[PATCH] D83181: [x86] improve codegen for non-splat bit-masked vector compare and select (PR46531)

[Sanjay Patel via Phabricator via llvm-commits]

spatel marked an inline comment as done.
spatel added inline comments.


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.cpp:40260
+        TLI.isTypeLegal(VT) && ((Subtarget.hasAVX() && EltBitWidth == 32) ||
+                                (Subtarget.hasAVX2() && EltBitWidth == 64));
+    if (CanShiftBlend &&
----------------
RKSimon wrote:
> XOP has more vector shifts and vpcmov which should allow 8/16-bit cases as well - I added testing at rGd6c72bdca2f2
Ok - I'll enable XOP for all legal types, and we can decide if we need to exclude any types based on those diffs. I don't have a good sense of what's good/bad/possible with those instructions. 


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D83181/new/

https://reviews.llvm.org/D83181