-
Notifications
You must be signed in to change notification settings - Fork 10.7k
/
BufferizationOps.cpp
641 lines (556 loc) · 24.1 KB
/
BufferizationOps.cpp
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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
//===----------------------------------------------------------------------===//
//
// 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/Arithmetic/IR/Arithmetic.h"
#include "mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h"
#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
#include "mlir/Dialect/MemRef/IR/MemRef.h"
#include "mlir/Dialect/MemRef/Utils/MemRefUtils.h"
#include "mlir/Dialect/Tensor/IR/Tensor.h"
#include "mlir/IR/Matchers.h"
using namespace mlir;
using namespace mlir::bufferization;
//===----------------------------------------------------------------------===//
// Helper functions
//===----------------------------------------------------------------------===//
FailureOr<Value>
mlir::bufferization::castOrReallocMemRefValue(OpBuilder &b, Value value,
MemRefType destType) {
auto srcType = value.getType().cast<MemRefType>();
// Element type, rank and memory space must match.
if (srcType.getElementType() != destType.getElementType())
return failure();
if (srcType.getMemorySpaceAsInt() != destType.getMemorySpaceAsInt())
return failure();
if (srcType.getRank() != destType.getRank())
return failure();
// In case the affine maps are different, we may need to use a copy if we go
// from dynamic to static offset or stride (the canonicalization cannot know
// at this point that it is really cast compatible).
auto isGuaranteedCastCompatible = [](MemRefType source, MemRefType target) {
int64_t sourceOffset, targetOffset;
SmallVector<int64_t, 4> sourceStrides, targetStrides;
if (failed(getStridesAndOffset(source, sourceStrides, sourceOffset)) ||
failed(getStridesAndOffset(target, targetStrides, targetOffset)))
return false;
auto dynamicToStatic = [](int64_t a, int64_t b) {
return a == MemRefType::getDynamicStrideOrOffset() &&
b != MemRefType::getDynamicStrideOrOffset();
};
if (dynamicToStatic(sourceOffset, targetOffset))
return false;
for (auto it : zip(sourceStrides, targetStrides))
if (dynamicToStatic(std::get<0>(it), std::get<1>(it)))
return false;
return true;
};
// Note: If `areCastCompatible`, a cast is valid, but may fail at runtime. To
// ensure that we only generate casts that always succeed at runtime, we check
// a fix extra conditions in `isGuaranteedCastCompatible`.
if (memref::CastOp::areCastCompatible(srcType, destType) &&
isGuaranteedCastCompatible(srcType, destType)) {
Value casted = b.create<memref::CastOp>(value.getLoc(), destType, value);
return casted;
}
auto loc = value.getLoc();
SmallVector<Value, 4> dynamicOperands;
for (int i = 0; i < destType.getRank(); ++i) {
if (destType.getShape()[i] != ShapedType::kDynamicSize)
continue;
auto index = b.createOrFold<arith::ConstantIndexOp>(loc, i);
Value size = b.create<memref::DimOp>(loc, value, index);
dynamicOperands.push_back(size);
}
// TODO: Use alloc/memcpy callback from BufferizationOptions if called via
// BufferizableOpInterface impl of ToMemrefOp.
Value copy = b.create<memref::AllocOp>(loc, destType, dynamicOperands);
b.create<memref::CopyOp>(loc, value, copy);
return copy;
}
/// Try to fold to_memref(to_tensor(x)). If x's type and the result type of the
/// to_memref op are different, a memref.cast is needed.
LogicalResult mlir::bufferization::foldToMemrefToTensorPair(
RewriterBase &rewriter, ToMemrefOp toMemref, bool allowSameType) {
auto memrefToTensor = toMemref.getTensor().getDefiningOp<ToTensorOp>();
if (!memrefToTensor)
return failure();
Type srcType = memrefToTensor.getMemref().getType();
Type destType = toMemref.getType();
// Directly rewrite if the type did not change.
if (srcType == destType) {
// Function can be configured to only handle cases where a cast is needed.
