
Merged
Merged
Conversation
Add an NCHW (trans=0) SVE2 path that processes F tiles with contiguous loads, svuzp deinterleave, and predicated padding/tails instead of the scalar Base fallback. Co-authored-by: igor.ermolaev <igor.ermolaev@altezza.org>
ermig1979
marked this pull request as ready for review
August 20, 2026 10:40
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The SVE2 implementation of
SimdWinogradKernel3x3Block2x2SetInputused the scalar Base path for NCHW (trans = 0), so it was much slower than NEON.This change adds a vectorized NCHW path that processes
F = svcntw()tiles at a time:svld1loads of overlapping windows, thensvuzp1/svuzp2to deinterleave even/odd spatial positions (nosvld2/svld4)PadNose1/PadTail1/PadTail2)srcHeight < 4orsrcWidth < 4) still fall back to Basetrans = 1) is unchangedRelease notes for 7.2.165 list the improvement.
No macros were added.
Correctness was checked by cross-compiling for AArch64 and running under QEMU at three SVE lengths:
neoverse-n2:svcntw=4(128-bit)neoverse-v1:svcntw=8(256-bit)max/a64fx:svcntw=16(512-bit)Each length ran 248 cases (31 shapes × pad 0/1 × trans 0/1 × Base-vs-Sve2 and Neon-vs-Sve2), including AutoTest sizes such as
512x7x6,256x14x12,128x28x24, and64x56x48. All passed.