Line data Source code
1 : //===- llvm/Support/Parallel.h - Parallel algorithms ----------------------===//
2 : //
3 : // The LLVM Compiler Infrastructure
4 : //
5 : // This file is distributed under the University of Illinois Open Source
6 : // License. See LICENSE.TXT for details.
7 : //
8 : //===----------------------------------------------------------------------===//
9 :
10 : #ifndef LLVM_SUPPORT_PARALLEL_H
11 : #define LLVM_SUPPORT_PARALLEL_H
12 :
13 : #include "llvm/ADT/STLExtras.h"
14 : #include "llvm/Config/llvm-config.h"
15 : #include "llvm/Support/MathExtras.h"
16 :
17 : #include <algorithm>
18 : #include <condition_variable>
19 : #include <functional>
20 : #include <mutex>
21 :
22 : #if defined(_MSC_VER) && LLVM_ENABLE_THREADS
23 : #pragma warning(push)
24 : #pragma warning(disable : 4530)
25 : #include <concrt.h>
26 : #include <ppl.h>
27 : #pragma warning(pop)
28 : #endif
29 :
30 : namespace llvm {
31 :
32 : namespace parallel {
33 : struct sequential_execution_policy {};
34 : struct parallel_execution_policy {};
35 :
36 : template <typename T>
37 : struct is_execution_policy
38 : : public std::integral_constant<
39 : bool, llvm::is_one_of<T, sequential_execution_policy,
40 : parallel_execution_policy>::value> {};
41 :
42 : constexpr sequential_execution_policy seq{};
43 : constexpr parallel_execution_policy par{};
44 :
45 : namespace detail {
46 :
47 : #if LLVM_ENABLE_THREADS
48 :
49 : class Latch {
50 : uint32_t Count;
51 : mutable std::mutex Mutex;
52 : mutable std::condition_variable Cond;
53 :
54 : public:
55 732432 : explicit Latch(uint32_t Count = 0) : Count(Count) {}
56 366323 : ~Latch() { sync(); }
57 :
58 : void inc() {
59 42451 : std::lock_guard<std::mutex> lock(Mutex);
60 42451 : ++Count;
61 : }
62 :
63 64568 : void dec() {
64 64568 : std::lock_guard<std::mutex> lock(Mutex);
65 64571 : if (--Count == 0)
66 10173 : Cond.notify_all();
67 64570 : }
68 :
69 366323 : void sync() const {
70 366323 : std::unique_lock<std::mutex> lock(Mutex);
71 376015 : Cond.wait(lock, [&] { return Count == 0; });
72 366323 : }
73 : };
74 :
75 : class TaskGroup {
76 : Latch L;
77 :
78 : public:
79 : void spawn(std::function<void()> f);
80 :
81 : void sync() const { L.sync(); }
82 : };
83 :
84 : #if defined(_MSC_VER)
85 : template <class RandomAccessIterator, class Comparator>
86 : void parallel_sort(RandomAccessIterator Start, RandomAccessIterator End,
87 : const Comparator &Comp) {
88 : concurrency::parallel_sort(Start, End, Comp);
89 : }
90 : template <class IterTy, class FuncTy>
91 : void parallel_for_each(IterTy Begin, IterTy End, FuncTy Fn) {
92 : concurrency::parallel_for_each(Begin, End, Fn);
93 : }
94 :
95 : template <class IndexTy, class FuncTy>
96 : void parallel_for_each_n(IndexTy Begin, IndexTy End, FuncTy Fn) {
97 : concurrency::parallel_for(Begin, End, Fn);
98 : }
99 :
100 : #else
101 : const ptrdiff_t MinParallelSize = 1024;
102 :
103 : /// Inclusive median.
104 : template <class RandomAccessIterator, class Comparator>
105 1762 : RandomAccessIterator medianOf3(RandomAccessIterator Start,
106 : RandomAccessIterator End,
107 : const Comparator &Comp) {
108 1762 : RandomAccessIterator Mid = Start + (std::distance(Start, End) / 2);
109 1762 : return Comp(*Start, *(End - 1))
110 1762 : ? (Comp(*Mid, *(End - 1)) ? (Comp(*Start, *Mid) ? Mid : Start)
111 : : End - 1)
112 864 : : (Comp(*Mid, *Start) ? (Comp(*(End - 1), *Mid) ? Mid : End - 1)
113 1762 : : Start);
114 : }
115 0 :
116 : template <class RandomAccessIterator, class Comparator>
117 3693 : void parallel_quick_sort(RandomAccessIterator Start, RandomAccessIterator End,
118 0 : const Comparator &Comp, TaskGroup &TG, size_t Depth) {
119 : // Do a sequential sort for small inputs.
120 3693 : if (std::distance(Start, End) < detail::MinParallelSize || Depth == 0) {
121 : llvm::sort(Start, End, Comp);
122 1762 : return;
123 0 : }
124 :
125 0 : // Partition.
126 1762 : auto Pivot = medianOf3(Start, End, Comp);
127 : // Move Pivot to End.
128 1762 : std::swap(*(End - 1), *Pivot);
129 0 : Pivot = std::partition(Start, End - 1, [&Comp, End](decltype(*Start) V) {
130 3908396 : return Comp(V, *(End - 1));
131 : });
132 0 : // Move Pivot to middle of partition.
133 0 : std::swap(*Pivot, *(End - 1));
134 :
135 : // Recurse.
