[PATCH] D32953: [X86][SSE] Improve combineLogicBlendIntoPBLENDV to use general masks.

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: test/CodeGen/X86/vselect-pcmp.ll:144
 
 ; Sorry 16-bit, you'll never be important enough to support?
 
----------------
You can remove those snarky 16-bit comments. 😃 


================
Comment at: test/CodeGen/X86/vselect-pcmp.ll:154-156
 ; AVX1-NEXT:    vandnps %ymm1, %ymm2, %ymm1
 ; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
----------------
Looks like we have to go to extremes to get the AVX1 case although this may improve with a patch that I'm working on for PR32790:
https://bugs.llvm.org/show_bug.cgi?id=32790


Repository:
  rL LLVM

https://reviews.llvm.org/D32953