[clang] [MLIR][NVGPU] Introduce `nvgpu.wargroup.mma.store` Op for Hopper GPUs (PR #65441)

[via cfe-commits]

================
@@ -529,6 +530,39 @@ LogicalResult WarpgroupMmaOp::verify() {
   return success();
 }
 
+LogicalResult WarpgroupMmaStoreOp::verify() {
+  MemRefType dstMemrefType = getDstMemref().getType();
+  VectorType firstVtype = getMatrixD()
+                              .front()
+                              .getType()
+                              .cast<WarpgroupAccumulatorType>()
+                              .getFragmented();
+
+  int64_t totalFirstDimension = 0;
+  for (auto result : getMatrixD()) {
----------------
qcolombet wrote:

We still have the `auto` keyword here. I'd like to see the explicit type.

https://github.com/llvm/llvm-project/pull/65441