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
|
//===- MathToLLVM.cpp - Math to LLVM dialect conversion -------------------===//
//
// 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/Conversion/MathToLLVM/MathToLLVM.h"
#include "mlir/Conversion/ArithCommon/AttrToLLVMConverter.h"
#include "mlir/Conversion/ConvertToLLVM/ToLLVMInterface.h"
#include "mlir/Conversion/LLVMCommon/ConversionTarget.h"
#include "mlir/Conversion/LLVMCommon/Pattern.h"
#include "mlir/Conversion/LLVMCommon/VectorPattern.h"
#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
#include "mlir/Dialect/Math/IR/Math.h"
#include "mlir/IR/TypeUtilities.h"
#include "mlir/Pass/Pass.h"
namespace mlir {
#define GEN_PASS_DEF_CONVERTMATHTOLLVMPASS
#include "mlir/Conversion/Passes.h.inc"
} // namespace mlir
using namespace mlir;
namespace {
template <typename SourceOp, typename TargetOp>
using ConvertFastMath = arith::AttrConvertFastMathToLLVM<SourceOp, TargetOp>;
template <typename SourceOp, typename TargetOp>
using ConvertFMFMathToLLVMPattern =
VectorConvertToLLVMPattern<SourceOp, TargetOp, ConvertFastMath>;
using AbsFOpLowering = ConvertFMFMathToLLVMPattern<math::AbsFOp, LLVM::FAbsOp>;
using CeilOpLowering = ConvertFMFMathToLLVMPattern<math::CeilOp, LLVM::FCeilOp>;
using CopySignOpLowering =
ConvertFMFMathToLLVMPattern<math::CopySignOp, LLVM::CopySignOp>;
using CosOpLowering = ConvertFMFMathToLLVMPattern<math::CosOp, LLVM::CosOp>;
using CtPopFOpLowering =
VectorConvertToLLVMPattern<math::CtPopOp, LLVM::CtPopOp>;
using Exp2OpLowering = ConvertFMFMathToLLVMPattern<math::Exp2Op, LLVM::Exp2Op>;
using ExpOpLowering = ConvertFMFMathToLLVMPattern<math::ExpOp, LLVM::ExpOp>;
using FloorOpLowering =
ConvertFMFMathToLLVMPattern<math::FloorOp, LLVM::FFloorOp>;
using FmaOpLowering = ConvertFMFMathToLLVMPattern<math::FmaOp, LLVM::FMAOp>;
using Log10OpLowering =
ConvertFMFMathToLLVMPattern<math::Log10Op, LLVM::Log10Op>;
using Log2OpLowering = ConvertFMFMathToLLVMPattern<math::Log2Op, LLVM::Log2Op>;
using LogOpLowering = ConvertFMFMathToLLVMPattern<math::LogOp, LLVM::LogOp>;
using PowFOpLowering = ConvertFMFMathToLLVMPattern<math::PowFOp, LLVM::PowOp>;
using FPowIOpLowering =
ConvertFMFMathToLLVMPattern<math::FPowIOp, LLVM::PowIOp>;
using RoundEvenOpLowering =
ConvertFMFMathToLLVMPattern<math::RoundEvenOp, LLVM::RoundEvenOp>;
using RoundOpLowering =
ConvertFMFMathToLLVMPattern<math::RoundOp, LLVM::RoundOp>;
using SinOpLowering = ConvertFMFMathToLLVMPattern<math::SinOp, LLVM::SinOp>;
using SqrtOpLowering = ConvertFMFMathToLLVMPattern<math::SqrtOp, LLVM::SqrtOp>;
using FTruncOpLowering =
ConvertFMFMathToLLVMPattern<math::TruncOp, LLVM::FTruncOp>;
// A `CtLz/CtTz/absi(a)` is converted into `CtLz/CtTz/absi(a, false)`.
