[Mlir-commits] [mlir] [mlir][ArmSME] Remove `ConvertIllegalShapeCastOpsToTransposes` (PR #139706)

[Benjamin Maxwell]

MacDue wrote:

I think you could just update the `LiftIllegalVectorTransposeToMemory` rewrite to match `vector.shape_cast` for this case (rather than `vector.transpose`). That should lead to less duplication as the existing lowerings for `memref.transpose` and the vector transfer operations would introduce the necessary loops.

https://github.com/llvm/llvm-project/pull/139706