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 explicit Latch(uint32_t Count = 0) : Count(Count) {} 56 ~Latch() { sync(); } 57 58 void inc() { 59 std::lock_guard<std::mutex> lock(Mutex); 60 ++Count; 61 } 62 63 void dec() { 64 std::lock_guard<std::mutex> lock(Mutex); 65 if (--Count == 0) 66 Cond.notify_all(); 67 } 68 69 void sync() const { 70 std::unique_lock<std::mutex> lock(Mutex); 71 Cond.wait(lock, [&] { return Count == 0; }); 72 } 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 RandomAccessIterator medianOf3(RandomAccessIterator Start, 106 RandomAccessIterator End, 107 const Comparator &Comp) { 108 RandomAccessIterator Mid = Start + (std::distance(Start, End) / 2); 109 return Comp(*Start, *(End - 1)) 110 ? (Comp(*Mid, *(End - 1)) ? (Comp(*Start, *Mid) ? Mid : Start) 111 : End - 1) 112 : (Comp(*Mid, *Start) ? (Comp(*(End - 1), *Mid) ? Mid : End - 1) 113 : Start); 114 } 115 116 template <class RandomAccessIterator, class Comparator> 117 void parallel_quick_sort(RandomAccessIterator Start, RandomAccessIterator End, 118 const Comparator &Comp, TaskGroup &TG, size_t Depth) { 119 // Do a sequential sort for small inputs. 120 if (std::distance(Start, End) < detail::MinParallelSize || Depth == 0) { 121 llvm::sort(Start, End, Comp); 122 return; 123 } 124 125 // Partition. 126 auto Pivot = medianOf3(Start, End, Comp); 127 // Move Pivot to End. 128 std::swap(*(End - 1), *Pivot); 129 Pivot = std::partition(Start, End - 1, [&Comp, End](decltype(*Start) V) { 130 return Comp(V, *(End - 1)); 131 }); 132 // Move Pivot to middle of partition. 133 std::swap(*Pivot, *(End - 1)); 134 135 // Recurse. 136 TG.spawn([=, &Comp, &TG] { 137 parallel_quick_sort(Start, Pivot, Comp, TG, Depth - 1); 138 }); 139 parallel_quick_sort(Pivot + 1, End, Comp, TG, Depth - 1); 140 } 141 142 template <class RandomAccessIterator, class Comparator> 143 void parallel_sort(RandomAccessIterator Start, RandomAccessIterator End, 144 const Comparator &Comp) { 145 TaskGroup TG; 146 parallel_quick_sort(Start, End, Comp, TG, 147 llvm::Log2_64(std::distance(Start, End)) + 1); 148 } 149 150 template <class IterTy, class FuncTy> 151 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 ptrdiff_t TaskSize = std::distance(Begin, End) / 1024; 157 if (TaskSize == 0) 158 TaskSize = 1; 159 160 TaskGroup TG; 161 while (TaskSize < std::distance(Begin, End)) { 162 TG.spawn([=, &Fn] { std::for_each(Begin, Begin + TaskSize, Fn); }); 163 Begin += TaskSize; 164 } 165 std::for_each(Begin, End, Fn); 166 } 167 168 template <class IndexTy, class FuncTy> 169 void parallel_for_each_n(IndexTy Begin, IndexTy End, FuncTy Fn) { 170 ptrdiff_t TaskSize = (End - Begin) / 1024; 171 if (TaskSize == 0) 172 TaskSize = 1; 173 174 TaskGroup TG; 175 IndexTy I = Begin; 176 for (; I + TaskSize < End; I += TaskSize) { 177 TG.spawn([=, &Fn] { 178 for (IndexTy J = I, E = I + TaskSize; J != E; ++J) 179 Fn(J); 180 }); 181 } 182 for (IndexTy J = I; J < End; ++J) 183 Fn(J); 184 } 185 186 #endif 187 188 #endif 189 190 template <typename Iter> 191 using DefComparator = 192 std::less<typename std::iterator_traits<Iter>::value_type>; 193 194 } // namespace detail 195 196 // sequential algorithm implementations. 197 template <class Policy, class RandomAccessIterator, 198 class Comparator = detail::DefComparator<RandomAccessIterator>> 199 void sort(Policy policy, RandomAccessIterator Start, RandomAccessIterator End, 200 const Comparator &Comp = Comparator()) { 201 static_assert(is_execution_policy<Policy>::value, 202 "Invalid execution policy!"); 203 llvm::sort(Start, End, Comp); 204 } 205 206 template <class Policy, class IterTy, class FuncTy> 207 void for_each(Policy policy, IterTy Begin, IterTy End, FuncTy Fn) { 208 static_assert(is_execution_policy<Policy>::value, 209 "Invalid execution policy!"); 210 std::for_each(Begin, End, Fn); 211 } 212 213 template <class Policy, class IndexTy, class FuncTy> 214 void for_each_n(Policy policy, IndexTy Begin, IndexTy End, FuncTy Fn) { 215 static_assert(is_execution_policy<Policy>::value, 216 "Invalid execution policy!"); 217 for (IndexTy I = Begin; I != End; ++I) 218 Fn(I); 219 } 220 221 // Parallel algorithm implementations, only available when LLVM_ENABLE_THREADS 222 // is true. 223 #if LLVM_ENABLE_THREADS 224 template <class RandomAccessIterator, 225 class Comparator = detail::DefComparator<RandomAccessIterator>> 226 void sort(parallel_execution_policy policy, RandomAccessIterator Start, 227 RandomAccessIterator End, const Comparator &Comp = Comparator()) { 228 detail::parallel_sort(Start, End, Comp); 229 } 230 231 template <class IterTy, class FuncTy> 232 void for_each(parallel_execution_policy policy, IterTy Begin, IterTy End, 233 FuncTy Fn) { 234 detail::parallel_for_each(Begin, End, Fn); 235 } 236 237 template <class IndexTy, class FuncTy> 238 void for_each_n(parallel_execution_policy policy, IndexTy Begin, IndexTy End, 239 FuncTy Fn) { 240 detail::parallel_for_each_n(Begin, End, Fn); 241 } 242 #endif 243 244 } // namespace parallel 245 } // namespace llvm 246 247 #endif // LLVM_SUPPORT_PARALLEL_H 248
