[Mlir-commits] [mlir] [mlir][tensor] Update `GeneralizeOuterUnitDimsPackOpPattern` (PR #115312)

[Andrzej Warzyński]

banach-space wrote:

> It is not a restriction in our use cases because we can data-tile any ops that implements ContractionOpInterface. E.g., a contraction generic op which has batch dimension being the innermost dimension. In this case, we don't data-tile batch dimension.
> 
> These patterns were built for the first take of pack/unpack vectorization, and they are no longer used on IREE CPU x86 codegen path. Because today we have direct vectorization and masking supports.

Thanks for the context and for the explanation! Any preference how to proceed here?

>From my perspective, without tests in-tree, it tends to be tricky to produce good ref examples. So, if I can simplify things without breaking any tests, I'd go with that.

https://github.com/llvm/llvm-project/pull/115312