[PATCH] D74863: AMDGPU/GlobalISel: Fix SALU mapping for v2s16 min/max

[Matt Arsenault via Phabricator via llvm-commits]

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


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp:1563
+                                               Register Src1) {
+  const LLT CmpType = LLT::scalar(32);
+  auto Cmp = B.buildICmp(Pred, CmpType, Src0, Src1);
----------------
foad wrote:
> Just curious: why s32 instead of s1 for comparison results?
Scalar compares need to use s32, and vector need to use s1: https://github.com/llvm/llvm-project/blob/master/llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp#L28



CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74863/new/

https://reviews.llvm.org/D74863