[PATCH] D86578: [TargetLowering] Combine known bits for icmp in SimplifySetCC (PR41182)

Mon Jan 17 09:59:02 PST 2022

RKSimon added a subscriber: foad.
RKSimon added inline comments.

================
Comment at: llvm/test/CodeGen/AMDGPU/wave32.ll:720
+; GFX1032: s_and_saveexec_b32 [[SAVE1:s[0-9]+]], [[SAVE2]]
+; GFX1032: s_or_saveexec_b32 [[SAVE2]], -1
 ; GFX1032: s_mov_b32 exec_lo, [[SAVE2]]
----------------
@foad I guess there's no good way around this if on wave32 targets - unless we duplicate the test with a 16 divergent variant?

================
Comment at: llvm/test/CodeGen/X86/avx512-mask-op.ll:623
+; X86-NEXT:    testb %al, %al
 ; X86-NEXT:    retl
 allocas:
----------------
test7 (and test6) don't look great (and the codegen was already redundant) - I wonder if we should at least have them return something?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86578/new/

https://reviews.llvm.org/D86578