[Mlir-commits] [mlir] Introduce new Unroll And Jam loop transform for SCF/Affine loops (PR #94142)

[Oleksandr Alex Zinenko]

================
@@ -0,0 +1,56 @@
+//===- LoopUtils.h -  LoopUtils Support ---------------------*- C++
+//-*-=============//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains definitions for the action framework. This framework
+// allows for external entities to control certain actions taken by the compiler
+// by registering handler functions.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_IR_LOOP_UTILS_H
+#define MLIR_IR_LOOP_UTILS_H
+
+#include "mlir/IR/BuiltinOps.h"
+
+namespace mlir {
+
+// Gathers all maximal sub-blocks of operations that do not themselves
+// include a `OpTy` (an operation could have a descendant `OpTy` though
+// in its tree). Ignore the block terminators.
+template <typename OpTy>
+struct JamBlockGatherer {
+  // Store iterators to the first and last op of each sub-block found.
+  llvm::SmallVector<std::pair<Block::iterator, Block::iterator>> subBlocks;
+
+  // This is a linear time walk.
+  void walk(Operation *op) {
+    for (auto &region : op->getRegions())
----------------
ftynse wrote:

Please expand `auto` unless the RHS is a case or the type is obnoxious to spell.

https://github.com/llvm/llvm-project/pull/94142