1[/ 2 / Copyright (c) 2014 Vicente J. Botet Escriba 3 / 4 / Distributed under the Boost Software License, Version 1.0. (See accompanying 5 / file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) 6 /] 7 8[//////////////////////////////////////////////////////////] 9[section:parallel Parallel - Fork-Join -- EXPERIMENTAL] 10 11[section:fork_join Fork-Join] 12 13[warning These features are experimental and subject to change in future versions. There are not too much tests yet, so it is possible that you can find out some trivial bugs :(] 14 15[note These features are based on the [@http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4088.pdf [* n4088 - Task Region R3]] C++1y proposal from P. Halpern, A. Robison, A. Laksberg, H. Sutter, et al. The text that follows has been adapted from this paper to show the differences.] 16 17The major difference respect to the standard proposal is that we are able to use a common executor for several task regions. 18 19[note 20Up to now, Boost.Thread doesn't implement the parallel algorithms as defined in [@http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4105.pdf [* n4105 - Information technology – Programming languages, their environments and system software interfaces – Technical Specification for C++ Extensions for Parallelism]]. 21] 22 23[////////////////////] 24[section Introduction] 25 26 27This module introduces a C++11/c++14 library function template `task_region` and a library class `task_region_handle` 28with member functions `run` and `wait` that together enable developers to write expressive and portable fork-join 29parallel code. 30 31The working draft for the Parallelism TS [@http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4105.pdf [*N4105]] augments the STL algorithms with the inclusion of parallel execution policies. Programmers use these as a basis to write additional high-level algorithms that can be implemented in terms of the provided parallel algorithms. However, the scope of n4105 does not include lower-level mechanisms to express arbitrary fork-join parallelism 32 33The `task_region`, `run` and the `wait` functions provided by this library are based on the `task_group` concept that is a part of the common subset of the PPL and the TBB libraries. 34 35[endsect] [/ Introduction] 36 37 38[/////////////////////////] 39[section:tutorial Tutorial] 40 41Consider an example of a parallel traversal of a tree, where a user-provided function compute is applied to each node of the tree, returning the sum of the results: 42 43 template<typename Func> 44 int traverse(node *n, Func&& compute) 45 { 46 int left = 0, right = 0; 47 task_region([&](task_region_handle& tr) { 48 if (n->left) 49 tr.run([&] { left = traverse(n->left, compute); }); 50 if (n->right) 51 tr.run([&] { right = traverse(n->right, compute); }); 52 }); 53 return compute(n) + left + right; 54 } 55 56The example above demonstrates the use of two of the functions proposed in this paper, `task_region` and 57`task_region_handle::run`. 58The `task_region` function delineates a region in a program code potentially containing invocations of tasks 59spawned by the `run` member function of the `task_region_handle` class. 60 61The run function spawns a task, a unit of work that is allowed to execute in parallel with respect to the caller. 62Any parallel tasks spawned by `run` within the `task_region` are joined back to a single thread of execution at 63the end of the `task_region`. 64 65`run` takes a user-provided function object `f` and starts it asynchronously - i.e. it may return before the 66execution of `f` completes. The implementation's scheduler may choose to run `f` immediately or delay running 67`f` until compute resources become available. 68 69A `task_region_handle` can be constructed only by `task_region` because it has no public constructors. 70Thus, `run` can be invoked (directly or indirectly) only from a user-provided function passed to `task_region`: 71 72 void g(); 73 void f(task_region_handle& tr) 74 { 75 tr.run(g); // OK, invoked from within task_region in h 76 } 77 void h() 78 { 79 task_region(f); 80 } 81 82 int main() 83 { 84 task_region_handle tr; // Error: no public constructor 85 tr.run(g); // No way to call run outside of a task_region 86 return 0; 87 } 88 89[endsect] [/ Tutorial] 90 91[////////////////] 92[section:examples Examples] 93 94[section:fib Parallel Fibonacci] 95 96 97This is surely the worst implementation of the Fibonacci function. Anyway, here it is, as it is simple and shows the fork-join structure clearly. `Fibonacci(n) = Fibonacci(n-1) + Fibonacci(n-2)`, so the task decomposition is trivial. 