136 1762 : TG.spawn([=, &Comp, &TG] {
137 45 : parallel_quick_sort(Start, Pivot, Comp, TG, Depth - 1);
138 : });
139 1762 : parallel_quick_sort(Pivot + 1, End, Comp, TG, Depth - 1);
140 45 : }
141 :
142 45 : template <class RandomAccessIterator, class Comparator>
143 169 : void parallel_sort(RandomAccessIterator Start, RandomAccessIterator End,
144 : const Comparator &Comp) {
145 : TaskGroup TG;
146 169 : parallel_quick_sort(Start, End, Comp, TG,
147 338 : llvm::Log2_64(std::distance(Start, End)) + 1);
148 169 : }
149 :
150 : template <class IterTy, class FuncTy>
151 5301 : void parallel_for_each(IterTy Begin, IterTy End, FuncTy Fn) {
152 : // TaskGroup has a relatively high overhead, so we want to reduce
153 : // the number of spawn() calls. We'll create up to 1024 tasks here.
154 : // (Note that 1024 is an arbitrary number. This code probably needs
155 : // improving to take the number of available cores into account.)
156 5301 : ptrdiff_t TaskSize = std::distance(Begin, End) / 1024;
157 5301 : if (TaskSize == 0)
158 : TaskSize = 1;
159 0 :
160 : TaskGroup TG;
161 21036 : while (TaskSize < std::distance(Begin, End)) {
162 33090 : TG.spawn([=, &Fn] { std::for_each(Begin, Begin + TaskSize, Fn); });
163 411 : Begin += TaskSize;
164 1 : }
165 4670 : std::for_each(Begin, End, Fn);
166 5302 : }
167 123 :
168 : template <class IndexTy, class FuncTy>
169 15 : void parallel_for_each_n(IndexTy Begin, IndexTy End, FuncTy Fn) {
170 15 : ptrdiff_t TaskSize = (End - Begin) / 1024;
171 15 : if (TaskSize == 0)
172 123 : TaskSize = 1;
173 123 :
174 : TaskGroup TG;
175 : IndexTy I = Begin;
176 1056 : for (; I + TaskSize < End; I += TaskSize) {
177 2598 : TG.spawn([=, &Fn] {
178 3850 : for (IndexTy J = I, E = I + TaskSize; J != E; ++J)
179 2056 : Fn(J);
180 0 : });
181 105 : }
182 153 : for (IndexTy J = I; J < End; ++J)
183 1922 : Fn(J);
184 15 : }
185 44 :
186 0 : #endif
187 0 :
188 1951 : #endif
189 1907 :
190 44 : template <typename Iter>
191 : using DefComparator =
192 0 : std::less<typename std::iterator_traits<Iter>::value_type>;
193 14120 :
194 24402 : } // namespace detail
195 24 :
196 0 : // sequential algorithm implementations.
197 1848 : template <class Policy, class RandomAccessIterator,
198 1907 : class Comparator = detail::DefComparator<RandomAccessIterator>>
199 215 : void sort(Policy policy, RandomAccessIterator Start, RandomAccessIterator End,
200 0 : const Comparator &Comp = Comparator()) {
201 0 : static_assert(is_execution_policy<Policy>::value,
202 0 : "Invalid execution policy!");
203 0 : llvm::sort(Start, End, Comp);
204 215 : }
205 215 :
206 : template <class Policy, class IterTy, class FuncTy>
207 0 : void for_each(Policy policy, IterTy Begin, IterTy End, FuncTy Fn) {
208 0 : static_assert(is_execution_policy<Policy>::value,
209 1393 : "Invalid execution policy!");
210 1977 : std::for_each(Begin, End, Fn);
211 424 : }
212 :
213 118 : template <class Policy, class IndexTy, class FuncTy>
214 260 : void for_each_n(Policy policy, IndexTy Begin, IndexTy End, FuncTy Fn) {
215 371 : static_assert(is_execution_policy<Policy>::value,
216 45 : "Invalid execution policy!");
217 1 : for (IndexTy I = Begin; I != End; ++I)
218 0 : Fn(I);
219 0 : }
220 282 :
221 283 : // Parallel algorithm implementations, only available when LLVM_ENABLE_THREADS
222 1 : // is true.
223 44 : #if LLVM_ENABLE_THREADS
224 : template <class RandomAccessIterator,
225 1393 : class Comparator = detail::DefComparator<RandomAccessIterator>>
226 2268 : void sort(parallel_execution_policy policy, RandomAccessIterator Start,
227 88 : RandomAccessIterator End, const Comparator &Comp = Comparator()) {
228 213 : detail::parallel_sort(Start, End, Comp);
229 278 : }
230 281 :
231 9648 : template <class IterTy, class FuncTy>
232 0 : void for_each(parallel_execution_policy policy, IterTy Begin, IterTy End,
233 350692 : FuncTy Fn) {
234 350790 : detail::parallel_for_each(Begin, End, Fn);
235 350692 : }
236 9648 :
237 9648 : template <class IndexTy, class FuncTy>
238 0 : void for_each_n(parallel_execution_policy policy, IndexTy Begin, IndexTy End,
239 : FuncTy Fn) {
240 370140 : detail::parallel_for_each_n(Begin, End, Fn);
241 53011 : }
242 11336 : #endif
243 :
244 : } // namespace parallel
245 2312 : } // namespace llvm
246 711019 :
247 351695 : #endif // LLVM_SUPPORT_PARALLEL_H
|
