parallel_foreach.hpp 23.6 KB
Newer Older
Sylvain Thery's avatar
Sylvain Thery committed
1 2 3
/*******************************************************************************
* CGoGN: Combinatorial and Geometric modeling with Generic N-dimensional Maps  *
* version 0.1                                                                  *
4
* Copyright (C) 2009-2012, IGG Team, LSIIT, University of Strasbourg           *
Sylvain Thery's avatar
Sylvain Thery committed
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
*                                                                              *
* This library is free software; you can redistribute it and/or modify it      *
* under the terms of the GNU Lesser General Public License as published by the *
* Free Software Foundation; either version 2.1 of the License, or (at your     *
* option) any later version.                                                   *
*                                                                              *
* This library is distributed in the hope that it will be useful, but WITHOUT  *
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or        *
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License  *
* for more details.                                                            *
*                                                                              *
* You should have received a copy of the GNU Lesser General Public License     *
* along with this library; if not, write to the Free Software Foundation,      *
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301 USA.           *
*                                                                              *
20
* Web site: http://cgogn.unistra.fr/                                           *
Sylvain Thery's avatar
Sylvain Thery committed
21 22 23 24 25
* Contact information: cgogn@unistra.fr                                        *
*                                                                              *
*******************************************************************************/
#include <boost/thread.hpp>
#include <boost/thread/barrier.hpp>
sylvain thery's avatar
sylvain thery committed
26
#include <vector>
Sylvain Thery's avatar
Sylvain Thery committed
27 28 29 30 31 32 33 34 35 36 37

namespace CGoGN
{

namespace Algo
{

namespace Parallel
{


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
/**
 *
 */
template<typename MAP>
class ThreadFunction
{
protected:
	std::vector<Dart>& m_darts;
	boost::barrier& m_sync1;
	boost::barrier& m_sync2;
	bool& m_finished;
	FunctorMapThreaded<MAP>* m_functor;
public:
	ThreadFunction(FunctorMapThreaded<MAP>& func, std::vector<Dart>& vd, boost::barrier& s1, boost::barrier& s2, bool& finished, unsigned int id):
		m_darts(vd), m_sync1(s1), m_sync2(s2), m_finished(finished)
	{
		m_functor = func.duplicate(id);
	}

	ThreadFunction(const ThreadFunction<MAP>& tf):
		m_darts(tf.m_darts), m_sync1(tf.m_sync1), m_sync2(tf.m_sync2), m_finished(tf.m_finished), m_functor(tf.m_functor){}

	void operator()()
	{
		while (!m_finished)
		{
			for (std::vector<Dart>::const_iterator it = m_darts.begin(); it != m_darts.end(); ++it)
				m_functor->operator()(*it);
			m_sync1.wait();
			m_sync2.wait();
		}
	}

	void clean()
	{
		delete m_functor;
	}
Sylvain Thery's avatar
Sylvain Thery committed
75

76
};
Sylvain Thery's avatar
Sylvain Thery committed
77

78 79
template<typename MAP, typename T>
class ThreadFunctionResult
Sylvain Thery's avatar
Sylvain Thery committed
80 81
{
protected:
82
	std::vector<Dart>& m_darts;
Sylvain Thery's avatar
Sylvain Thery committed
83 84 85
	boost::barrier& m_sync1;
	boost::barrier& m_sync2;
	bool& m_finished;
86 87
	FunctorMapThreadedResult<MAP,T>* m_functor;
	T& m_result;
Sylvain Thery's avatar
Sylvain Thery committed
88
public:
89 90 91 92 93 94 95 96 97 98 99 100 101
	ThreadFunctionResult(FunctorMapThreadedResult<MAP,T>& func, std::vector<Dart>& vd, boost::barrier& s1, boost::barrier& s2, bool& finished, unsigned int id, T& result):
		m_darts(vd), m_sync1(s1), m_sync2(s2), m_finished(finished), m_result(result)
	{
		m_functor = reinterpret_cast< FunctorMapThreadedResult<MAP,T>* >(func.duplicate(id));
	}