template <typename MathOp, typename LLVMOp>
struct IntOpWithFlagLowering : public ConvertOpToLLVMPattern<MathOp> {
using ConvertOpToLLVMPattern<MathOp>::ConvertOpToLLVMPattern;
using Super = IntOpWithFlagLowering<MathOp, LLVMOp>;
LogicalResult
matchAndRewrite(MathOp op, typename MathOp::Adaptor adaptor,
ConversionPatternRewriter &rewriter) const override {
auto operandType = adaptor.getOperand().getType();
if (!operandType || !LLVM::isCompatibleType(operandType))
return failure();
auto loc = op.getLoc();
auto resultType = op.getResult().getType();
if (!isa<LLVM::LLVMArrayType>(operandType)) {
rewriter.replaceOpWithNewOp<LLVMOp>(op, resultType, adaptor.getOperand(),
false);
return success();
}
auto vectorType = dyn_cast<VectorType>(resultType);
if (!vectorType)
return failure();
return LLVM::detail::handleMultidimensionalVectors(
op.getOperation(), adaptor.getOperands(), *this->getTypeConverter(),
[&](Type llvm1DVectorTy, ValueRange operands) {
return rewriter.create<LLVMOp>(loc, llvm1DVectorTy, operands[0],
false);
},
rewriter);
}
};
using CountLeadingZerosOpLowering =
IntOpWithFlagLowering<math::CountLeadingZerosOp, LLVM::CountLeadingZerosOp>;
using CountTrailingZerosOpLowering =
IntOpWithFlagLowering<math::CountTrailingZerosOp,
LLVM::CountTrailingZerosOp>;
using AbsIOpLowering = IntOpWithFlagLowering<math::AbsIOp, LLVM::AbsOp>;
// A `expm1` is converted into `exp - 1`.
struct ExpM1OpLowering : public ConvertOpToLLVMPattern<math::ExpM1Op> {
using ConvertOpToLLVMPattern<math::ExpM1Op>::ConvertOpToLLVMPattern;
LogicalResult
matchAndRewrite(math::ExpM1Op op, OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const override {
auto operandType = adaptor.getOperand().getType();
if (!operandType || !LLVM::isCompatibleType(operandType))
return failure();
auto loc = op.getLoc();
auto resultType = op.getResult().getType();
auto floatType = cast<FloatType>(getElementTypeOrSelf(resultType));
auto floatOne = rewriter.getFloatAttr(floatType, 1.0);
ConvertFastMath<math::ExpM1Op, LLVM::ExpOp> expAttrs(op);
ConvertFastMath<math::ExpM1Op, LLVM::FSubOp> subAttrs(op);
if (!isa<LLVM::LLVMArrayType>(operandType)) {
LLVM::ConstantOp one;
if (LLVM::isCompatibleVectorType(operandType)) {
one = rewriter.create<LLVM::ConstantOp>(
loc, operandType,
SplatElementsAttr::get(cast<ShapedType>(resultType), floatOne));
} else {
one = rewriter.create<LLVM::ConstantOp>(loc, operandType, floatOne);
}
auto exp = rewriter.create<LLVM::ExpOp>(loc, adaptor.getOperand(),
expAttrs.getAttrs());
rewriter.replaceOpWithNewOp<LLVM::FSubOp>(
op, operandType, ValueRange{exp, one}, subAttrs.getAttrs());
return success();
}
auto vectorType = dyn_cast<VectorType>(resultType);
if (!vectorType)
return rewriter.notifyMatchFailure(op, "expected vector result type");
return LLVM::detail::handleMultidimensionalVectors(
op.getOperation(), adaptor.getOperands(), *getTypeConverter(),
[&](Type llvm1DVectorTy, ValueRange operands) {
auto splatAttr = SplatElementsAttr::get(
mlir::VectorType::get(
{LLVM::getVectorNumElements(llvm1DVectorTy).getFixedValue()},
floatType),
floatOne);
auto one =
rewriter.create<LLVM::ConstantOp>(loc, llvm1DVectorTy, splatAttr);
auto exp = rewriter.create<LLVM::ExpOp>(
loc, llvm1DVectorTy, operands[0], expAttrs.getAttrs());
return rewriter.create<LLVM::FSubOp>(
loc, llvm1DVectorTy, ValueRange{exp, one}, subAttrs.getAttrs());
},
rewriter);
}
};
// A `log1p` is converted into `log(1 + ...)`.
