[Mlir-commits] [mlir] [MLIR][XeGPU] Optimize instruction granularity for elemwise (PR #220986)

[Jianhui Li]

Jianhui-Li wrote:

The motivation may be mitigated once multi-reduction increases the vector-size.  Please hold this PR for now. 
BTW: We may still need this optimization (a neat idea!). Implemenation-wise I am not sure that it is part of ResolveLayoutConflicts().

https://github.com/llvm/llvm-project/pull/220986