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
|
/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include "tensorflow/compiler/xla/service/cpu/compiler_functor.h"
#include <algorithm>
#include <iterator>
#include <memory>
#include <string>
#include <utility>
#include <vector>
#include "llvm/ADT/StringRef.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/IR/LegacyPassManager.h"
#include "llvm/IR/Verifier.h"
#include "llvm/MC/MCContext.h"
#include "llvm/Object/ObjectFile.h"
#include "llvm/Support/SmallVectorMemoryBuffer.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Transforms/IPO/AlwaysInliner.h"
#include "llvm/Transforms/IPO/PassManagerBuilder.h"
#include "tensorflow/compiler/xla/ptr_util.h"
#include "tensorflow/compiler/xla/service/cpu/cpu_runtime.h"
#include "tensorflow/compiler/xla/service/cpu/llvm_ir_runtime.h"
#include "tensorflow/compiler/xla/service/llvm_ir/llvm_util.h"
#include "tensorflow/compiler/xla/statusor.h"
#include "tensorflow/compiler/xla/types.h"
#include "tensorflow/compiler/xla/util.h"
#include "tensorflow/core/platform/logging.h"
namespace xla {
namespace cpu {
/* Create filtered versions of the LLVM Pass Managers to filter out some
of the expensive passes.
Profiling:
learning/brain/google/xla/benchmarks:inception_cpu_benchmark
learning/brain/google/xla/benchmarks:cifarnet
pointed to LICM and IndVarSimplify as the hottest passes.
LICM is known to exhibit O(n^2) time in the number of instructions.
IndVarSimplify is slow due to SCEV. If loops are emitted in canonical form,
this pass is not necessary.
Disabling these as a starting point.
*/
// TODO(b/64227304) Creating a custom pass pipeline will replace this.
namespace {
class FilteredFunctionPassManager : public llvm::legacy::FunctionPassManager {
public:
FilteredFunctionPassManager(llvm::Module* m, bool disable_expensive_passes)
: llvm::legacy::FunctionPassManager(m),
disable_expensive_passes_(disable_expensive_passes) {}
void add(llvm::Pass* p) override {
llvm::legacy::FunctionPassManager::add(p);
}
private:
bool disable_expensive_passes_;
};
class FilteredPassManager : public llvm::legacy::PassManager {
public:
explicit FilteredPassManager(bool disable_expensive_passes)
: disable_expensive_passes_(disable_expensive_passes) {}
void add(llvm::Pass* p) override {
if (disable_expensive_passes_) {
llvm::StringRef PassName = p->getPassName();
if (PassName.contains("Unroll loops")) {
return;
}
}
llvm::legacy::PassManager::add(p);
}
private:
bool disable_expensive_passes_;
};
} // anonymous namespace
std::unique_ptr<llvm::MemoryBuffer> CompilerFunctor::operator()(
llvm::Module& module) const {
FilteredPassManager module_passes(disable_expensive_passes_);
FilteredFunctionPassManager function_passes(&module,
disable_expensive_passes_);
VLOG(2) << "IR before optimizations";
XLA_VLOG_LINES(2, llvm_ir::DumpModuleToString(module));
if (pre_optimization_hook_) {
TF_CHECK_OK(pre_optimization_hook_(module));
}
// Add the appropriate TargetLibraryInfo and TargetTransformInfo.
AddTargetInfoPasses(&module_passes);
// Build up optimization pipeline.
if (optimize_for_size_) {
// Optimizing for size turns on -O2 level optimizations.
//
// TODO(b/64153864): Although the code generator supports size_level = 2 to
// turn on more aggressive code size optimizations than size_level = 1, we
// pass size_level = 1 because in many cases a size_level of 2 does
// worse. Investigate why.