if (!allowSameType)
return failure();
rewriter.replaceOp(toMemref, memrefToTensor.getMemref());
return success();
}
auto rankedSrcType = srcType.dyn_cast<MemRefType>();
auto rankedDestType = destType.dyn_cast<MemRefType>();
auto unrankedSrcType = srcType.dyn_cast<UnrankedMemRefType>();
// Ranked memref -> Ranked memref cast.
if (rankedSrcType && rankedDestType) {
FailureOr<Value> replacement = castOrReallocMemRefValue(
rewriter, memrefToTensor.getMemref(), rankedDestType);
if (failed(replacement))
return failure();
rewriter.replaceOp(toMemref, *replacement);
return success();
}
// Unranked memref -> Ranked memref cast: May require a copy.
// TODO: Not implemented at the moment.
if (unrankedSrcType && rankedDestType)
return failure();
// Unranked memref -> unranked memref cast
// Ranked memref -> unranked memref cast: No copy needed.
assert(memref::CastOp::areCastCompatible(srcType, destType) &&
"expected that types are cast compatible");
rewriter.replaceOpWithNewOp<memref::CastOp>(toMemref, destType,
memrefToTensor.getMemref());
return success();
}
void mlir::bufferization::populateDynamicDimSizes(
OpBuilder &b, Location loc, Value shapedValue,
SmallVector<Value> &dynamicDims) {
auto shapedType = shapedValue.getType().cast<ShapedType>();
for (int64_t i = 0; i < shapedType.getRank(); ++i) {
if (shapedType.isDynamicDim(i)) {
if (shapedType.isa<MemRefType>()) {
dynamicDims.push_back(b.create<memref::DimOp>(loc, shapedValue, i));
} else {
assert(shapedType.isa<RankedTensorType>() && "expected tensor");
dynamicDims.push_back(b.create<tensor::DimOp>(loc, shapedValue, i));
}
}
}
}
//===----------------------------------------------------------------------===//
// AllocTensorOp
//===----------------------------------------------------------------------===//
LogicalResult AllocTensorOp::bufferize(RewriterBase &rewriter,
const BufferizationOptions &options) {
OpBuilder::InsertionGuard g(rewriter);
Operation *op = this->getOperation();
Location loc = getLoc();
// Nothing to do for dead AllocTensorOps.
if (getOperation()->getUses().empty()) {
rewriter.eraseOp(getOperation());
return success();
}
// Create buffer allocation.
Value copyBuffer;
if (getCopy())
copyBuffer = getBuffer(rewriter, getCopy(), options);
auto allocType =
MemRefType::get(getType().getShape(), getType().getElementType());
SmallVector<Value> dynamicDims = getDynamicSizes();
if (getCopy()) {
assert(dynamicDims.empty() && "expected either `copy` or `dynamicDims`");
populateDynamicDimSizes(rewriter, loc, copyBuffer, dynamicDims);
}
FailureOr<Value> alloc =
options.createAlloc(rewriter, loc, allocType, dynamicDims);
if (failed(alloc))
return failure();
// Create memory copy (if any).
if (getCopy()) {
if (failed(options.createMemCpy(rewriter, loc, copyBuffer, *alloc)))
return failure();
}
// Should the buffer be deallocated?
AnalysisState analysisState(options);
bool dealloc;
if (op->hasAttr(BufferizationDialect::kEscapeAttrName)) {
// AllocTensorOp has one result.
ArrayAttr escapeAttr =
op->getAttr(BufferizationDialect::kEscapeAttrName).cast<ArrayAttr>();
dealloc = !escapeAttr[0].cast<BoolAttr>().getValue();
} else {
// No "escape" annotation found.
if (options.createDeallocs) {
// Perform an ad-hoc analysis.
dealloc = !analysisState.isTensorYielded(getResult());
} else {
dealloc = false;
}
}
// Replace op.
replaceOpWithBufferizedValues(rewriter, getOperation(), *alloc);
// Create buffer deallocation (if requested).
if (!dealloc)
return success();
rewriter.setInsertionPoint(rewriter.getInsertionBlock()->getTerminator());
if (failed(options.createDealloc(rewriter, loc, *alloc)))
return failure();
return success();
}
bool AllocTensorOp::isMemoryWrite(OpResult opResult,
const AnalysisState &state) {
// AllocTensorOps do not write unless they have a `copy` value.
