[PATCH] D33107: AMDGPU/SI: Don't promote to vector if the load/store is volatile

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp:405
   case Instruction::BitCast:
   case Instruction::AddrSpaceCast:
     return true;
----------------
As a separate patch this should reject addrspacecasts


================
Comment at: lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp:410-412
+    if (SI->isVolatile())
+      return false;
     return SI->getPointerOperand() == User;
----------------
Can combine into return condition


https://reviews.llvm.org/D33107