[Mlir-commits] [mlir] [mlir][linalg][nfc] Fix `linalg.matmul_transpose_a` def. (PR #97690)

[Oleksandr Alex Zinenko]

https://github.com/ftynse commented:

Looks reasonable, could we see the result of converting this to `linalg.generic` to double-check?  The fact that the K dimension is leading in `shape_map` is rather confusing, though I do think it's just the order in which dimensions appear rather than the order of loop iterators.

https://github.com/llvm/llvm-project/pull/97690