[Mlir-commits] [mlir] [MLIR][XeGPU] Add distribution pattern for xegpu.load & store for sg to wi pass (PR #181917)

[Jianhui Li]

================
@@ -395,6 +395,71 @@ struct SgToWiPrefetchNd : public OpConversionPattern<xegpu::PrefetchNdOp> {
   }
 };
 
+/// Distributes a subgroup-level LoadGather (xegpu.load) op to workitem-level.
+struct SgToWiLoadGather : public OpConversionPattern<xegpu::LoadGatherOp> {
----------------
Jianhui-Li wrote:

separate to this PR:  need to keep consistent to use "lane" instead of "work-item" or "wi".  

https://github.com/llvm/llvm-project/pull/181917