	ThreadFunctionResult(const ThreadFunctionResult<MAP,T>& tf):
		m_darts(tf.m_darts),
		m_sync1(tf.m_sync1),
		m_sync2(tf.m_sync2),
		m_finished(tf.m_finished),
		m_functor(tf.m_functor),
		m_result(tf.m_result){}
Sylvain Thery's avatar
Sylvain Thery committed
102 103 104

	void operator()()
	{
105

Sylvain Thery's avatar
Sylvain Thery committed
106 107 108
		while (!m_finished)
		{
			for (std::vector<Dart>::const_iterator it = m_darts.begin(); it != m_darts.end(); ++it)
109
				m_functor->operator()(*it);
Sylvain Thery's avatar
Sylvain Thery committed
110 111 112
			m_sync1.wait();
			m_sync2.wait();
		}
113 114 115 116 117 118
		m_result = this->m_functor->getResult();
	}

	void clean()
	{
		delete m_functor;
Sylvain Thery's avatar
Sylvain Thery committed
119 120 121
	}
};

122 123


124 125
template <typename PFP, unsigned int ORBIT>
void foreach_orbit(typename PFP::MAP& map, FunctorMapThreaded<typename PFP::MAP>& func, unsigned int nbth, unsigned int szbuff, bool needMarkers, const FunctorSelect& good)
Sylvain Thery's avatar
Sylvain Thery committed
126
{
127 128
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
Sylvain Thery's avatar
Sylvain Thery committed
129 130

	DartMarker dm(map);
131
	Dart d = map.begin();
Sylvain Thery's avatar
Sylvain Thery committed
132 133 134 135 136 137

	// nbth new functions, new thread (with good darts !)
	for (unsigned int i=0; i<nbth; ++i)
		vd[i].reserve(szbuff);

	// fill each vd buffers with 4096 darts
138 139
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
140 141 142
	{
		if (good(d) && (!dm.isMarked(d)))
		{
143
			dm.markOrbit<ORBIT>(d);
Sylvain Thery's avatar
Sylvain Thery committed
144 145 146 147 148 149
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

150 151 152 153 154 155 156 157
	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

Sylvain Thery's avatar
Sylvain Thery committed
158 159
	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
160
	bool finished = false;
Sylvain Thery's avatar
Sylvain Thery committed
161
	// lauch threads
162 163
	if (needMarkers)
	{
164
		for (unsigned int i = 0; i < nbth; ++i)
165 166 167 168
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,1+i));
	}
	else
	{
169
		for (unsigned int i = 0; i < nbth; ++i)
170 171
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,0));
	}
Sylvain Thery's avatar
Sylvain Thery committed
172 173

	// and continue to traverse the map
174 175
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];

176
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
177 178
		tempo[i].reserve(szbuff);

179
	while (d != map.end())
Sylvain Thery's avatar
Sylvain Thery committed
180
	{
181
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
182 183
			tempo[i].clear();

184 185
		unsigned int nb = 0;
		while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
186 187 188
		{
			if (good(d) && (!dm.isMarked(d)))
			{
189
				dm.markOrbit<ORBIT>(d);
Sylvain Thery's avatar
Sylvain Thery committed
190 191 192 193 194 195
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

196
		// sync and swap the two vectors
Sylvain Thery's avatar
Sylvain Thery committed
197
		sync1.wait();
198
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
199 200 201 202 203 204 205 206 207
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	//wait for all theads to be finished
208
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
209 210 211 212
	{
		threads[i]->join();
		delete threads[i];
	}
213 214 215 216 217

	// and release memory
	delete[] threads;
	delete[] vd;
	delete[] tempo;
Sylvain Thery's avatar
Sylvain Thery committed
218 219 220 221 222 223 224 225 226 227 228
}