AddOptimizationPasses(&module_passes, &function_passes, /*opt_level=*/2,
/*size_level=*/1);
} else {
AddOptimizationPasses(&module_passes, &function_passes,
/*opt_level=*/opt_level_, /*size_level=*/0);
}
// Run optimization passes on module.
function_passes.doInitialization();
CHECK(!llvm::verifyModule(module, &llvm::dbgs()));
for (auto func = module.begin(); func != module.end(); ++func) {
function_passes.run(*func);
}
function_passes.doFinalization();
module_passes.run(module);
CHECK(!llvm::verifyModule(module, &llvm::dbgs()));
runtime::RewriteIRRuntimeFunctions(&module, enable_fast_math_);
// Buffer for holding machine code prior to constructing the ObjectFile.
llvm::SmallVector<char, 0> stream_buffer;
llvm::raw_svector_ostream ostream(stream_buffer);
VLOG(2) << "IR after optimizations";
XLA_VLOG_LINES(2, llvm_ir::DumpModuleToString(module));
if (post_optimization_hook_) {
TF_CHECK_OK(post_optimization_hook_(module));
}
// Generate code.
llvm::MCContext* mc_context;
llvm::legacy::PassManager codegen_passes;
target_machine_->addPassesToEmitMC(codegen_passes, mc_context, ostream);
codegen_passes.run(module);
// Construct ObjectFile from machine code buffer.
return std::unique_ptr<llvm::MemoryBuffer>(
new llvm::SmallVectorMemoryBuffer(std::move(stream_buffer)));
}
static std::vector<llvm::VecDesc> VectorFunctionsForTargetLibraryInfoImpl() {
std::vector<llvm::VecDesc> result = {
{"tanhf", runtime::kTanhV4F32SymbolName, 4},
{"llvm.tanh.f32", runtime::kTanhV4F32SymbolName, 4},
{"tanhf", runtime::kTanhV8F32SymbolName, 8},
{"llvm.tanh.f32", runtime::kTanhV8F32SymbolName, 8},
{"expf", runtime::kExpV4F32SymbolName, 4},
{"llvm.exp.f32", runtime::kExpV4F32SymbolName, 4},
{"expf", runtime::kExpV8F32SymbolName, 8},
{"llvm.exp.f32", runtime::kExpV8F32SymbolName, 8},
{"logf", runtime::kLogV4F32SymbolName, 4},
{"llvm.log.f32", runtime::kLogV4F32SymbolName, 4},
{"logf", runtime::kLogV8F32SymbolName, 8},
{"llvm.log.f32", runtime::kLogV8F32SymbolName, 8},
};
return result;
}
void CompilerFunctor::AddTargetInfoPasses(
llvm::legacy::PassManagerBase* passes) const {
llvm::Triple target_triple(target_machine_->getTargetTriple());
auto target_library_info_impl =
MakeUnique<llvm::TargetLibraryInfoImpl>(target_triple);
target_library_info_impl->addVectorizableFunctions(
VectorFunctionsForTargetLibraryInfoImpl());
passes->add(
new llvm::TargetLibraryInfoWrapperPass(*target_library_info_impl));
passes->add(createTargetTransformInfoWrapperPass(
target_machine_->getTargetIRAnalysis()));
}
void CompilerFunctor::AddOptimizationPasses(
llvm::legacy::PassManagerBase* module_passes,
llvm::legacy::FunctionPassManager* function_passes, unsigned opt_level,
unsigned size_level) const {
llvm::PassManagerBuilder builder;
builder.OptLevel = opt_level;
builder.SizeLevel = size_level;
if (opt_level > 1) {
builder.Inliner = llvm::createFunctionInliningPass();
} else {
// Only inline functions marked with "alwaysinline".
builder.Inliner = llvm::createAlwaysInlinerLegacyPass();
}
builder.DisableUnitAtATime = false;
builder.DisableUnrollLoops = opt_level == 0;
builder.LoopVectorize = opt_level > 0 && size_level == 0;
builder.SLPVectorize = opt_level > 1 && size_level == 0;
builder.populateFunctionPassManager(*function_passes);
builder.populateModulePassManager(*module_passes);
}
} // namespace cpu
} // namespace xla
|
