[PATCH] D102404: [InstCombine] Add instcombine fold for extractelement + splat for scalable vectors

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/IR/ConstantFold.cpp:875-879
+  if (Constant *SplatVal = Val->getSplatValue()) {
+    unsigned NumElts = ValVTy->getElementCount().getKnownMinValue();
+    if (CIdx->getZExtValue() < NumElts)
+      return SplatVal;
+  }
----------------
It looks like this handles some of the same cases as the "CAZ of ScalableVectorType" case below. Can you combine them?


================
Comment at: llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp:350
 
+    if (auto *I = dyn_cast<Instruction>(SrcVec))
+      if (auto *SE = dyn_cast<ShuffleVectorInst>(I))
----------------
This needs a comment explaining what case you are trying to handle, e.g. "(extractelement (shufflevector ...), ...) -> (...)"





================
Comment at: llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp:351
+    if (auto *I = dyn_cast<Instruction>(SrcVec))
+      if (auto *SE = dyn_cast<ShuffleVectorInst>(I))
+        if (isSplatValue(SE) && IndexC->getZExtValue() < NumElts)
----------------
You can dyn_cast SrcVec directly to ShuffleVectorInst, no need to go via Instruction.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102404/new/

https://reviews.llvm.org/D102404