/**
 * Traverse cells of a map in parallel. Use embedding marker
 * Functor application must be independant
 * @param map the map
 * @param func the functor to apply
 * @param nbth number of thread to use (use twice as threads of processor)
 * @param szbuff size of buffers to store darts in each thread (default is 8192, use less for lower memory consumsion)
 * @param good a selector
 */
229 230
template <typename PFP, unsigned int CELL>
void foreach_cell(typename PFP::MAP& map, FunctorMapThreaded<typename PFP::MAP>& func, unsigned int nbth, unsigned int szbuff, bool needMarkers, const FunctorSelect& good)
Sylvain Thery's avatar
Sylvain Thery committed
231
{
sylvain thery's avatar
sylvain thery committed
232 233
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
Sylvain Thery's avatar
Sylvain Thery committed
234

235 236
	CellMarker<CELL> cm(map);
	Dart d = map.begin();
Sylvain Thery's avatar
Sylvain Thery committed
237 238

	// nbth new functions, new thread (with good darts !)
239
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
240 241 242
		vd[i].reserve(szbuff);

	// fill each vd buffers with 4096 darts
243 244
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
245 246 247 248 249 250 251 252 253 254
	{
		if (good(d) && (!cm.isMarked(d)))
		{
			cm.mark(d);
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

255 256 257 258 259 260 261 262
	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

Sylvain Thery's avatar
Sylvain Thery committed
263 264 265 266
	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
	bool finished=false;
	// lauch threads
267 268
	if (needMarkers)
	{
269
		for (unsigned int i = 0; i < nbth; ++i)
270 271 272 273
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,1+i));
	}
	else
	{
274
		for (unsigned int i = 0; i < nbth; ++i)
275 276
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,0));
	}
Sylvain Thery's avatar
Sylvain Thery committed
277
	// and continue to traverse the map
sylvain thery's avatar
sylvain thery committed
278 279
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];

280
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
281 282
		tempo[i].reserve(szbuff);

283
	while (d != map.end())
Sylvain Thery's avatar
Sylvain Thery committed
284
	{
285
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
286 287
			tempo[i].clear();

288 289
		unsigned int nb = 0;
		while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
290 291 292 293 294 295 296 297 298 299 300
		{
			if (good(d) && (!cm.isMarked(d)))
			{
				cm.mark(d);
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

		sync1.wait();
301
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
302 303 304 305 306 307 308 309 310
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	//wait for all theads to be finished
311
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
312 313 314 315
	{
		threads[i]->join();
		delete threads[i];
	}
sylvain thery's avatar
sylvain thery committed
316 317 318
	delete[] threads;
	delete[] vd;
	delete[] tempo;
Sylvain Thery's avatar
Sylvain Thery committed
319 320
}

Sylvain Thery's avatar
Sylvain Thery committed
321
template <typename PFP>
322
void foreach_dart(typename PFP::MAP& map, FunctorMapThreaded<typename PFP::MAP>& func, unsigned int nbth, unsigned int szbuff, bool needMarkers, const FunctorSelect& good)
Sylvain Thery's avatar
Sylvain Thery committed
323
{
324 325
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
Sylvain Thery's avatar
Sylvain Thery committed
326

327
	Dart d = map.begin();
Sylvain Thery's avatar
Sylvain Thery committed
328 329

	// nbth new functions, new thread (with good darts !)
330
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
331 332 333
		vd[i].reserve(szbuff);

	// fill each vd buffers with 4096 darts
334 335
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
336 337 338 339 340 341 342 343 344
	{
		if (good(d))
		{
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

345 346 347 348 349 350 351 352
	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

Sylvain Thery's avatar
Sylvain Thery committed
353 354
	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
355
	bool finished = false;
Sylvain Thery's avatar
Sylvain Thery committed
356
	// lauch threads
357 358
	if (needMarkers)
	{
359
		for (unsigned int i = 0; i < nbth; ++i)
360 361 362 363
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,1+i));
	}
	else
	{
364
		for (unsigned int i = 0; i < nbth; ++i)
365 366
			threads[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(func, vd[i],sync1,sync2, finished,0));
	}
Sylvain Thery's avatar
Sylvain Thery committed
367
	// and continue to traverse the map
368
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];
369
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
370 371
		tempo[i].reserve(szbuff);

