[Mlir-commits] [mlir] [mlir][Linalg] Refine how broadcast dims are treated (PR #99015)

[Andrzej Warzyński]

banach-space wrote:

> I would also revisit the commit description as this seems to be a reduction problem, not a broadcast problem? Or maybe both broadcast and reduction?

Both :) We already test: 
* [reductions](https://github.com/llvm/llvm-project/blob/2ba3fe7356f065757a2279f65e4ef5c8f1476293/mlir/test/Dialect/Linalg/vectorization.mlir#L229-L265), and 
* [broadcasts](https://github.com/llvm/llvm-project/blob/2ba3fe7356f065757a2279f65e4ef5c8f1476293/mlir/test/Dialect/Linalg/vectorization.mlir#L114-L128). 

However, there's no single test that would cover both.

https://github.com/llvm/llvm-project/pull/99015