[PATCH] D108757: [X86][Codegen] PR51615: don't replace wide volatile load with narrow broadcast-from-memory

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.cpp:5047
+  return !Ld->isVolatile() ||
+         Ld->getValueSizeInBits(0) == EltVT.getScalarSizeInBits();
+}
----------------
I'm not sure if the semantics on this are settled - what happens if the overall size matches, but the "shape" isn't the same?
I'm imagining something like:
  %i = load volatile <2 x float>, <2 x float>* @test5_id12345, align 16
  %b = bitcast <2 x float> %i to <1 x double>
  %i1 = shufflevector <1 x double> %b, <1 x double> poison, <4 x i32> <i32 undef, i32 0, i32 undef, i32 0>


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108757/new/

https://reviews.llvm.org/D108757