[Mlir-commits] [mlir] [MLIR][NVVM] Add S2G and Reduce override NVVM Dialect ops (PR #216481)

[Durgadoss R]

================
@@ -146,6 +172,24 @@ LogicalResult CpAsyncBulkTensorSharedCTAToGlobalOp::verify() {
   return success();
 }
 
+LogicalResult CpAsyncBulkTensorSharedCTAToGlobalOverrideAddrOp::verify() {
+  TMAStoreMode mode = getMode();
+  bool isIm2Col = mode == TMAStoreMode::IM2COL;
+  bool isTile = mode == TMAStoreMode::TILE;
+
+  LogicalResult commonRes = cpAsyncBulkTensorCommonVerifier(
+      getCoordinates().size(), isIm2Col, 0, getLoc());
+
+  LogicalResult overrideAddrRes = CpAsyncBulkTensorOverrideAddrCommonVerifier(
+      getCoordinates(), getTensorSize(), getLowerStride(), getUpperStride(),
+      isTile, getLoc());
+
+  if (mode == TMAStoreMode::TILE_SCATTER4 && getCoordinates().size() != 5)
+    overrideAddrRes = emitError("Mode tile scatter4 expects 5 coordinates");
----------------
durga4github wrote:

That's fine too

https://github.com/llvm/llvm-project/pull/216481