[Mlir-commits] [mlir] [MLIR][XeGPU] Add XeGPU scattered ops (PR #86594)

[Chao Chen]

================
@@ -114,6 +149,29 @@ LogicalResult CreateNdDescOp::verify() {
     return emitOpError("TensorDesc should have the same element "
                        "type with the source if it is a memref.\n");
 
+  if (getType().getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
+
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_PrefetchNdOp
+//===----------------------------------------------------------------------===//
+LogicalResult PrefetchNdOp::verify() {
+  auto tdescTy = getTensorDescType();
+  if (tdescTy.getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
+
+  if (!isReadHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isReadHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isReadHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
----------------
chencha3 wrote:

correct case has been in current test cases. I added some invalid cases for them in new commits. 

https://github.com/llvm/llvm-project/pull/86594