372
	while (d != map.end())
Sylvain Thery's avatar
Sylvain Thery committed
373
	{
374
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
375 376 377
			tempo[i].clear();

		unsigned int nb =0;
378
		while ((d != map.end()) && (nb < nbth*szbuff) )
Sylvain Thery's avatar
Sylvain Thery committed
379 380 381 382 383 384 385 386 387 388
		{
			if (good(d))
			{
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

		sync1.wait();
389
		for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
390 391 392 393 394 395 396 397 398
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	//wait for all theads to be finished
399
	for (unsigned int i = 0; i < nbth; ++i)
Sylvain Thery's avatar
Sylvain Thery committed
400 401 402 403
	{
		threads[i]->join();
		delete threads[i];
	}
404 405 406 407
	
	delete vd;
	delete threads;
	delete tempo;
Sylvain Thery's avatar
Sylvain Thery committed
408 409
}

410 411
template <typename PFP, unsigned int ORBIT, typename T>
void foreach_orbit_res(typename PFP::MAP& map, FunctorMapThreadedResult<typename PFP::MAP, T>& func, unsigned int nbth, unsigned int szbuff, std::vector<T>& results, bool needMarkers, const FunctorSelect& good)
412 413 414
{
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
415
	ThreadFunctionResult<typename PFP::MAP,T>** th_funcs= new ThreadFunctionResult<typename PFP::MAP,T>*[nbth];
416 417

	DartMarker dm(map);
418
	Dart d = map.begin();
419 420

	// nbth new functions, new thread (with good darts !)
421
	for (unsigned int i = 0; i < nbth; ++i)
422 423 424
		vd[i].reserve(szbuff);

	// fill each vd buffers with 4096 darts
425 426
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
427 428 429
	{
		if (good(d) && (!dm.isMarked(d)))
		{
430
			dm.markOrbit<ORBIT>(d);
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

	// prepare some space pour results
	results.resize(nbth);

	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
	bool finished=false;
	// lauch threads
	if (needMarkers)
	{
454
		for (unsigned int i = 0; i < nbth; ++i)
455 456 457 458 459 460 461 462
		{
			// here dynamic allocation to allow the freeing of m_functor (clean) at the end
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,1+i,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}
	else
	{
463
		for (unsigned int i = 0; i < nbth; ++i)
464 465 466 467 468 469 470 471 472
		{
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,0,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}

	// and continue to traverse the map
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];

473
	for (unsigned int i = 0; i < nbth; ++i)
474 475
		tempo[i].reserve(szbuff);

476
	while (d != map.end())
477
	{
478
		for (unsigned int i = 0; i < nbth; ++i)
479 480
			tempo[i].clear();

481 482
		unsigned int nb = 0;
		while ((d != map.end()) && (nb < nbth*szbuff) )
483 484 485
		{
			if (good(d) && (!dm.isMarked(d)))
			{
486
				dm.markOrbit<ORBIT>(d);
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

		// sync and swap the two vectors
		sync1.wait();
		for (unsigned int i=0; i<nbth; ++i)
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	std::vector<T> res;
	//wait for all theads to be finished and get results
506
	for (unsigned int i = 0; i < nbth; ++i)
507 508 509 510 511 512 513 514 515 516 517 518 519
	{
		threads[i]->join();
		delete threads[i];
		th_funcs[i]->clean();
	}

	//release dynamic allocation
	delete[] threads;
	delete[] vd;
	delete[] th_funcs;
	delete[] tempo;
}

520 521
template <typename PFP, unsigned int CELL, typename T>
void foreach_cell_res(typename PFP::MAP& map, FunctorMapThreadedResult<typename PFP::MAP, T>& func,  unsigned int nbth, unsigned int szbuff, std::vector<T>& results, bool needMarkers, const FunctorSelect& good)
522 523 524 525 526
{
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
	ThreadFunctionResult<typename PFP::MAP,T>**  th_funcs= new ThreadFunctionResult<typename PFP::MAP,T>*[nbth];

