Subversion Repositories QNX 8.QNX8 LLVM/Clang compiler suite

Rev

Details | Last modification | View Log | RSS feed

Rev Author Line No. Line
14 pmbaty 1
//===- llvm/Support/Parallel.h - Parallel algorithms ----------------------===//
2
//
3
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4
// See https://llvm.org/LICENSE.txt for license information.
5
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6
//
7
//===----------------------------------------------------------------------===//
8
 
9
#ifndef LLVM_SUPPORT_PARALLEL_H
10
#define LLVM_SUPPORT_PARALLEL_H
11
 
12
#include "llvm/ADT/STLExtras.h"
13
#include "llvm/Config/llvm-config.h"
14
#include "llvm/Support/Error.h"
15
#include "llvm/Support/MathExtras.h"
16
#include "llvm/Support/Threading.h"
17
 
18
#include <algorithm>
19
#include <condition_variable>
20
#include <functional>
21
#include <mutex>
22
 
23
namespace llvm {
24
 
25
namespace parallel {
26
 
27
// Strategy for the default executor used by the parallel routines provided by
28
// this file. It defaults to using all hardware threads and should be
29
// initialized before the first use of parallel routines.
30
extern ThreadPoolStrategy strategy;
31
 
32
#if LLVM_ENABLE_THREADS
33
#ifdef _WIN32
34
// Direct access to thread_local variables from a different DLL isn't
35
// possible with Windows Native TLS.
36
unsigned getThreadIndex();
37
#else
38
// Don't access this directly, use the getThreadIndex wrapper.
39
extern thread_local unsigned threadIndex;
40
 
41
inline unsigned getThreadIndex() { return threadIndex; }
42
#endif
43
#else
44
inline unsigned getThreadIndex() { return 0; }
45
#endif
46
 
47
namespace detail {
48
class Latch {
49
  uint32_t Count;
50
  mutable std::mutex Mutex;
51
  mutable std::condition_variable Cond;
52
 
53
public:
54
  explicit Latch(uint32_t Count = 0) : Count(Count) {}
55
  ~Latch() {
56
    // Ensure at least that sync() was called.
57
    assert(Count == 0);
58
  }
59
 
60
  void inc() {
61
    std::lock_guard<std::mutex> lock(Mutex);
62
    ++Count;
63
  }
64
 
65
  void dec() {
66
    std::lock_guard<std::mutex> lock(Mutex);
67
    if (--Count == 0)
68
      Cond.notify_all();
69
  }
70
 
71
  void sync() const {
72
    std::unique_lock<std::mutex> lock(Mutex);
73
    Cond.wait(lock, [&] { return Count == 0; });
74
  }
75
};
76
} // namespace detail
77
 
78
class TaskGroup {
79
  detail::Latch L;
80
  bool Parallel;
81
 
82
public:
83
  TaskGroup();
84
  ~TaskGroup();
85
 
86
  // Spawn a task, but does not wait for it to finish.
87
  void spawn(std::function<void()> f);
88
 
89
  // Similar to spawn, but execute the task immediately when ThreadsRequested ==
90
  // 1. The difference is to give the following pattern a more intuitive order
91
  // when single threading is requested.
92
  //
93
  // for (size_t begin = 0, i = 0, taskSize = 0;;) {
94
  //   taskSize += ...
95
  //   bool done = ++i == end;
96
  //   if (done || taskSize >= taskSizeLimit) {
97
  //     tg.execute([=] { fn(begin, i); });
98
  //     if (done)
99
  //       break;
100
  //     begin = i;
101
  //     taskSize = 0;
102
  //   }
103
  // }
104
  void execute(std::function<void()> f);
105
 
106
  void sync() const { L.sync(); }
107
};
108
 
109
namespace detail {
110
 
111
#if LLVM_ENABLE_THREADS
112
const ptrdiff_t MinParallelSize = 1024;
113
 
114
/// Inclusive median.
115
template <class RandomAccessIterator, class Comparator>
116
RandomAccessIterator medianOf3(RandomAccessIterator Start,
117
                               RandomAccessIterator End,
118
                               const Comparator &Comp) {
119
  RandomAccessIterator Mid = Start + (std::distance(Start, End) / 2);
120
  return Comp(*Start, *(End - 1))
121
             ? (Comp(*Mid, *(End - 1)) ? (Comp(*Start, *Mid) ? Mid : Start)
122
                                       : End - 1)
123
             : (Comp(*Mid, *Start) ? (Comp(*(End - 1), *Mid) ? Mid : End - 1)
124
                                   : Start);
125
}
126
 