98 99 int fib_task_region(int n) 100 { 101 using boost::experimental::parallel::task_region; 102 using boost::experimental::parallel::task_region_handle; 103 104 if (n == 0) return 0; 105 if (n == 1) return 1; 106 107 int n1; 108 int n2; 109 110 task_region([&](task_region_handle& trh) 111 { 112 trh.run([&] 113 { 114 n1 = fib_task_region(n - 1); 115 }); 116 117 n2 = fib_task_region(n - 2); 118 }); 119 120 return n1 + n2; 121 } 122 123 int main() 124 { 125 for (int i = 0; i<10; ++i) { 126 std::cout << fib_task_region(i) << " "; 127 } 128 std::cout << std::endl; 129 } 130 131[endsect] [/ Fib] 132[section:fibex Parallel Fibonacci - Specific executor] 133 134The previous example make use of an implementation defined way to spawn the tasks. Often the user wants to master how the task must be spawned. There is an overload of `task_region` that accept an additional `Executor` parameter and a function that takes as parameter a `task_region_handle_gen<Executor>`. `task_region_handle_gen<Executor>` run uses this executor to spawn the tasks. 135 136 template <class Ex> 137 int fib_task_region_gen( Ex& ex, int n) 138 { 139 using boost::experimental::parallel::task_region; 140 using boost::experimental::parallel::task_region_handle_gen; 141 142 if (n == 0) return 0; 143 if (n == 1) return 1; 144 145 int n1; 146 int n2; 147 148 task_region(ex, [&](task_region_handle_gen<Ex>& trh) // (2) 149 { 150 trh.run([&] 151 { 152 n1 = fib_task_region(n - 1); 153 }); 154 155 n2 = fib_task_region(n - 2); 156 }); 157 158 return n1 + n2; 159 } 160 161 int main() 162 { 163 boost::basic_thread_pool tp; // (1) 164 for (int i = 0; i<10; ++i) { 165 std::cout << fib_task_region_gen(tp,i) << " "; 166 } 167 std::cout << std::endl; 168 return 0; 169 } 170 171 172The specific executor is declared in line (1) and it is used in line (2). 173 174[endsect] [/ Fib ex] 175 176 177 178 179[section:quick_sort Parallel Accumulate] 180 181 182[endsect] [/ Accumulate] 183 184[section:quick_sort Parallel Quick Sort] 185 186 187 188[endsect] [/ QuickSort] 189[endsect] [/ Examples] 190 191 192[////////////////////////] 193[section:rationale Design Rationale] 194 195 196[endsect] [/ Design Rationale] 197[endsect] [/ Fork-Join] 198 199[/////////////////////] 200[section:ref Reference -- EXPERIMENTAL] 201 202[/////////////////////////] 203[section:v1 Parallel V1] 204 205[section:exception_list Header `<experimental/exception_list.hpp>`] 206 207 namespace boost 208 { 209 namespace experimental 210 { 211 namespace parallel 212 { 213 inline namespace v1 214 { 215 216 class exception_list; 217 218 } // v1 219 } // parallel 220 } // experimental 221 } // boost 222 223 224[/////////////////////////] 225[section:exception_list Class `exception_list`] 226 227 namespace boost 228 { 229 namespace experimental 230 { 231 namespace parallel 232 { 233 inline namespace v1 234 { 235 236 class exception_list: public std::exception 237 { 238 public: 239 typedef 'implementation defined' const_iterator; 240 241 ~exception_list() noexcept {} 242 243 void add(exception_ptr const& e); 244 size_t size() const noexcept; 245 const_iterator begin() const noexcept; 246 const_iterator end() const noexcept; 247 const char* what() const noexcept; 248 249 }; 250 251 } // v1 252 } // parallel 253 } // experimental 254 } // boost 255 256 257[endsect] [/ exception_list] 258 259[endsect] [/ exception_list.hpp] 260 261[endsect] [/ Parallel V1] 262 263[////////////////////////////////////////////////////////////////////] 264[section:v2 Parallel V2] 265[////////////////////////////////////////////////////////////////////] 266[section:concepts Concepts] 267[////////////////////////////////////////////////////////////////////] 268[section:regionCallable Concept `Region_Callable`] 269 270[endsect] [/ Region_Callable] 271[////////////////////////////////////////////////////////////////////] 272[section:taskCallable Concept `Task_Callable`] 273 274 275[endsect] [/ Task_Callable] 276[////////////////////////////////////////////////////////////////////] 277[endsect] [/ Concepts] 278[////////////////////////////////////////////////////////////////////] 279[section:task_region Header `<experimental/task_region.hpp>`] 280 281 namespace boost 282 { 283 namespace experimental 284 { 285 namespace parallel 286 { 