[Mlir-commits] [mlir] [mlir][linalg] Produce canonical linalg.generic for im2col (PR #134675)

[Quinn Dawkins]

qedawkins wrote:

> Any thoughts on this @qedawkins @Groverkss ? Is there any specific pass or test that I can run to check that this wouldn't break the current pipelines relying on tile-and-fuse?

I'm not aware of anyone relying on these patterns (we aren't downstream) so I can't point you at a pipeline. Since it's always easier to go from indexing_map -> explicit indexing, this PR is fine to land. I would just caution against complex indexing maps as most transformations don't/can't support them.

https://github.com/llvm/llvm-project/pull/134675