527 528
	CellMarker<CELL> cm(map);
	Dart d = map.begin();
529 530

	// nbth new functions, new thread (with good darts !)
531
	for (unsigned int i = 0; i < nbth; ++i)
532 533 534
		vd[i].reserve(szbuff);

	// fill each vd buffers with 4096 darts
535 536
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
	{
		if (good(d) && (!cm.isMarked(d)))
		{
			cm.mark(d);
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

	// prepare some space pour results
	results.resize(nbth);

	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
	bool finished=false;
	// lauch threads
	if (needMarkers)
	{
564
		for (unsigned int i = 0; i < nbth; ++i)
565 566 567 568 569 570 571
		{
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,1+i,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}
	else
	{
572
		for (unsigned int i = 0; i < nbth; ++i)
573 574 575 576 577 578 579 580 581
		{
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,0,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}

	// and continue to traverse the map
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];

582
	for (unsigned int i = 0; i < nbth; ++i)
583 584 585 586
		tempo[i].reserve(szbuff);

	while ( d != map.end())
	{
587
		for (unsigned int i = 0; i < nbth; ++i)
588 589
			tempo[i].clear();

590 591
		unsigned int nb = 0;
		while ((d != map.end()) && (nb < nbth*szbuff) )
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
		{
			if (good(d) && (!cm.isMarked(d)))
			{
				cm.mark(d);
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

		sync1.wait();
		for (unsigned int i=0; i<nbth; ++i)
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	std::vector<T> res;
	//wait for all theads to be finished and get results
614
	for (unsigned int i = 0; i < nbth; ++i)
615 616 617 618 619 620 621 622 623 624 625 626 627
	{
		threads[i]->join();
		delete threads[i];
		th_funcs[i]->clean();
	}

	delete[] threads;
	delete[] vd;
	delete[] th_funcs;
	delete[] tempo;
}

template <typename PFP, typename T>
628
void foreach_dart_res(typename PFP::MAP& map, FunctorMapThreadedResult<typename PFP::MAP, T>& func, unsigned int nbth, unsigned int szbuff, std::vector<T>& results, bool needMarkers, const FunctorSelect& good)
629 630 631
{
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	boost::thread** threads = new boost::thread*[nbth];
632
	ThreadFunctionResult<typename PFP::MAP,T>** th_funcs= new ThreadFunctionResult<typename PFP::MAP,T>*[nbth];
633

634
	Dart d = map.begin();
635 636 637 638 639 640

	// nbth new functions, new thread (with good darts !)
	for (unsigned int i=0; i<nbth; ++i)
		vd[i].reserve(szbuff);

	// fill each vd buffers with szbuff darts
641 642
	unsigned int nb = 0;
	while ((d != map.end()) && (nb < nbth*szbuff) )
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
	{
		if (good(d))
		{
			vd[nb%nbth].push_back(d);
			nb++;
		}
		map.next(d);
	}

	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

	// prepare some space pour results
	results.resize(nbth);

	boost::barrier sync1(nbth+1);
	boost::barrier sync2(nbth+1);
665
	bool finished = false;
666 667 668
	// lauch threads
	if (needMarkers)
	{
669
		for (unsigned int i = 0; i < nbth; ++i)
670 671 672 673 674 675 676
		{
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,1+i,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}
	else
	{
677
		for (unsigned int i = 0; i < nbth; ++i)
678 679 680 681 682 683 684 685 686
		{
			th_funcs[i] = new ThreadFunctionResult<typename PFP::MAP,T>(func, vd[i],sync1,sync2, finished,0,results[i]);
			threads[i] = new boost::thread(*(th_funcs[i]));
		}
	}

	// and continue to traverse the map
	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];

687
	for (unsigned int i = 0; i < nbth; ++i)
688 689
		tempo[i].reserve(szbuff);

690
	while (d != map.end())
691 692 693 694
	{
		for (unsigned int i=0; i<nbth; ++i)
			tempo[i].clear();