287 inline namespace v2 288 { 289 290 class task_canceled_exception; 291 292 template <class Executor> 293 class task_region_handle_gen; 294 295 using default_executor = 'implementation defined'; 296 297 class task_region_handle; 298 299 template <typename Executor, typename F> 300 void task_region_final(Executor& ex, F&& f); 301 template <typename F> 302 void task_region_final(F&& f); 303 304 template <typename Executor, typename F> 305 void task_region(Executor& ex, F&& f); 306 template <typename F> 307 void task_region(F&& f); 308 309 } // v2 310 } // parallel 311 } // experimental 312 } // boost 313 314 315[////////////////////////////////////////////////////////////////////] 316[section:task_canceled_exception Class `task_canceled_exception `] 317 318 namespace boost 319 { 320 namespace experimental 321 { 322 namespace parallel 323 { 324 inline namespace v2 325 { 326 327 class task_canceled_exception: public std::exception 328 { 329 public: 330 task_canceled_exception() noexcept; 331 task_canceled_exception(const task_canceled_exception&) noexcept; 332 task_canceled_exception& operator=(const task_canceled_exception&) noexcept; 333 virtual const char* what() const noexcept; 334 }; 335 336 } // v2 337 } // parallel 338 } // experimental 339 } // boost 340 341[endsect] [/ task_canceled_exception] 342[////////////////////////////////////////////////////////////////////] 343[section:task_region_handle_gen Template Class `task_region_handle_gen<>`] 344 345 namespace boost 346 { 347 namespace experimental 348 { 349 namespace parallel 350 { 351 inline namespace v2 352 { 353 354 template <class Executor> 355 class task_region_handle_gen 356 { 357 protected: 358 task_region_handle_gen(Executor& ex); 359 360 ~task_region_handle_gen(); 361 362 public: 363 task_region_handle_gen(const task_region_handle_gen&) = delete; 364 task_region_handle_gen& operator=(const task_region_handle_gen&) = delete; 365 task_region_handle_gen* operator&() const = delete; 366 367 template<typename F> 368 void run(F&& f); 369 370 void wait(); 371 }; 372 373 } // v2 374 } // parallel 375 } // experimental 376 } // boost 377 378[endsect] [/ task_region_handle_gen] 379[////////////////////////////////////////////////////////////////////] 380[section:default_executor Class `default_executor `] 381 382 namespace boost 383 { 384 namespace experimental 385 { 386 namespace parallel 387 { 388 inline namespace v2 389 { 390 391 using default_executor = 'implementation defined'; 392 393 } // v2 394 } // parallel 395 } // experimental 396 } // boost 397 398[endsect] [/ default_executor] 399[////////////////////////////////////////////////////////////////////] 400[section:task_region_handle Class `task_region_handle `] 401 402 namespace boost 403 { 404 namespace experimental 405 { 406 namespace parallel 407 { 408 inline namespace v2 409 { 410 411 class task_region_handle : 412 public task_region_handle_gen<default_executor> 413 { 414 protected: 415 task_region_handle(); 416 task_region_handle(const task_region_handle&) = delete; 417 task_region_handle& operator=(const task_region_handle&) = delete; 418 task_region_handle* operator&() const = delete; 419 420 }; 421 422 } // v2 423 } // parallel 424 } // experimental 425 } // boost 426 427[endsect] [/ task_region_handle] 428[////////////////////////////////////////////////////////////////////] 429[section:task_region_final Template Function `task_region_final `] 430 431 namespace boost 432 { 433 namespace experimental 434 { 435 namespace parallel 436 { 437 inline namespace v2 438 { 439 440 template <typename Executor, typename F> 441 void task_region_final(Executor& ex, F&& f); 442 template <typename F> 443 void task_region_final(F&& f); 444 445 } // v2 446 } // parallel 447 } // experimental 448 } // boost 449 450[endsect] [/ task_region_final] 451[////////////////////////////////////////////////////////////////////] 452[section:task_region Template Function `task_region `] 453 454 namespace boost 455 { 456 namespace experimental 457 { 458 namespace parallel 459 { 460 inline namespace v2 461 { 462 463 template <typename Executor, typename F> 464 void task_region(Executor& ex, F&& f); 465 template <typename F> 466 void task_region(F&& f); 467 468 } // v2 469 } // parallel 470 } // experimental 471 } // boost 472 473[endsect] [/ task_region] 474 475 476 477 478[endsect] [/ task_region.hpp] 479[endsect] [/ Parallel V2] 480[endsect] [/ Reference] 481 482[endsect] [/ Parallel] 483