return static_cast<bool>(getCopy());
}
bool AllocTensorOp::bufferizesToMemoryRead(OpOperand &opOperand,
const AnalysisState &state) {
assert(opOperand.getOperandNumber() == getNumOperands() - 1 &&
"expected copy operand");
return true;
}
bool AllocTensorOp::bufferizesToMemoryWrite(OpOperand &opOperand,
const AnalysisState &state) {
assert(opOperand.getOperandNumber() == getNumOperands() - 1 &&
"expected copy operand");
return false;
}
SmallVector<OpResult>
AllocTensorOp::getAliasingOpResult(OpOperand &opOperand,
const AnalysisState &state) {
// This is a new allocation. It does not alias with any other buffer.
return {};
}
LogicalResult AllocTensorOp::verify() {
if (getCopy() && !getDynamicSizes().empty())
return emitError("dynamic sizes not needed when copying a tensor");
if (!getCopy() && getType().getNumDynamicDims() !=
static_cast<int64_t>(getDynamicSizes().size()))
return emitError("expected ")
<< getType().getNumDynamicDims() << " dynamic sizes";
if (getCopy() && getCopy().getType() != getType())
return emitError("expected that `copy` and return type match");
return success();
}
void AllocTensorOp::build(OpBuilder &builder, OperationState &result,
RankedTensorType type, ValueRange dynamicSizes) {
build(builder, result, type, dynamicSizes, /*copy=*/Value());
}
namespace {
/// Change the type of the result of a `bufferization.alloc_tensor` by making
/// the result type statically sized along dimension that in the original
/// operation where defined as dynamic, but the size was defined using a
/// `constant` op. For example:
///
/// %c5 = arith.constant 5: index
/// %0 = bufferization.alloc_tensor(%arg0, %c5) : tensor<?x?xf32>
///
/// to
///
/// %0 = bufferization.alloc_tensor(%arg0) : tensor<?x5xf32>
struct ReplaceStaticShapeDims : OpRewritePattern<AllocTensorOp> {
using OpRewritePattern<AllocTensorOp>::OpRewritePattern;
LogicalResult matchAndRewrite(AllocTensorOp op,
PatternRewriter &rewriter) const override {
if (op.getCopy())
return failure();
SmallVector<int64_t> newShape = llvm::to_vector(op.getType().getShape());
SmallVector<Value> newDynamicSizes;
unsigned int dynValCounter = 0;
for (int64_t i = 0; i < op.getType().getRank(); ++i) {
if (!op.isDynamicDim(i))
continue;
Value value = op.getDynamicSizes()[dynValCounter++];
APInt intVal;
if (matchPattern(value, m_ConstantInt(&intVal))) {
newShape[i] = intVal.getSExtValue();
} else {
newDynamicSizes.push_back(value);
}
}
RankedTensorType newType = RankedTensorType::get(
newShape, op.getType().getElementType(), op.getType().getEncoding());
if (newType == op.getType())
return failure();
auto newOp = rewriter.create<AllocTensorOp>(
op.getLoc(), newType, newDynamicSizes, /*copy=*/Value());
rewriter.replaceOpWithNewOp<tensor::CastOp>(op, op.getType(), newOp);
return success();
}
};
struct FoldDimOfAllocTensorOp : public OpRewritePattern<tensor::DimOp> {
using OpRewritePattern<tensor::DimOp>::OpRewritePattern;
LogicalResult matchAndRewrite(tensor::DimOp dimOp,
PatternRewriter &rewriter) const override {
Optional<int64_t> maybeConstantIndex = dimOp.getConstantIndex();
auto allocTensorOp = dimOp.source().getDefiningOp<AllocTensorOp>();
if (!allocTensorOp || !maybeConstantIndex)
return failure();
if (!allocTensorOp.getType().isDynamicDim(*maybeConstantIndex))
return failure();
rewriter.replaceOp(
dimOp, allocTensorOp.getDynamicSize(rewriter, *maybeConstantIndex));
return success();
}
};
} // namespace
void AllocTensorOp::getCanonicalizationPatterns(RewritePatternSet &results,
MLIRContext *ctx) {
results.add<FoldDimOfAllocTensorOp, ReplaceStaticShapeDims>(ctx);