695 696
		unsigned int nb = 0;
		while ((d != map.end()) && (nb < nbth*szbuff) )
697 698 699 700 701 702 703 704 705 706
		{
			if (good(d))
			{
				tempo[nb%nbth].push_back(d);
				nb++;
			}
			map.next(d);
		}

		sync1.wait();
707
		for (unsigned int i = 0; i < nbth; ++i)
708 709 710 711 712 713 714 715 716 717
			vd[i].swap(tempo[i]);
		sync2.wait();
	}

	sync1.wait();
	finished = true;
	sync2.wait();

	std::vector<T> res;
	//wait for all theads to be finished and get results
718
	for (unsigned int i = 0; i < nbth; ++i)
719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
	{
		threads[i]->join();
		delete threads[i];
		th_funcs[i]->clean();
	}

	delete[] threads;
	delete[] vd;
	delete[] th_funcs;
	delete[] tempo;
}

template <typename T>
T sumResult(const std::vector<T>& res)
{
	T sum(res.front());
735
	typename std::vector<T>::const_iterator it = res.begin();
736 737 738 739 740 741 742 743 744 745 746 747 748 749
	it++;
	while (it != res.end())
	{
		sum += *it;
		++it;
	}
	return sum;
}

template <typename T1, typename T2>
std::pair<T1,T2> sumPairResult(const std::vector< std::pair<T1,T2> >& res)
{
	T1 sum1(res.front().first);
	T2 sum2(res.front().second);
750
	typename std::vector< std::pair<T1,T2> >::const_iterator it = res.begin();
751 752 753 754 755 756 757 758 759 760 761 762 763 764
	it++;
	while (it != res.end())
	{
		sum1 += it->first;
		sum2 += it->second;
		++it;
	}
	return std::pair<T1,T2>(sum1,sum2);
}

template <typename T>
T maxResult(const std::vector<T>& res)
{
	T maxr(res.front());
765
	typename std::vector<T>::const_iterator it = res.begin();
766 767 768 769 770 771 772 773 774 775 776 777 778 779
	it++;
	while (it != res.end())
	{
		if (*it > maxr)
		maxr = *it;
		++it;
	}
	return maxr;
}

template <typename T>
T minResult(const std::vector<T>& res)
{
	T minr(res.front());
780
	typename std::vector<T>::const_iterator it = res.begin();
781 782 783 784 785 786 787 788 789 790
	it++;
	while (it != res.end())
	{
		if (*it < minr)
		minr = *it;
		++it;
	}
	return minr;
}

791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983




/**
 * Traverse cells of a map in parallel. Use embedding marker
 * Functor application must be independant
 * @param map the map
 * @param func the functor to apply
 * @param nbth number of thread to use (use twice as threads of processor)
 * @param szbuff size of buffers to store darts in each thread (default is 8192, use less for lower memory consumsion)
 * @param good a selector
 */
template <typename PFP, unsigned int CELL>
void foreach_cell2Pass(typename PFP::MAP& map, FunctorMapThreaded<typename PFP::MAP>& funcFront, FunctorMapThreaded<typename PFP::MAP>& funcBack, unsigned int nbLoops, unsigned int nbth, unsigned int szbuff, bool needMarkers, const FunctorSelect& good)
{
	std::vector<Dart>* vd = new std::vector<Dart>[nbth];
	for (unsigned int i = 0; i < nbth; ++i)
		vd[i].reserve(szbuff);

	std::vector<Dart>* tempo = new std::vector<Dart>[nbth];
	for (unsigned int i = 0; i < nbth; ++i)
		tempo[i].reserve(szbuff);

	boost::thread** threadsA = new boost::thread*[nbth];
	boost::thread** threadsB = new boost::thread*[nbth];

	if (needMarkers)
	{
		// ensure that there is enough threads
		unsigned int nbth_prec = map.getNbThreadMarkers();
		if (nbth_prec < nbth+1)
			map.addThreadMarker(nbth+1-nbth_prec);
	}

