[PATCH] D152440: [AMDGPU] Trim trailing undefs from the end of image and buffer store

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp:387-388
 
+  if (!isa<Constant>(UseV) && !isa<Instruction>(UseV))
+    return DemandedElts;
+
----------------
I don't think there's any need for this. Almost everything is an Instruction or a Constant, and anything that is not will just break out of the loop on the first iteration anyway.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp:394
+    if (auto *ConstElt = dyn_cast<Constant>(Elt)) {
+      if (!ConstElt->isZeroValue() && !isa<UndefValue>(Elt))
+        break;
----------------
Should use isNullValue because we don't want to match -0.0.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152440/new/

https://reviews.llvm.org/D152440