[Mlir-commits] [mlir] Revert "[mlir][vector] Move transpose with unit-dim to shape_cast pattern (#72493)" (PR #72918)

[Andrzej Warzyński]

banach-space wrote:

> Sorry, maybe I pushed the button too hastily. 

No worries, we did agree to have this reverted. I just wanted to make sure that we use this opportunity to improve test coverage in MLIR - ideally there would be a test that would prevent us from landing this in the first place.

> Both are judgment calls and a matter of design.

I think that there's also a friction between "canonical for CPUs" and "canonical for GPUs" - currently there is no mechanism to distinguish between the two.

This has been a very useful discussion - I didn't realise that things would be so different for GPUs. Apologies for missing that.

https://github.com/llvm/llvm-project/pull/72918