}
LogicalResult AllocTensorOp::reifyResultShapes(
OpBuilder &builder, ReifiedRankedShapedTypeDims &reifiedReturnShapes) {
auto shapes = llvm::to_vector<4>(llvm::map_range(
llvm::seq<int64_t>(0, getType().getRank()), [&](int64_t dim) -> Value {
if (isDynamicDim(dim))
return getDynamicSize(builder, dim);
return builder.create<arith::ConstantIndexOp>(getLoc(),
getStaticSize(dim));
}));
reifiedReturnShapes.emplace_back(std::move(shapes));
return success();
}
ParseResult AllocTensorOp::parse(OpAsmParser &parser, OperationState &result) {
SmallVector<OpAsmParser::UnresolvedOperand> dynamicSizesOperands;
if (parser.parseLParen() || parser.parseOperandList(dynamicSizesOperands) ||
parser.parseRParen())
return failure();
ParseResult copyKeyword = parser.parseOptionalKeyword("copy");
OpAsmParser::UnresolvedOperand copyOperand;
if (copyKeyword.succeeded())
if (parser.parseLParen() || parser.parseOperand(copyOperand) ||
parser.parseRParen())
return failure();
if (parser.parseOptionalAttrDict(result.attributes) || parser.parseColon())
return failure();
TensorType type;
if (parser.parseCustomTypeWithFallback(type))
return failure();
result.addTypes(type);
Type indexType = parser.getBuilder().getIndexType();
if (parser.resolveOperands(dynamicSizesOperands, indexType, result.operands))
return failure();
if (copyKeyword.succeeded())
if (parser.resolveOperand(copyOperand, type, result.operands))
return failure();
result.addAttribute(AllocTensorOp::getOperandSegmentSizeAttr(),
parser.getBuilder().getI32VectorAttr(
{static_cast<int32_t>(dynamicSizesOperands.size()),
static_cast<int32_t>(copyKeyword.succeeded())}));
return success();
}
void AllocTensorOp::print(OpAsmPrinter &p) {
p << "(" << getDynamicSizes() << ")";
if (getCopy())
p << " copy(" << getCopy() << ")";
p.printOptionalAttrDict((*this)->getAttrs(), /*elidedAttrs=*/{
AllocTensorOp::getOperandSegmentSizeAttr()});
p << " : ";
auto type = getResult().getType();
if (auto validType = type.dyn_cast<::mlir::TensorType>())
p.printStrippedAttrOrType(validType);
else
p << type;
}
Value AllocTensorOp::getDynamicSize(OpBuilder &b, unsigned idx) {
assert(isDynamicDim(idx) && "expected dynamic dim");
if (getCopy())
return b.create<tensor::DimOp>(getLoc(), getCopy(), idx);
return getOperand(getIndexOfDynamicSize(idx));
}
//===----------------------------------------------------------------------===//
// CloneOp
//===----------------------------------------------------------------------===//
void CloneOp::getEffects(
SmallVectorImpl<SideEffects::EffectInstance<MemoryEffects::Effect>>
&effects) {
effects.emplace_back(MemoryEffects::Read::get(), getInput(),
SideEffects::DefaultResource::get());
effects.emplace_back(MemoryEffects::Write::get(), getOutput(),
SideEffects::DefaultResource::get());
effects.emplace_back(MemoryEffects::Allocate::get(), getOutput(),
SideEffects::DefaultResource::get());
}
OpFoldResult CloneOp::fold(ArrayRef<Attribute> operands) {
return succeeded(memref::foldMemRefCast(*this)) ? getResult() : Value();
}
namespace {
/// Merge the clone and its source (by converting the clone to a cast) when
/// possible.
struct SimplifyClones : public OpRewritePattern<CloneOp> {
using OpRewritePattern<CloneOp>::OpRewritePattern;
LogicalResult matchAndRewrite(CloneOp cloneOp,
PatternRewriter &rewriter) const override {
if (cloneOp.use_empty()) {
rewriter.eraseOp(cloneOp);
return success();
}
Value source = cloneOp.getInput();
// This only finds dealloc operations for the immediate value. It should
// also consider aliases. That would also make the safety check below
// redundant.
