[Mlir-commits] [mlir] [MLIR][XeGPU] Allow load/store/prefetch uses [memref+offset] instead of tdesc (PR #150576)

[Jianhui Li]

================
@@ -546,7 +547,8 @@ struct UnrollPrefetchOp : public UnrollPattern<xegpu::PrefetchOp> {
     Location loc = op.getLoc();
     xegpu::TensorDescType tdescTy = op.getTensorDescType();
 
-    if (!tdescTy.isScattered())
+    // TODO: handle the unstructure source case (!tdesTy)
+    if (!tdescTy || !tdescTy.isScattered())
----------------
Jianhui-Li wrote:

Layout propagation works the same way.  no need to check

https://github.com/llvm/llvm-project/pull/150576