[Mlir-commits] [mlir] [MLIR][XeGPU] Add uArch limitation to scatter load store (PR #172845)

[Jianhui Li]

================
@@ -1220,7 +1264,8 @@ static LogicalResult updateOp(mlir::OpBuilder &builder, mlir::Operation *op,
     }
     // If the result is a vector type, add a temporary layout attribute to the
     // op.
-    xegpu::setDistributeLayoutAttr(result, layout);
+    if (!isa<xegpu::LoadGatherOp>(op))
----------------
Jianhui-Li wrote:

should skip all the anchor op here. There is no need to assign temp layout to anchor op's result. 

https://github.com/llvm/llvm-project/pull/172845