llvm::Optional<Operation *> maybeCloneDeallocOp =
memref::findDealloc(cloneOp.getOutput());
// Skip if either of them has > 1 deallocate operations.
if (!maybeCloneDeallocOp.hasValue())
return failure();
llvm::Optional<Operation *> maybeSourceDeallocOp =
memref::findDealloc(source);
if (!maybeSourceDeallocOp.hasValue())
return failure();
Operation *cloneDeallocOp = *maybeCloneDeallocOp;
Operation *sourceDeallocOp = *maybeSourceDeallocOp;
// If both are deallocated in the same block, their in-block lifetimes
// might not fully overlap, so we cannot decide which one to drop.
if (cloneDeallocOp && sourceDeallocOp &&
cloneDeallocOp->getBlock() == sourceDeallocOp->getBlock())
return failure();
Block *currentBlock = cloneOp->getBlock();
Operation *redundantDealloc = nullptr;
if (cloneDeallocOp && cloneDeallocOp->getBlock() == currentBlock) {
redundantDealloc = cloneDeallocOp;
} else if (sourceDeallocOp && sourceDeallocOp->getBlock() == currentBlock) {
redundantDealloc = sourceDeallocOp;
}
if (!redundantDealloc)
return failure();
// Safety check that there are no other deallocations inbetween
// cloneOp and redundantDealloc, as otherwise we might deallocate an alias
// of source before the uses of the clone. With alias information, we could
// restrict this to only fail of the dealloc's operand is an alias
// of the source.
for (Operation *pos = cloneOp->getNextNode(); pos != redundantDealloc;
pos = pos->getNextNode()) {
auto effectInterface = dyn_cast<MemoryEffectOpInterface>(pos);
if (!effectInterface)
continue;
if (effectInterface.hasEffect<MemoryEffects::Free>())
return failure();
}
rewriter.replaceOpWithNewOp<memref::CastOp>(cloneOp, cloneOp.getType(),
source);
rewriter.eraseOp(redundantDealloc);
return success();
}
};
} // namespace
void CloneOp::getCanonicalizationPatterns(RewritePatternSet &results,
MLIRContext *context) {
results.add<SimplifyClones>(context);
}
//===----------------------------------------------------------------------===//
// ToTensorOp
//===----------------------------------------------------------------------===//
OpFoldResult ToTensorOp::fold(ArrayRef<Attribute>) {
if (auto toMemref = getMemref().getDefiningOp<ToMemrefOp>())
// Approximate alias analysis by conservatively folding only when no there
// is no interleaved operation.
if (toMemref->getBlock() == this->getOperation()->getBlock() &&
toMemref->getNextNode() == this->getOperation())
return toMemref.getTensor();
return {};
}
namespace {
struct DimOfToTensorFolder : public OpRewritePattern<tensor::DimOp> {
using OpRewritePattern<tensor::DimOp>::OpRewritePattern;
LogicalResult matchAndRewrite(tensor::DimOp dimOp,
PatternRewriter &rewriter) const override {
auto memrefToTensorOp = dimOp.source().getDefiningOp<ToTensorOp>();
if (!memrefToTensorOp)
return failure();
rewriter.replaceOpWithNewOp<memref::DimOp>(
dimOp, memrefToTensorOp.getMemref(), dimOp.index());
return success();
}
};
} // namespace
void ToTensorOp::getCanonicalizationPatterns(RewritePatternSet &results,
MLIRContext *context) {
results.add<DimOfToTensorFolder>(context);
}
//===----------------------------------------------------------------------===//
// ToMemrefOp
//===----------------------------------------------------------------------===//
OpFoldResult ToMemrefOp::fold(ArrayRef<Attribute>) {
if (auto memrefToTensor = getTensor().getDefiningOp<ToTensorOp>())
if (memrefToTensor.getMemref().getType() == getType())
return memrefToTensor.getMemref();
return {};
}
namespace {
/// Replace tensor.cast + to_memref by to_memref + memref.cast.
