fix: route avxvnni FMA ops through fma3<avx2> kernels#1368
Open
DiamonDinoia wants to merge 1 commit into
Open
Conversation
batch<T, avxvnni> derived from avx2, so fnma/fnms fell back to the generic neg(x*y)+z form (vxorpd + vfmadd) instead of the hardware vfnmadd/vfnmsub kernels registered for fma3<avx2>. This bites -march=native on Alder/Meteor Lake and Zen 5, where default_arch resolves to avxvnni. Derive avxvnni from fma3<avx2> instead. fma3<avx2> always derives from avx2 and its kernels are guarded by XSIMD_WITH_FMA3_AVX2, so when FMA is disabled the base is transparent and dispatch falls through to avx2 unchanged. Validated under Intel SDE across avxvnni+FMA (-adl), fma3<avx2> (-hsw), avx2-only (-hsw) and avxvnni-without-FMA (-adl): correct results, full test_batch suite passes, vfnmadd emitted only where FMA is enabled.
DiamonDinoia
force-pushed
the
fix/avxvnni-fma3-dispatch
branch
2 times, most recently
from
c3bcc77 to
3a25348
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
batch<T, avxvnni> derived from avx2, so fnma/fnms fell back to the generic neg(x*y)+z form (vxorpd + vfmadd) instead of the hardware vfnmadd/vfnmsub kernels registered for fma3. This bites -march=native on Alder/Meteor Lake and Zen 5, where default_arch resolves to avxvnni.
Derive avxvnni from fma3 instead. fma3 always derives from avx2 and its kernels are guarded by XSIMD_WITH_FMA3_AVX2, so when FMA is disabled the base is transparent and dispatch falls through to avx2 unchanged.
Validated with Intel SDE across avxvnni+FMA (-adl), fma3 (-hsw), avx2-only (-hsw) and avxvnni-without-FMA (-adl).