[llvm] [mlir] [AMDGPU][MLIR]Add shmem-optimization as an op using transform dialect (PR #81550)

[via llvm-commits]

================
@@ -0,0 +1,46 @@
+//===- AMDGPUTransformOps.td - AMDGPU transform ops ----------*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef AMDGPU_TRANSFORM_OPS
+#define AMDGPU_TRANSFORM_OPS
+
+include "mlir/Dialect/Transform/IR/TransformAttrs.td"
+include "mlir/Dialect/Transform/IR/TransformDialect.td"
+include "mlir/Dialect/Transform/IR/TransformInterfaces.td"
+include "mlir/Dialect/Transform/IR/TransformTypes.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
+
+//===----------------------------------------------------------------------===//
+// ApplyOptimizeSharedMemoryReadsAndWritesOp
+//===----------------------------------------------------------------------===//
+
+def ApplyOptimizeSharedMemoryReadsAndWritesOp :
+  Op<Transform_Dialect, "amdgpu.optimize_shared_memory_reads_and_writes",
+    [DeclareOpInterfaceMethods<MemoryEffectsOpInterface>,
+     TransformOpInterface, TransformEachOpTrait]> {
+  let summary = "Reduce shared memory bank conflicts";
+  let description = [{ This opp adds a transformation and pass to the AMDGPU 
----------------
harsh-nod wrote:

typo: op. Also,  the description should be what the op does which is optimize reads/writes. I would remove the transformation and pass to AMDGPU dialect. So something like "This op attempts to optimize GPU shared memory reads/writes with the goal of avoiding bank conflicts."

https://github.com/llvm/llvm-project/pull/81550