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
|
//===- TensorCopyInsertion.cpp - Resolve Bufferization Conflicts w/ Copies ===//
//
// 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/Bufferization/Transforms/Passes.h"
#include "mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h"
#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
#include "mlir/Dialect/Bufferization/Transforms/Bufferize.h"
#include "mlir/Dialect/Bufferization/Transforms/OneShotAnalysis.h"
#include "mlir/Dialect/Bufferization/Transforms/OneShotModuleBufferize.h"
#include "mlir/Dialect/Bufferization/Transforms/Transforms.h"
#include "mlir/Dialect/Func/IR/FuncOps.h"
namespace mlir {
namespace bufferization {
#define GEN_PASS_DEF_TENSORCOPYINSERTION
#include "mlir/Dialect/Bufferization/Transforms/Passes.h.inc"
} // namespace bufferization
} // namespace mlir
using namespace mlir;
using namespace mlir::bufferization;
/// Resolve all operands that are also used inside of repetitive regions of the
/// same op. Such cases are not fully supported by One-Shot Bufferize.
///
/// E.g.:
/// %r = scf.for ... iter_args(%t = %tensor) -> tensor<?xf32> {
/// "some_use"(%tensor)
/// ...
/// }
///
/// Is converted to:
/// %tensor_copy = bufferization.alloc_tensor copy(%tensor)
/// %r = scf.for ... iter_args(%t = %tensor) -> tensor<?xf32> {
/// "some_use"(%tensor_copy)
/// ...
/// }
static void
resolveUsesInRepetitiveRegions(Operation *op,
const BufferizationOptions &options) {
IRRewriter rewriter(op->getContext());
AnalysisState state(options);
// Look for repetitive ops (loops).
op->walk([&](BufferizableOpInterface bufferizableOp) {
// Skip filtered ops.
if (!options.isOpAllowed(bufferizableOp.getOperation()))
return WalkResult::advance();
// Find all operands that are also used inside of a repetitive region of
// this op.
for (OpOperand &opOperand : bufferizableOp->getOpOperands()) {
Value operand = opOperand.get();
// Skip non-tensor operands.
if (!isa<TensorType>(operand.getType()))
continue;
// Skip operands that do not bufferize to memory writes.
if (!bufferizableOp.bufferizesToMemoryWrite(opOperand, state))
continue;
// Gather all uses inside repetitive regions.
SmallVector<OpOperand *> usesInsideRegion;
for (OpOperand &use : operand.getUses()) {
Operation *owner = use.getOwner();
if (!bufferizableOp->isProperAncestor(owner))
continue;
for (Region &r : bufferizableOp->getRegions()) {
if (r.findAncestorOpInRegion(*owner) &&
bufferizableOp.isRepetitiveRegion(r.getRegionNumber())) {
usesInsideRegion.push_back(&use);
break;
}
}
}
// Nothing to do if the operand is not used inside a repetitive region.
if (usesInsideRegion.empty())
continue;
// Insert a tensor copy and replace all uses inside of repetitive regions.
rewriter.setInsertionPoint(bufferizableOp);
auto tensorCopy = rewriter.create<AllocTensorOp>(
bufferizableOp->getLoc(), cast<TensorType>(operand.getType()),
/*dynamicSizes=*/ValueRange(),
/*copy=*/operand, /*memory_space=*/IntegerAttr());
for (OpOperand *use : usesInsideRegion)
use->set(tensorCopy);
}
return WalkResult::advance();
});
}
LogicalResult mlir::bufferization::insertTensorCopies(
Operation *op, const OneShotBufferizationOptions &options,
BufferizationStatistics *statistics) {
// Preprocessing: Resolve currently unsupported bufferization cases.
resolveUsesInRepetitiveRegions(op, options);
OneShotAnalysisState state(op, options);
// Run normal One-Shot Bufferize analysis or One-Shot Module Bufferize
// analysis depending on whether function boundary bufferization is enabled or
// not.
if (options.bufferizeFunctionBoundaries) {
if (failed(analyzeModuleOp(cast<ModuleOp>(op), state, statistics)))
return failure();
} else {
if (failed(analyzeOp(op, state, statistics)))
return failure();
}
if (options.testAnalysisOnly)
return success();
return insertTensorCopies(op, state);
}
LogicalResult
mlir::bufferization::insertTensorCopies(Operation *op,
const AnalysisState &state) {
IRRewriter rewriter(op->getContext());
StringRef escapeAttrName = BufferizationDialect::kEscapeAttrName;
WalkResult result = op->walk([&](Operation *op) {
auto bufferizableOp = state.getOptions().dynCastBufferizableOp(op);
if (!bufferizableOp)
return WalkResult::skip();
// Find allocations without an `escape` attribute and add the attribute
// based on analysis results.
if (!op->hasAttr(escapeAttrName)) {
SmallVector<bool> escapeAttrValue;
bool foundTensorResult = false;
for (OpResult opResult : op->getOpResults()) {
if (!isa<TensorType>(opResult.getType()) ||
!bufferizableOp.bufferizesToAllocation(opResult)) {
escapeAttrValue.push_back(false);
continue;
}
foundTensorResult = true;
bool escape = !state.getOptions().createDeallocs ||
state.isTensorYielded(opResult);
escapeAttrValue.push_back(escape);
}
if (foundTensorResult)
op->setAttr(escapeAttrName, rewriter.getBoolArrayAttr(escapeAttrValue));
}
// Find inplacability conflicts and resolve them. (Typically with explicit
// tensor copies in the form of AllocTensorOps.)
rewriter.setInsertionPoint(op);
if (failed(bufferizableOp.resolveConflicts(rewriter, state)))
return WalkResult::interrupt();
return WalkResult::advance();
});
return failure(result.wasInterrupted());
}
|