struct Log1pOpLowering : public ConvertOpToLLVMPattern<math::Log1pOp> {
using ConvertOpToLLVMPattern<math::Log1pOp>::ConvertOpToLLVMPattern;
LogicalResult
matchAndRewrite(math::Log1pOp op, OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const override {
auto operandType = adaptor.getOperand().getType();
if (!operandType || !LLVM::isCompatibleType(operandType))
return rewriter.notifyMatchFailure(op, "unsupported operand type");
auto loc = op.getLoc();
auto resultType = op.getResult().getType();
auto floatType = cast<FloatType>(getElementTypeOrSelf(resultType));
auto floatOne = rewriter.getFloatAttr(floatType, 1.0);
ConvertFastMath<math::Log1pOp, LLVM::FAddOp> addAttrs(op);
ConvertFastMath<math::Log1pOp, LLVM::LogOp> logAttrs(op);
if (!isa<LLVM::LLVMArrayType>(operandType)) {
LLVM::ConstantOp one =
LLVM::isCompatibleVectorType(operandType)
? rewriter.create<LLVM::ConstantOp>(
loc, operandType,
SplatElementsAttr::get(cast<ShapedType>(resultType),
floatOne))
: rewriter.create<LLVM::ConstantOp>(loc, operandType, floatOne);
auto add = rewriter.create<LLVM::FAddOp>(
loc, operandType, ValueRange{one, adaptor.getOperand()},
addAttrs.getAttrs());
rewriter.replaceOpWithNewOp<LLVM::LogOp>(op, operandType, ValueRange{add},
logAttrs.getAttrs());
return success();
}
auto vectorType = dyn_cast<VectorType>(resultType);
if (!vectorType)
return rewriter.notifyMatchFailure(op, "expected vector result type");
return LLVM::detail::handleMultidimensionalVectors(
op.getOperation(), adaptor.getOperands(), *getTypeConverter(),
[&](Type llvm1DVectorTy, ValueRange operands) {
auto splatAttr = SplatElementsAttr::get(
mlir::VectorType::get(
{LLVM::getVectorNumElements(llvm1DVectorTy).getFixedValue()},
floatType),
floatOne);
auto one =
rewriter.create<LLVM::ConstantOp>(loc, llvm1DVectorTy, splatAttr);
auto add = rewriter.create<LLVM::FAddOp>(loc, llvm1DVectorTy,
ValueRange{one, operands[0]},
addAttrs.getAttrs());
return rewriter.create<LLVM::LogOp>(
loc, llvm1DVectorTy, ValueRange{add}, logAttrs.getAttrs());
},
rewriter);
}
};
// A `rsqrt` is converted into `1 / sqrt`.
