[Mlir-commits] [mlir] [MLIR][XeGPU] Lower arith.truncf wider than one xevm.truncf group (PR #217130)

[Artem Kroviakov]

https://github.com/akroviakov approved this pull request.

Functionally looks fine.
Design-wise, can `arith-expand-ops` be called before inst_data propagation if we know that the pass introduces shapes that are subject to target restrictions?

https://github.com/llvm/llvm-project/pull/217130