Skip to content

[X86] lowerV64I8Shuffle - avoid lowerShuffleAsRepeatedMaskAndLanePermute call on VBMI targets#183109

Merged
RKSimon merged 1 commit into
llvm:mainfrom
RKSimon:x86-pr137422
Feb 24, 2026
Merged

[X86] lowerV64I8Shuffle - avoid lowerShuffleAsRepeatedMaskAndLanePermute call on VBMI targets#183109
RKSimon merged 1 commit into
llvm:mainfrom
RKSimon:x86-pr137422

Conversation

@RKSimon
Copy link
Copy Markdown
Contributor

@RKSimon RKSimon commented Feb 24, 2026

Shuffle combining fails to fold the inner shuffles first, but luckily the LanePermuteAnd* methods are enough if we have VPERMB as a fallback

Fixes #137422

…ute call on VBMI targets

Shuffle combining fails to fold the inner shuffles first, but luckily the LanePermuteAnd* methods are enough if we have VPERMB as a fallback

Fixes llvm#137422
@RKSimon RKSimon added the skip-precommit-approval PR for CI feedback, not intended for review label Feb 24, 2026
@RKSimon RKSimon enabled auto-merge (squash) February 24, 2026 17:25
@RKSimon RKSimon merged commit 1b9fea0 into llvm:main Feb 24, 2026
12 of 13 checks passed
@RKSimon RKSimon deleted the x86-pr137422 branch February 24, 2026 18:03
c-rhodes pushed a commit to llvmbot/llvm-project that referenced this pull request May 27, 2026
…ute call on VBMI targets (llvm#183109)

Shuffle combining fails to fold the inner shuffles first, but luckily the LanePermuteAnd* methods are enough if we have VPERMB as a fallback

Fixes llvm#137422

(cherry picked from commit 1b9fea0)
c-rhodes pushed a commit to llvmbot/llvm-project that referenced this pull request Jun 1, 2026
…ute call on VBMI targets (llvm#183109)

Shuffle combining fails to fold the inner shuffles first, but luckily the LanePermuteAnd* methods are enough if we have VPERMB as a fallback

Fixes llvm#137422

(cherry picked from commit 1b9fea0)
daunabomba pushed a commit to daunabomba/llvm-project that referenced this pull request Jun 2, 2026
…ute call on VBMI targets (llvm#183109)

Shuffle combining fails to fold the inner shuffles first, but luckily the LanePermuteAnd* methods are enough if we have VPERMB as a fallback

Fixes llvm#137422

(cherry picked from commit 1b9fea0)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

backend:X86 skip-precommit-approval PR for CI feedback, not intended for review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[X86] Suboptimal code for AVX-512 narrowing / blended permutex2var

2 participants