[Mlir-commits] [mlir] [mlir][vector] don't emit non-rank 1 masked load and store (PR #67656)

[Diego Caballero]

dcaballe wrote:

This is probably a restriction we want to remove as `vector.load/store` can deal with n-D vectors. This restriction comes from when some of the vector ops were a carbon copy of the LLVM ones because we couldn't use the LLVM dialect directly but that's no longer the case (similar to the vector.extract/vector.extractelement situation).

https://github.com/llvm/llvm-project/pull/67656