[Mlir-commits] [mlir] [mlir][AMDGPU] Add canonicalization pattern to pack scales for ScaledMFMAOp (PR #155951)

[Krzysztof Drewniak]

krzysz00 wrote:

... Oh, and while I'm here, the `setOperand()`s and such should be wrapped in a `rewriter.modifyOpInplace()`

https://github.com/llvm/llvm-project/pull/155951