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

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp:398
                             &IC.getDominatorTree());
-    if (KnownFPClass.KnownFPClasses != fcPosZero)
+    if (KnownFPClass.KnownFPClasses != fcPosZero && !isa<UndefValue>(Elt))
       break;
----------------
You can skip the whole computeKnownFPClass call in this case


================
Comment at: llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-simplify-image-buffer-stores.ll:93
+;
+  %newvdata1 = insertelement <4 x float> undef, float %vdata1, i32 0
+  %newvdata2 = insertelement <4 x float> %newvdata1, float 1.0, i32 1
----------------
Should use poison in place of undef


Repository:
  rG LLVM Github Monorepo

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

https://reviews.llvm.org/D152440