1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
|
//===- ConvertToDestinationStyle.cpp - Convert non-DPS to DPS ops ---------===//
//
// 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 patterns to convert non-DPS ops to DPS ops. New
// tensor.empty ops are inserted as a destination. Such tensor.empty can be
// eliminated with "empty tensor elimination", allowing them to bufferize
// without an allocation (assuming there are no further conflicts).
//
//===----------------------------------------------------------------------===//
//
#include "mlir/Dialect/Arith/IR/Arith.h"
#include "mlir/Dialect/Linalg/IR/Linalg.h"
#include "mlir/Dialect/Linalg/Transforms/Transforms.h"
#include "mlir/Dialect/Tensor/IR/Tensor.h"
#include "mlir/IR/Matchers.h"
#include "mlir/IR/PatternMatch.h"
#include "llvm/Support/Debug.h"
using namespace mlir;
using namespace mlir::tensor;
namespace {
/// Lower tensor.generate to linalg.generic.
struct GenerateOpConverter : public OpRewritePattern<GenerateOp> {
using OpRewritePattern<GenerateOp>::OpRewritePattern;
LogicalResult matchAndRewrite(GenerateOp generateOp,
PatternRewriter &rewriter) const override {
// Only ops with exactly one block are supported.
if (!generateOp.getBody().hasOneBlock())
return failure();
Location loc = generateOp.getLoc();
RankedTensorType tensorType = generateOp.getType().cast<RankedTensorType>();
// Create tensor.empty.
auto emptyOp = rewriter.create<EmptyOp>(loc, tensorType,
generateOp.getDynamicExtents());
// Create linalg.generic.
SmallVector<utils::IteratorType> iteratorTypes(
tensorType.getRank(), utils::IteratorType::parallel);
SmallVector<AffineMap> indexingMaps(
1, rewriter.getMultiDimIdentityMap(tensorType.getRank()));
auto genericOp = rewriter.create<linalg::GenericOp>(
loc, tensorType, /*inputs=*/ValueRange(),
/*outputs=*/ValueRange{emptyOp.getResult()}, /*indexingMaps=*/
indexingMaps, iteratorTypes);
Block *body = rewriter.createBlock(&genericOp->getRegion(0), {},
tensorType.getElementType(), loc);
rewriter.setInsertionPointToStart(body);
SmallVector<Value> bbArgReplacements;
for (int64_t i = 0; i < tensorType.getRank(); ++i)
bbArgReplacements.push_back(rewriter.create<linalg::IndexOp>(loc, i));
rewriter.mergeBlocks(&generateOp.getBody().front(), body,
bbArgReplacements);
// Update terminator.
auto yieldOp = cast<tensor::YieldOp>(body->getTerminator());
rewriter.replaceOpWithNewOp<linalg::YieldOp>(yieldOp, yieldOp.getValue());
// Replace tensor.generate.
rewriter.replaceOp(generateOp, genericOp->getResult(0));
return success();
}
};
/// Lower tensor.pad to linalg.generic + tensor.insert_slice.
struct PadOpConverter : public OpRewritePattern<PadOp> {
using OpRewritePattern<PadOp>::OpRewritePattern;
LogicalResult matchAndRewrite(PadOp padOp,
PatternRewriter &rewriter) const override {
// Only ops with exactly one block are supported.
if (!padOp.getBodyRegion().hasOneBlock())
return failure();
// Create tensor.empty.
Location loc = padOp.getLoc();
RankedTensorType resultType = padOp.getResultType();
ReifiedRankedShapedTypeDims reifiedShape;
if (failed(cast<ReifyRankedShapedTypeOpInterface>(padOp.getOperation())
.reifyResultShapes(rewriter, reifiedShape)))
return rewriter.notifyMatchFailure(
padOp, "failed to reify tensor.pad op result shape");
SmallVector<Value> dynamicSizes;
for (int64_t i = 0; i < resultType.getRank(); ++i)
if (resultType.isDynamicDim(i))
dynamicSizes.push_back(reifiedShape[0][i]);
auto emptyOp = rewriter.create<EmptyOp>(loc, resultType, dynamicSizes);
// Examine the yielded value to decide if a linalg.generic is neede or a
// linalg.fill is sufficient.