127
template <class RandomAccessIterator, class Comparator>
128
void parallel_quick_sort(RandomAccessIterator Start, RandomAccessIterator End,
129
                         const Comparator &Comp, TaskGroup &TG, size_t Depth) {
130
  // Do a sequential sort for small inputs.
131
  if (std::distance(Start, End) < detail::MinParallelSize || Depth == 0) {
132
    llvm::sort(Start, End, Comp);
133
    return;
134
  }
135
 
136
  // Partition.
137
  auto Pivot = medianOf3(Start, End, Comp);
138
  // Move Pivot to End.
139
  std::swap(*(End - 1), *Pivot);
140
  Pivot = std::partition(Start, End - 1, [&Comp, End](decltype(*Start) V) {
141
    return Comp(V, *(End - 1));
142
  });
143
  // Move Pivot to middle of partition.
144
  std::swap(*Pivot, *(End - 1));
145
 
146
  // Recurse.
147
  TG.spawn([=, &Comp, &TG] {
148
    parallel_quick_sort(Start, Pivot, Comp, TG, Depth - 1);
149
  });
150
  parallel_quick_sort(Pivot + 1, End, Comp, TG, Depth - 1);
151
}
152
 
153
template <class RandomAccessIterator, class Comparator>
154
void parallel_sort(RandomAccessIterator Start, RandomAccessIterator End,
155
                   const Comparator &Comp) {
156
  TaskGroup TG;
157
  parallel_quick_sort(Start, End, Comp, TG,
158
                      llvm::Log2_64(std::distance(Start, End)) + 1);
159
}
160
 
161
// TaskGroup has a relatively high overhead, so we want to reduce
162
// the number of spawn() calls. We'll create up to 1024 tasks here.
163
// (Note that 1024 is an arbitrary number. This code probably needs
164
// improving to take the number of available cores into account.)
165
enum { MaxTasksPerGroup = 1024 };
166
 
167
template <class IterTy, class ResultTy, class ReduceFuncTy,
168
          class TransformFuncTy>
169
ResultTy parallel_transform_reduce(IterTy Begin, IterTy End, ResultTy Init,
170
                                   ReduceFuncTy Reduce,
171
                                   TransformFuncTy Transform) {
172
  // Limit the number of tasks to MaxTasksPerGroup to limit job scheduling
173
  // overhead on large inputs.
174
  size_t NumInputs = std::distance(Begin, End);
175
  if (NumInputs == 0)
176
    return std::move(Init);
177
  size_t NumTasks = std::min(static_cast<size_t>(MaxTasksPerGroup), NumInputs);
178
  std::vector<ResultTy> Results(NumTasks, Init);
179
  {
180
    // Each task processes either TaskSize or TaskSize+1 inputs. Any inputs
181
    // remaining after dividing them equally amongst tasks are distributed as
182
    // one extra input over the first tasks.
183
    TaskGroup TG;
184
    size_t TaskSize = NumInputs / NumTasks;
185
    size_t RemainingInputs = NumInputs % NumTasks;
186
    IterTy TBegin = Begin;
187
    for (size_t TaskId = 0; TaskId < NumTasks; ++TaskId) {
188
      IterTy TEnd = TBegin + TaskSize + (TaskId < RemainingInputs ? 1 : 0);
189
      TG.spawn([=, &Transform, &Reduce, &Results] {
190
        // Reduce the result of transformation eagerly within each task.
191
        ResultTy R = Init;
192
        for (IterTy It = TBegin; It != TEnd; ++It)
193
          R = Reduce(R, Transform(*It));
194
        Results[TaskId] = R;
195
      });
196
      TBegin = TEnd;
197
    }
198
    assert(TBegin == End);
199
  }
200
 
