[clang] [llvm] [LLVM][SROA] Teach SROA how to "bitcast" between fixed and scalable vectors. (PR #130973)

[Paul Walker via cfe-commits]

================
@@ -554,6 +554,22 @@ class VectorType : public Type {
     return VectorType::get(VTy->getElementType(), EltCnt * 2);
   }
 
+  /// This static method returns a VectorType with the same size-in-bits as
+  /// SizeTy but with an element type that matches the scalar type of EltTy.
+  static VectorType *getWithSizeAndScalar(VectorType *SizeTy, Type *EltTy) {
+    if (SizeTy->getScalarType() == EltTy->getScalarType())
+      return SizeTy;
+
+    unsigned EltSize = EltTy->getScalarSizeInBits();
+    if (!SizeTy->getPrimitiveSizeInBits().isKnownMultipleOf(EltSize))
----------------
paulwalker-arm wrote:

Are you sure? The intent of this function is to create a VectorType that can be bitcasted, typically from/to `SizeTy`. When looking at `CastInst::castIsValid()` I can see it uses `getPrimitiveSizeInBits()` to determine validity.

To put another way, the check is ensuring the IR is consistent, essentially protecting against the case where the ElementCount calculation would be invalid. So I don't think this function cares about data layout?


https://github.com/llvm/llvm-project/pull/130973