[PATCH] D68865: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Oct 11 12:08:51 PDT 2019


arsenm added inline comments.


================
Comment at: lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp:975
+
+  // FIXME: Allow v3i16/v3f16 in buffer intrinsics when the types are fully supported.
+  if (DMaskIdx < 0 &&
----------------
I think I have these working in GlobalISel already


================
Comment at: lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp:977
+  if (DMaskIdx < 0 &&
+      II->getType()->getScalarSizeInBits() == 16 &&
+      DemandedElts.getActiveBits() == 3)
----------------
!= 32 would be a bit safer


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68865/new/

https://reviews.llvm.org/D68865





More information about the llvm-commits mailing list