struct ToMemrefOfCast : public OpRewritePattern<ToMemrefOp> {
using OpRewritePattern<ToMemrefOp>::OpRewritePattern;
LogicalResult matchAndRewrite(ToMemrefOp toMemref,
PatternRewriter &rewriter) const final {
auto tensorCastOperand =
toMemref.getOperand().getDefiningOp<tensor::CastOp>();
if (!tensorCastOperand)
return failure();
auto srcTensorType =
tensorCastOperand.getOperand().getType().dyn_cast<RankedTensorType>();
if (!srcTensorType)
return failure();
auto memrefType = MemRefType::get(srcTensorType.getShape(),
srcTensorType.getElementType());
Value memref = rewriter.create<ToMemrefOp>(toMemref.getLoc(), memrefType,
tensorCastOperand.getOperand());
rewriter.replaceOpWithNewOp<memref::CastOp>(toMemref, toMemref.getType(),
memref);
return success();
}
};
/// Canonicalize bufferization.to_tensor + bufferization.to_memref to
/// memref.cast when type mismatches prevent `ToMemrefOp::fold` to kick in.
struct TensorLoadToMemref : public OpRewritePattern<ToMemrefOp> {
using OpRewritePattern<ToMemrefOp>::OpRewritePattern;
LogicalResult matchAndRewrite(ToMemrefOp toMemref,
PatternRewriter &rewriter) const final {
// Only handle cases where a cast is needed. The other case is handled by
// the folder.
return foldToMemrefToTensorPair(rewriter, toMemref,
/*allowSameType=*/false);
}
};
/// Fold a load on a to_memref operation into an tensor.extract on the
/// corresponding tensor.
struct LoadOfToMemref : public OpRewritePattern<memref::LoadOp> {
using OpRewritePattern<memref::LoadOp>::OpRewritePattern;
LogicalResult matchAndRewrite(memref::LoadOp load,
PatternRewriter &rewriter) const override {
auto toMemref = load.memref().getDefiningOp<ToMemrefOp>();
if (!toMemref)
return failure();
rewriter.replaceOpWithNewOp<tensor::ExtractOp>(load, toMemref.getTensor(),
load.indices());
return success();
}
};
/// Fold dim of a to_memref into the dim of the tensor.
struct DimOfCastOp : public OpRewritePattern<memref::DimOp> {
using OpRewritePattern<memref::DimOp>::OpRewritePattern;
LogicalResult matchAndRewrite(memref::DimOp dimOp,
PatternRewriter &rewriter) const override {
auto castOp = dimOp.source().getDefiningOp<ToMemrefOp>();
if (!castOp)
return failure();
Value newSource = castOp.getOperand();
rewriter.replaceOpWithNewOp<tensor::DimOp>(dimOp, newSource, dimOp.index());
return success();
}
};
} // namespace
void ToMemrefOp::getCanonicalizationPatterns(RewritePatternSet &results,
MLIRContext *context) {
results.add<DimOfCastOp, LoadOfToMemref, ToMemrefOfCast, TensorLoadToMemref>(
context);
}
LogicalResult ToMemrefOp::bufferize(RewriterBase &rewriter,
const BufferizationOptions &options) {
// Fold to_memref(to_tensor(x)) to x. Insert a cast if necessary.
(void)foldToMemrefToTensorPair(rewriter, *this);
// Note: The return value of `bufferize` indicates whether there was an error
// or not. (And not whether the pattern matched or not.)
return success();
}
Optional<Operation *> CloneOp::buildDealloc(OpBuilder &builder, Value alloc) {
return builder.create<memref::DeallocOp>(alloc.getLoc(), alloc)
.getOperation();
}
Optional<Value> CloneOp::buildClone(OpBuilder &builder, Value alloc) {
return builder.create<CloneOp>(alloc.getLoc(), alloc).getResult();
}
//===----------------------------------------------------------------------===//
// TableGen'd op method definitions
//===----------------------------------------------------------------------===//
#define GET_OP_CLASSES
#include "mlir/Dialect/Bufferization/IR/BufferizationOps.cpp.inc"