[clang] [llvm] [LV] Mask off possibly aliasing vector lanes (PR #100579)

[Graham Hunter via cfe-commits]

================
@@ -2778,6 +2808,60 @@ void VPWidenPointerInductionRecipe::print(raw_ostream &O, const Twine &Indent,
 }
 #endif
 
+void VPAliasLaneMaskRecipe::execute(VPTransformState &State) {
+  IRBuilderBase Builder = State.Builder;
+  Value *SinkValue = State.get(getSinkValue(), 0, true);
+  Value *SourceValue = State.get(getSourceValue(), 0, true);
+
+  unsigned ElementSize = 0;
+  auto *ReadInsn = cast<Instruction>(SourceValue);
+  auto *ReadCast = dyn_cast<CastInst>(SourceValue);
+  if (ReadInsn->getOpcode() == Instruction::Add)
+    ReadCast = dyn_cast<CastInst>(ReadInsn->getOperand(0));
+
+  if (ReadCast && ReadCast->getOpcode() == Instruction::PtrToInt) {
+    Value *Ptr = ReadCast->getOperand(0);
+    for (auto *Use : Ptr->users()) {
+      if (auto *GEP = dyn_cast<GetElementPtrInst>(Use)) {
+        auto *EltVT = GEP->getSourceElementType();
+        if (EltVT->isArrayTy())
+          ElementSize = EltVT->getArrayElementType()->getScalarSizeInBits() *
+                        EltVT->getArrayNumElements();
+        else
+          ElementSize = GEP->getSourceElementType()->getScalarSizeInBits() / 8;
+        break;
+      }
+    }
+  }
+  assert(ElementSize > 0 && "Couldn't get element size from pointer");
----------------
huntergr-arm wrote:

This assert fires when compiling `MicroBenchmarks/ImageProcessing/Dilate/CMakeFiles/Dilate.dir/dilateKernel.c` from the test suite, so that needs fixing at least.

But I think we shouldn't be trying to figure out the element size at recipe execution time; it should be determined when trying to build the recipe and the value stored as part of the class. If we can't find the size then we stop trying to build the recipe and fall back to normal RT checks. Does that sound sensible to you?

https://github.com/llvm/llvm-project/pull/100579