	CellMarkerNoUnmark<CELL> cm(map);

	for (unsigned int loop=0; loop< nbLoops; ++loop)
	{
		// PASS FRONT (A)
		{
			Dart d = map.begin();
			// fill each vd buffers with szbuff darts
			unsigned int nb = 0;
			while ((d != map.end()) && (nb < nbth*szbuff) )
			{
				if (good(d) && (!cm.isMarked(d)))
				{
					cm.mark(d);
					vd[nb%nbth].push_back(d);
					nb++;
				}
				map.next(d);
			}

			boost::barrier sync1(nbth+1);
			boost::barrier sync2(nbth+1);
			bool finished=false;
			// lauch threads
			if (needMarkers)
			{
				for (unsigned int i = 0; i < nbth; ++i)
					threadsA[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(funcFront, vd[i],sync1,sync2, finished,1+i));
			}
			else
			{
				for (unsigned int i = 0; i < nbth; ++i)
					threadsA[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(funcFront, vd[i],sync1,sync2, finished,0));
			}
			// and continue to traverse the map

			while (d != map.end())
			{
				for (unsigned int i = 0; i < nbth; ++i)
					tempo[i].clear();

				unsigned int nb = 0;
				while ((d != map.end()) && (nb < nbth*szbuff) )
				{
					if (good(d) && (!cm.isMarked(d)))
					{
						cm.mark(d);
						tempo[nb%nbth].push_back(d);
						nb++;
					}
					map.next(d);
				}

				sync1.wait();
				for (unsigned int i = 0; i < nbth; ++i)
					vd[i].swap(tempo[i]);
				sync2.wait();
			}

			sync1.wait();
			finished = true;
			sync2.wait();

			//wait for all theads to be finished
			for (unsigned int i = 0; i < nbth; ++i)
				threadsA[i]->join();
		}
		// PASS BACK (B)
		{
			for (unsigned int i = 0; i < nbth; ++i)
				vd[i].clear();
			Dart d = map.begin();
			// fill each vd buffers with szbuff darts
			unsigned int nb = 0;
			while ((d != map.end()) && (nb < nbth*szbuff) )
			{
				if (good(d) && (cm.isMarked(d)))
				{
					cm.unmark(d);
					vd[nb%nbth].push_back(d);
					nb++;
				}
				map.next(d);
			}

			boost::barrier sync1(nbth+1);
			boost::barrier sync2(nbth+1);
			bool finished=false;
			// lauch threads
			if (needMarkers)
			{
				for (unsigned int i = 0; i < nbth; ++i)
					threadsB[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(funcBack, vd[i],sync1,sync2, finished,1+i));
			}
			else
			{
				for (unsigned int i = 0; i < nbth; ++i)
					threadsB[i] = new boost::thread(ThreadFunction<typename PFP::MAP>(funcBack, vd[i],sync1,sync2, finished,0));
			}
			// and continue to traverse the map

			while (d != map.end())
			{
				for (unsigned int i = 0; i < nbth; ++i)
					tempo[i].clear();

				unsigned int nb = 0;
				while ((d != map.end()) && (nb < nbth*szbuff) )
				{
					if (good(d) && (cm.isMarked(d)))
					{
						cm.unmark(d);
						tempo[nb%nbth].push_back(d);
						nb++;
					}
					map.next(d);
				}

				sync1.wait();
				for (unsigned int i = 0; i < nbth; ++i)
					vd[i].swap(tempo[i]);
				sync2.wait();
			}

			sync1.wait();
			finished = true;
			sync2.wait();

			//wait for all theads to be finished
			for (unsigned int i = 0; i < nbth; ++i)
				threadsB[i]->join();
		}
	}

	// free buffers and threads
	for (unsigned int i = 0; i < nbth; ++i)
	{
		delete threadsA[i];
		delete threadsB[i];
	}
	delete[] threadsA;
	delete[] threadsB;
	delete[] vd;
	delete[] tempo;
}













984
} // namespace Parallel
985

986
} // namespace Algo
987

988
} // namespace CGoGN