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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
|
//===- BufferizableOpInterfaceImpl.cpp - Impl. of BufferizableOpInterface -===//
//
// 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
//
//===----------------------------------------------------------------------===//
#include "mlir/Dialect/Arith/Transforms/BufferizableOpInterfaceImpl.h"
#include "mlir/Dialect/Arith/IR/Arith.h"
#include "mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h"
#include "mlir/Dialect/Bufferization/Transforms/BufferUtils.h"
#include "mlir/Dialect/MemRef/IR/MemRef.h"
#include "mlir/IR/Attributes.h"
#include "mlir/IR/Dialect.h"
#include "mlir/IR/Operation.h"
using namespace mlir;
using namespace mlir::bufferization;
namespace {
/// Bufferization of arith.constant. Replace with memref.get_global.
struct ConstantOpInterface
: public BufferizableOpInterface::ExternalModel<ConstantOpInterface,
arith::ConstantOp> {
LogicalResult bufferize(Operation *op, RewriterBase &rewriter,
const BufferizationOptions &options) const {
auto constantOp = cast<arith::ConstantOp>(op);
Attribute memorySpace;
if (options.defaultMemorySpace.has_value())
memorySpace = *options.defaultMemorySpace;
else
return constantOp->emitError("could not infer memory space");
// Only ranked tensors are supported.
if (!isa<RankedTensorType>(constantOp.getType()))
return failure();
// Only constants inside a module are supported.
auto moduleOp = constantOp->getParentOfType<ModuleOp>();
if (!moduleOp)
return failure();
// Create global memory segment and replace tensor with memref pointing to
// that memory segment.
FailureOr<memref::GlobalOp> globalOp =
getGlobalFor(constantOp, options.bufferAlignment, memorySpace);
if (failed(globalOp))
return failure();
memref::GlobalOp globalMemref = *globalOp;
replaceOpWithNewBufferizedOp<memref::GetGlobalOp>(
rewriter, op, globalMemref.getType(), globalMemref.getName());
return success();
}
bool isWritable(Operation *op, Value value,
const AnalysisState &state) const {
// Memory locations returned by memref::GetGlobalOp may not be written to.
assert(isa<OpResult>(value));
return false;
}
};
struct IndexCastOpInterface
: public BufferizableOpInterface::ExternalModel<IndexCastOpInterface,
arith::IndexCastOp> {
bool bufferizesToMemoryRead(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return false;
}
bool bufferizesToMemoryWrite(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return false;
}
AliasingOpResultList getAliasingOpResults(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return {{op->getResult(0), BufferRelation::Equivalent}};
}
LogicalResult bufferize(Operation *op, RewriterBase &rewriter,
const BufferizationOptions &options) const {
auto castOp = cast<arith::IndexCastOp>(op);
auto resultTensorType = cast<TensorType>(castOp.getType());
FailureOr<Value> source = getBuffer(rewriter, castOp.getIn(), options);
if (failed(source))
return failure();
auto sourceType = cast<BaseMemRefType>(source->getType());
// Result type should have same layout and address space as the source type.
BaseMemRefType resultType;
if (auto rankedMemRefType = dyn_cast<MemRefType>(sourceType)) {
resultType = MemRefType::get(
rankedMemRefType.getShape(), resultTensorType.getElementType(),
rankedMemRefType.getLayout(), rankedMemRefType.getMemorySpace());
} else {
auto unrankedMemrefType = cast<UnrankedMemRefType>(sourceType);
resultType = UnrankedMemRefType::get(resultTensorType.getElementType(),
unrankedMemrefType.getMemorySpace());
}
replaceOpWithNewBufferizedOp<arith::IndexCastOp>(rewriter, op, resultType,
*source);
return success();
}
};
/// Bufferization of arith.select. Just replace the operands.
struct SelectOpInterface
: public BufferizableOpInterface::ExternalModel<SelectOpInterface,
arith::SelectOp> {
bool bufferizesToMemoryRead(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return false;
}
bool bufferizesToMemoryWrite(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return false;
}
AliasingOpResultList getAliasingOpResults(Operation *op, OpOperand &opOperand,
const AnalysisState &state) const {
return {{op->getOpResult(0) /*result*/, BufferRelation::Equivalent,
/*isDefinite=*/false}};
}
LogicalResult bufferize(Operation *op, RewriterBase &rewriter,
const BufferizationOptions &options) const {
auto selectOp = cast<arith::SelectOp>(op);
Location loc = selectOp.getLoc();
// TODO: It would be more efficient to copy the result of the `select` op
// instead of its OpOperands. In the worst case, 2 copies are inserted at
// the moment (one for each tensor). When copying the op result, only one
// copy would be needed.
FailureOr<Value> maybeTrueBuffer =
getBuffer(rewriter, selectOp.getTrueValue(), options);
FailureOr<Value> maybeFalseBuffer =
getBuffer(rewriter, selectOp.getFalseValue(), options);
if (failed(maybeTrueBuffer) || failed(maybeFalseBuffer))
return failure();
Value trueBuffer = *maybeTrueBuffer;
Value falseBuffer = *maybeFalseBuffer;
// The "true" and the "false" operands must have the same type. If the
// buffers have different types, they differ only in their layout map. Cast
// both of them to the most dynamic MemRef type.
if (trueBuffer.getType() != falseBuffer.getType()) {
auto targetType =
bufferization::getBufferType(selectOp.getResult(), options);
if (failed(targetType))
return failure();
if (trueBuffer.getType() != *targetType)
trueBuffer =
rewriter.create<memref::CastOp>(loc, *targetType, trueBuffer);
if (falseBuffer.getType() != *targetType)
falseBuffer =
rewriter.create<memref::CastOp>(loc, *targetType, falseBuffer);
}
replaceOpWithNewBufferizedOp<arith::SelectOp>(
rewriter, op, selectOp.getCondition(), trueBuffer, falseBuffer);
return success();
}
FailureOr<BaseMemRefType>
getBufferType(Operation *op, Value value, const BufferizationOptions &options,
const DenseMap<Value, BaseMemRefType> &fixedTypes) const {
auto selectOp = cast<arith::SelectOp>(op);
assert(value == selectOp.getResult() && "invalid value");
auto trueType = bufferization::getBufferType(selectOp.getTrueValue(),
options, fixedTypes);
auto falseType = bufferization::getBufferType(selectOp.getFalseValue(),
options, fixedTypes);
if (failed(trueType) || failed(falseType))
return failure();
if (*trueType == *falseType)
return *trueType;
if (trueType->getMemorySpace() != falseType->getMemorySpace())
return op->emitError("inconsistent memory space on true/false operands");
// If the buffers have different types, they differ only in their layout
// map.
auto memrefType = llvm::cast<MemRefType>(*trueType);
return getMemRefTypeWithFullyDynamicLayout(
RankedTensorType::get(memrefType.getShape(),
memrefType.getElementType()),
memrefType.getMemorySpace());
}
};
} // namespace
void mlir::arith::registerBufferizableOpInterfaceExternalModels(
DialectRegistry ®istry) {
registry.addExtension(+[](MLIRContext *ctx, ArithDialect *dialect) {
ConstantOp::attachInterface<ConstantOpInterface>(*ctx);
IndexCastOp::attachInterface<IndexCastOpInterface>(*ctx);
SelectOp::attachInterface<SelectOpInterface>(*ctx);
});
}
|
