[Mlir-commits] [mlir] [MLIR][XeGPU] Add blocking and subgroup to lane distribution support for ConvertLayout operation (PR #183837)

[Jianhui Li]

================
@@ -1539,9 +1539,7 @@ def XeGPU_ConvertLayoutOp: XeGPU_Op<"convert_layout", [Pure, AllTypesMatch<["sou
 
     }];
 
-    let hasFolder = 1;
     let hasVerifier = 1;
-    let hasCanonicalizer = 1;
 }
----------------
Jianhui-Li wrote:

I don't find a use case. If input and target layout are same, there should not be inserted to the IR to begin with. 

https://github.com/llvm/llvm-project/pull/183837