Value filled;
Value yieldedValue =
cast<tensor::YieldOp>(padOp.getBody()->getTerminator()).getValue();
Attribute constYieldedValue;
// Is the yielded value a bbArg defined outside of the PadOp?
bool outsideBbArg =
yieldedValue.isa<BlockArgument>() &&
yieldedValue.cast<BlockArgument>().getOwner()->getParentOp() !=
padOp.getOperation();
// Is the yielded value an OpResult defined outside of the PadOp?
bool outsideOpResult =
yieldedValue.isa<OpResult>() &&
yieldedValue.getDefiningOp()->getParentOp() != padOp.getOperation();
bool invariantYieldedValue = outsideBbArg || outsideOpResult;
if (matchPattern(yieldedValue, m_Constant(&constYieldedValue))) {
// Padding with a constant: Create linalg.fill.
Dialect *arithDialect =
rewriter.getContext()->getLoadedDialect<arith::ArithDialect>();
Value fillValue = arithDialect
->materializeConstant(rewriter, constYieldedValue,
yieldedValue.getType(),
yieldedValue.getLoc())
->getResult(0);
auto fillOp = rewriter.create<linalg::FillOp>(
loc, ValueRange(fillValue), ValueRange(emptyOp.getResult()));
rewriter.setInsertionPointAfter(fillOp);
filled = fillOp.getResult(0);
} else if (invariantYieldedValue) {
// Padding with an invariant value.
auto fillOp = rewriter.create<linalg::FillOp>(
loc, ValueRange(yieldedValue), ValueRange(emptyOp.getResult()));
rewriter.setInsertionPointAfter(fillOp);
filled = fillOp.getResult(0);
} else {
// Create linalg.generic.
SmallVector<utils::IteratorType> iteratorTypes(
resultType.getRank(), utils::IteratorType::parallel);
SmallVector<AffineMap> indexingMaps(
1, rewriter.getMultiDimIdentityMap(resultType.getRank()));
auto genericOp = rewriter.create<linalg::GenericOp>(
loc, resultType, /*inputs=*/ValueRange(),
/*outputs=*/ValueRange{emptyOp.getResult()}, /*indexingMaps=*/
indexingMaps, iteratorTypes);
Block *body = rewriter.createBlock(&genericOp->getRegion(0), {},
resultType.getElementType(), loc);
rewriter.setInsertionPointToStart(body);
SmallVector<Value> bbArgReplacements;
for (int64_t i = 0; i < resultType.getRank(); ++i)
bbArgReplacements.push_back(rewriter.create<linalg::IndexOp>(loc, i));
rewriter.mergeBlocks(padOp.getBody(), body, bbArgReplacements);
// Update terminator.
auto yieldOp = cast<tensor::YieldOp>(body->getTerminator());
rewriter.replaceOpWithNewOp<linalg::YieldOp>(yieldOp, yieldOp.getValue());
rewriter.setInsertionPointAfter(genericOp);
filled = genericOp->getResult(0);
}
// Create tensor::InsertSliceOp.
SmallVector<OpFoldResult> sliceSizes =
getMixedSizes(rewriter, loc, padOp.getSource());
SmallVector<OpFoldResult> sliceStrides(resultType.getRank(),
rewriter.getIndexAttr(1));
rewriter.replaceOpWithNewOp<tensor::InsertSliceOp>(
padOp, padOp.getSource(), filled,
/*offsets=*/padOp.getMixedLowPad(), sliceSizes, sliceStrides);
return success();
}
};
} // namespace
void linalg::populateConvertToDestinationStylePatterns(
RewritePatternSet &patterns) {
patterns.insert<GenerateOpConverter, PadOpConverter>(patterns.getContext());
}
|