201
  // Do a final reduction. There are at most 1024 tasks, so this only adds
202
  // constant single-threaded overhead for large inputs. Hopefully most
203
  // reductions are cheaper than the transformation.
204
  ResultTy FinalResult = std::move(Results.front());
205
  for (ResultTy &PartialResult :
206
       MutableArrayRef(Results.data() + 1, Results.size() - 1))
207
    FinalResult = Reduce(FinalResult, std::move(PartialResult));
208
  return std::move(FinalResult);
209
}
210
 
211
#endif
212
 
213
} // namespace detail
214
} // namespace parallel
215
 
216
template <class RandomAccessIterator,
217
          class Comparator = std::less<
218
              typename std::iterator_traits<RandomAccessIterator>::value_type>>
219
void parallelSort(RandomAccessIterator Start, RandomAccessIterator End,
220
                  const Comparator &Comp = Comparator()) {
221
#if LLVM_ENABLE_THREADS
222
  if (parallel::strategy.ThreadsRequested != 1) {
223
    parallel::detail::parallel_sort(Start, End, Comp);
224
    return;
225
  }
226
#endif
227
  llvm::sort(Start, End, Comp);
228
}
229
 
230
void parallelFor(size_t Begin, size_t End, function_ref<void(size_t)> Fn);
231
 
232
template <class IterTy, class FuncTy>
233
void parallelForEach(IterTy Begin, IterTy End, FuncTy Fn) {
234
  parallelFor(0, End - Begin, [&](size_t I) { Fn(Begin[I]); });
235
}
236
 
237
template <class IterTy, class ResultTy, class ReduceFuncTy,
238
          class TransformFuncTy>
239
ResultTy parallelTransformReduce(IterTy Begin, IterTy End, ResultTy Init,
240
                                 ReduceFuncTy Reduce,
241
                                 TransformFuncTy Transform) {
242
#if LLVM_ENABLE_THREADS
243
  if (parallel::strategy.ThreadsRequested != 1) {
244
    return parallel::detail::parallel_transform_reduce(Begin, End, Init, Reduce,
245
                                                       Transform);
246
  }
247
#endif
248
  for (IterTy I = Begin; I != End; ++I)
249
    Init = Reduce(std::move(Init), Transform(*I));
250
  return std::move(Init);
251
}
252
 
253
// Range wrappers.
254
template <class RangeTy,
255
          class Comparator = std::less<decltype(*std::begin(RangeTy()))>>
256
void parallelSort(RangeTy &&R, const Comparator &Comp = Comparator()) {
257
  parallelSort(std::begin(R), std::end(R), Comp);
258
}
259
 
260
template <class RangeTy, class FuncTy>
261
void parallelForEach(RangeTy &&R, FuncTy Fn) {
262
  parallelForEach(std::begin(R), std::end(R), Fn);
263
}
264
 
265
template <class RangeTy, class ResultTy, class ReduceFuncTy,
266
          class TransformFuncTy>
267
ResultTy parallelTransformReduce(RangeTy &&R, ResultTy Init,
268
                                 ReduceFuncTy Reduce,
269
                                 TransformFuncTy Transform) {
270
  return parallelTransformReduce(std::begin(R), std::end(R), Init, Reduce,
271
                                 Transform);
272
}
273
 
274
// Parallel for-each, but with error handling.
275
template <class RangeTy, class FuncTy>
276
Error parallelForEachError(RangeTy &&R, FuncTy Fn) {
277
  // The transform_reduce algorithm requires that the initial value be copyable.
278
  // Error objects are uncopyable. We only need to copy initial success values,
279
  // so work around this mismatch via the C API. The C API represents success
280
  // values with a null pointer. The joinErrors discards null values and joins
281
  // multiple errors into an ErrorList.
282
  return unwrap(parallelTransformReduce(
283
      std::begin(R), std::end(R), wrap(Error::success()),
284
      [](LLVMErrorRef Lhs, LLVMErrorRef Rhs) {
285
        return wrap(joinErrors(unwrap(Lhs), unwrap(Rhs)));
286
      },
287
      [&Fn](auto &&V) { return wrap(Fn(V)); }));
288
}
289
 
290
} // namespace llvm
291
 
292
#endif // LLVM_SUPPORT_PARALLEL_H