[Mlir-commits] [mlir] [mlir][vector] Verify non-unit strides on `masked/expand/compress` ops (PR #210952)

[Federico Bruzzone]

FedericoBruzzone wrote:


> > A dynamic stride is still accepted, unlike vector.load/vector.store (tests added to document the asymmetry.
> 
> Why is this exception required?

Thanks for the comment @banach-space!  It gave me the chance to double-check.

The issue I had encountered was with the Sparse Vectorizer. That problem no longer arises.
This means we've managed to align the table (I’ve updated it, as well as the comment in the PR).
There's good code reuse now :D

https://github.com/llvm/llvm-project/pull/210952