1 //===- LoopUnrollAndJam.cpp - Code to perform loop unroll and jam ---------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements loop unroll and jam. Unroll and jam is a transformation
10 // that improves locality, in particular, register reuse, while also improving
11 // operation level parallelism. The example below shows what it does in nearly
12 // the general case. Loop unroll and jam currently works if the bounds of the
13 // loops inner to the loop being unroll-jammed do not depend on the latter.
14 //
15 // Before After unroll and jam of i by factor 2:
16 //
17 // for i, step = 2
18 // for i S1(i);
19 // S1; S2(i);
20 // S2; S1(i+1);
21 // for j S2(i+1);
22 // S3; for j
23 // S4; S3(i, j);
24 // S5; S4(i, j);
25 // S6; S3(i+1, j)
26 // S4(i+1, j)
27 // S5(i);
28 // S6(i);
29 // S5(i+1);
30 // S6(i+1);
31 //
32 // Note: 'if/else' blocks are not jammed. So, if there are loops inside if
33 // op's, bodies of those loops will not be jammed.
34 //===----------------------------------------------------------------------===//
35
36 #include "PassDetail.h"
37 #include "mlir/Analysis/LoopAnalysis.h"
38 #include "mlir/Dialect/Affine/IR/AffineOps.h"
39 #include "mlir/Dialect/Affine/Passes.h"
40 #include "mlir/IR/AffineExpr.h"
41 #include "mlir/IR/AffineMap.h"
42 #include "mlir/IR/BlockAndValueMapping.h"
43 #include "mlir/IR/Builders.h"
44 #include "mlir/Transforms/LoopUtils.h"
45 #include "llvm/ADT/DenseMap.h"
46 #include "llvm/Support/CommandLine.h"
47
48 using namespace mlir;
49
50 #define DEBUG_TYPE "affine-loop-unroll-jam"
51
52 namespace {
53 /// Loop unroll jam pass. Currently, this just unroll jams the first
54 /// outer loop in a Function.
55 struct LoopUnrollAndJam : public AffineLoopUnrollAndJamBase<LoopUnrollAndJam> {
LoopUnrollAndJam__anon96b7db9c0111::LoopUnrollAndJam56 explicit LoopUnrollAndJam(Optional<unsigned> unrollJamFactor = None) {
57 if (unrollJamFactor)
58 this->unrollJamFactor = *unrollJamFactor;
59 }
60
61 void runOnFunction() override;
62 };
63 } // end anonymous namespace
64
65 std::unique_ptr<OperationPass<FuncOp>>
createLoopUnrollAndJamPass(int unrollJamFactor)66 mlir::createLoopUnrollAndJamPass(int unrollJamFactor) {
67 return std::make_unique<LoopUnrollAndJam>(
68 unrollJamFactor == -1 ? None : Optional<unsigned>(unrollJamFactor));
69 }
70
runOnFunction()71 void LoopUnrollAndJam::runOnFunction() {
72 // Currently, just the outermost loop from the first loop nest is
73 // unroll-and-jammed by this pass. However, runOnAffineForOp can be called on
74 // any for operation.
75 auto &entryBlock = getFunction().front();
76 if (auto forOp = dyn_cast<AffineForOp>(entryBlock.front()))
77 loopUnrollJamByFactor(forOp, unrollJamFactor);
78 }
79