struct RsqrtOpLowering : public ConvertOpToLLVMPattern<math::RsqrtOp> {
using ConvertOpToLLVMPattern<math::RsqrtOp>::ConvertOpToLLVMPattern;
LogicalResult
matchAndRewrite(math::RsqrtOp op, OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const override {
auto operandType = adaptor.getOperand().getType();
if (!operandType || !LLVM::isCompatibleType(operandType))
return failure();
auto loc = op.getLoc();
auto resultType = op.getResult().getType();
auto floatType = cast<FloatType>(getElementTypeOrSelf(resultType));
auto floatOne = rewriter.getFloatAttr(floatType, 1.0);
ConvertFastMath<math::RsqrtOp, LLVM::SqrtOp> sqrtAttrs(op);
ConvertFastMath<math::RsqrtOp, LLVM::FDivOp> divAttrs(op);
if (!isa<LLVM::LLVMArrayType>(operandType)) {
LLVM::ConstantOp one;
if (LLVM::isCompatibleVectorType(operandType)) {
one = rewriter.create<LLVM::ConstantOp>(
loc, operandType,
SplatElementsAttr::get(cast<ShapedType>(resultType), floatOne));
} else {
one = rewriter.create<LLVM::ConstantOp>(loc, operandType, floatOne);
}
auto sqrt = rewriter.create<LLVM::SqrtOp>(loc, adaptor.getOperand(),
sqrtAttrs.getAttrs());
rewriter.replaceOpWithNewOp<LLVM::FDivOp>(
op, operandType, ValueRange{one, sqrt}, divAttrs.getAttrs());
return success();
}
auto vectorType = dyn_cast<VectorType>(resultType);
if (!vectorType)
return failure();
return LLVM::detail::handleMultidimensionalVectors(
op.getOperation(), adaptor.getOperands(), *getTypeConverter(),
[&](Type llvm1DVectorTy, ValueRange operands) {
auto splatAttr = SplatElementsAttr::get(
mlir::VectorType::get(
{LLVM::getVectorNumElements(llvm1DVectorTy).getFixedValue()},
floatType),
floatOne);
auto one =
rewriter.create<LLVM::ConstantOp>(loc, llvm1DVectorTy, splatAttr);
auto sqrt = rewriter.create<LLVM::SqrtOp>(
loc, llvm1DVectorTy, operands[0], sqrtAttrs.getAttrs());
return rewriter.create<LLVM::FDivOp>(
loc, llvm1DVectorTy, ValueRange{one, sqrt}, divAttrs.getAttrs());
},
rewriter);
}
};
struct ConvertMathToLLVMPass
: public impl::ConvertMathToLLVMPassBase<ConvertMathToLLVMPass> {
using Base::Base;
void runOnOperation() override {
RewritePatternSet patterns(&getContext());
LLVMTypeConverter converter(&getContext());
populateMathToLLVMConversionPatterns(converter, patterns, approximateLog1p);
LLVMConversionTarget target(getContext());
if (failed(applyPartialConversion(getOperation(), target,
std::move(patterns))))
signalPassFailure();
}
};
} // namespace
void mlir::populateMathToLLVMConversionPatterns(LLVMTypeConverter &converter,
RewritePatternSet &patterns,
bool approximateLog1p) {
if (approximateLog1p)
patterns.add<Log1pOpLowering>(converter);
// clang-format off
patterns.add<
AbsFOpLowering,
AbsIOpLowering,
CeilOpLowering,
CopySignOpLowering,
CosOpLowering,
CountLeadingZerosOpLowering,
CountTrailingZerosOpLowering,
CtPopFOpLowering,
Exp2OpLowering,
ExpM1OpLowering,
ExpOpLowering,
FPowIOpLowering,
FloorOpLowering,
FmaOpLowering,
Log10OpLowering,
Log2OpLowering,
LogOpLowering,
PowFOpLowering,
RoundEvenOpLowering,
RoundOpLowering,
RsqrtOpLowering,
SinOpLowering,
SqrtOpLowering,
FTruncOpLowering
>(converter);
// clang-format on
}
//===----------------------------------------------------------------------===//
// ConvertToLLVMPatternInterface implementation
//===----------------------------------------------------------------------===//
namespace {
/// Implement the interface to convert Math to LLVM.
struct MathToLLVMDialectInterface : public ConvertToLLVMPatternInterface {
using ConvertToLLVMPatternInterface::ConvertToLLVMPatternInterface;
void loadDependentDialects(MLIRContext *context) const final {
context->loadDialect<LLVM::LLVMDialect>();
}
/// Hook for derived dialect interface to provide conversion patterns
/// and mark dialect legal for the conversion target.
void populateConvertToLLVMConversionPatterns(
ConversionTarget &target, LLVMTypeConverter &typeConverter,
RewritePatternSet &patterns) const final {
populateMathToLLVMConversionPatterns(typeConverter, patterns);
}
};
} // namespace
void mlir::registerConvertMathToLLVMInterface(DialectRegistry ®istry) {
registry.addExtension(+[](MLIRContext *ctx, math::MathDialect *dialect) {
dialect->addInterfaces<MathToLLVMDialectInterface>();
});
}
|