Issue 244943
Summary in some situations, _mm256_andnot_pd() produces a vpxor + vpand sequence instead of a vpandn
Labels new issue
Assignees
Reporter Hamilcar34
    Since clang 16, the _mm256_andnot_pd() intrinsic produces a vpxor + vpand sequence instead of a vpandn instruction in some situations, it was ok with clang 15.
A repro case is available here: https://godbolt.org/z/d1xrG14rE
Thanks

_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs

Reply via email to