File: Subdomain.cpp

package info (click to toggle)
yade 2026.1.0-2
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid
  • size: 34,448 kB
  • sloc: cpp: 97,645; python: 52,173; sh: 677; makefile: 162
file content (840 lines) | stat: -rw-r--r-- 31,742 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
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
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
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
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
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
// (c) 2018 Bruno Chareyre <bruno.chareyre@grenoble-inp.fr>
// (c) 2019 Deepak Kunhappan, <deepak.kunhappan@3sr-grenoble.fr> <deepak.kn1990@gmail.com>
#ifdef YADE_MPI

#include "Subdomain.hpp"
#include <lib/serialization/ObjectIO.hpp>
#include <core/BodyContainer.hpp>
#include <core/Interaction.hpp>
#include <core/InteractionContainer.hpp>
#include <core/InteractionLoop.hpp>
#include <core/MPIBodyContainer.hpp>
#include <core/State.hpp>
#include <pkg/common/Sphere.hpp>
#include <boost/iostreams/device/array.hpp>
#include <boost/iostreams/device/back_inserter.hpp>
#include <boost/iostreams/stream.hpp>
#include <boost/iostreams/stream_buffer.hpp>
#include <preprocessing/dem/Shop.hpp>

namespace yade { // Cannot have #include directive inside.

YADE_PLUGIN((Subdomain));
CREATE_LOGGER(Subdomain);

YADE_PLUGIN((Bo1_Subdomain_Aabb) /*(Bo1_Facet_Aabb)(Bo1_Box_Aabb)*/);
CREATE_LOGGER(Bo1_Subdomain_Aabb);

void Subdomain::setMinMax()
{
	Scene* scene(Omega::instance().getScene().get()); // get scene
	                                                  // 	Vector3r min, max;
	Real inf  = std::numeric_limits<Real>::infinity();
	boundsMin = Vector3r(inf, inf, inf);
	boundsMax = Vector3r(-inf, -inf, -inf);
	if (ids.size() == 0) LOG_WARN("empty subdomain!");
	if (ids.size() > 0 and Body::byId(ids[0], scene)->subdomain != scene->subdomain)
		LOG_WARN("setMinMax executed with deprecated data (body->subdomain != scene->subdomain)");

	for (const auto& id : ids) {
		const shared_ptr<Body>& b = Body::byId(id, scene);
		if (!b or !b->bound) continue;
		if (!scene->isPeriodic) {
			boundsMax = boundsMax.cwiseMax(b->bound->max);
			boundsMin = boundsMin.cwiseMin(b->bound->min);
		} else {
			// if periodic, find the period of minbound, find size, wrap minbound based on period and add size to get maxbound (of body)
			Vector3r inVsz = Vector3r(1. / scene->cell->getSize()[0], 1. / scene->cell->getSize()[1], 1. / scene->cell->getSize()[2]);
			Vector3i period(Vector3i::Zero());
			for (int i = 0; i != 3; ++i) {
				period[i] = (int)(math::floor(b->state->pos[i] * inVsz[i]));
			}
			Vector3r wMax;
			Vector3r wMin;
			for (int i = 0; i != 3; ++i) {
				wMin[i] = (period[i]) != 0 ? (b->bound->min[i] / period[i]) : (b->bound->min[i]);
				wMax[i] = (period[i]) != 0 ? (b->bound->max[i] / period[i]) : (b->bound->max[i]);
			}
			boundsMax = boundsMax.cwiseMax(wMax);
			boundsMin = boundsMin.cwiseMin(wMin);
		}
	}
}

// inspired by Integrator::slaves_set (Integrator.hpp)
void Subdomain::intrs_set(const boost::python::list& source)
{
	int len = boost::python::len(source);
	intersections.clear();
	for (int i = 0; i < len; i++) {
		boost::python::extract<std::vector<Body::id_t>> serialGroup(source[i]);
		if (serialGroup.check()) {
			intersections.push_back(serialGroup());
			continue;
		}
		cerr << "  ... failed" << endl;
		PyErr_SetString(PyExc_TypeError, "intersections should be provided as a list of list of ids");
		boost::python::throw_error_already_set();
	}
}

void Subdomain::mIntrs_set(const boost::python::list& source)
{
	int len = boost::python::len(source);
	mirrorIntersections.clear();
	for (int i = 0; i < len; i++) {
		boost::python::extract<std::vector<Body::id_t>> serialGroup(source[i]);
		if (serialGroup.check()) {
			mirrorIntersections.push_back(serialGroup());
			continue;
		}
		cerr << "  ... failed" << endl;
		PyErr_SetString(PyExc_TypeError, "intersections should be provided as a list of list of ids");
		boost::python::throw_error_already_set();
	}
}

boost::python::list Subdomain::intrs_get()
{
	boost::python::list ret;
	for (const auto& grp : intersections) {
		ret.append(boost::python::list(grp));
	}
	return ret;
}

boost::python::list Subdomain::mIntrs_get()
{
	boost::python::list ret;
	for (const auto& grp : mirrorIntersections) {
		ret.append(boost::python::list(grp));
	}
	return ret;
}

void Subdomain::setSubdomainIds(std::vector<Body::id_t> subdIds) { subdomains = subdIds; }

std::vector<Body::id_t> Subdomain::getSubdomainIds() const { return subdomains; }

void Subdomain::append(Body::id_t bId) { subdomains.push_back(bId); }
void Subdomain::appendList(const boost::python::list& lst)
{
	unsigned sz = boost::python::len(lst);
	for (unsigned i = 0; i != sz; ++i) {
		subdomains.push_back(boost::python::extract<int>(lst[i]));
	}
}


void Bo1_Subdomain_Aabb::go(const shared_ptr<Shape>& cm, shared_ptr<Bound>& bv, const Se3r& /*se3*/, const Body* /*b*/)
{
	// 	LOG_WARN("Bo1_Subdomain_Aabb::go()")
	scene             = Omega::instance().getScene().get();
	Subdomain* domain = static_cast<Subdomain*>(cm.get());
	if (!bv) { bv = shared_ptr<Bound>(new Aabb); }
	Aabb* aabb = static_cast<Aabb*>(bv.get());
	aabb->min  = domain->boundsMin;
	aabb->max  = domain->boundsMax;
	return;
}

/********************dpk********************/

std::string Subdomain::serializeMPIBodyContainer(const shared_ptr<MPIBodyContainer>& container)
{
	std::string                                                                 strContainer;
	boost::iostreams::back_insert_device<std::string>                           inserter(strContainer);
	boost::iostreams::stream<boost::iostreams::back_insert_device<std::string>> s(inserter);
	yade::ObjectIO::save<decltype(container), boost::archive::binary_oarchive>(s, "container", container);
	s.flush();
	return strContainer;
}


shared_ptr<MPIBodyContainer> Subdomain::deSerializeMPIBodyContainer(const char* strContainer, int sizeC)
{
	shared_ptr<MPIBodyContainer>                                         container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
	boost::iostreams::basic_array_source<char>                           device(strContainer, sizeC);
	boost::iostreams::stream<boost::iostreams::basic_array_source<char>> s(device);
	yade::ObjectIO::load<decltype(container), boost::archive::binary_iarchive>(s, "container", container);
	return container;
}


string Subdomain::fillContainerGetString(shared_ptr<MPIBodyContainer>& container, const std::vector<Body::id_t>& ids2)
{
	container->insertBodyList(ids2);
	std::string containerString = serializeMPIBodyContainer(container);
	return containerString;
}

string Subdomain::idsToSerializedMPIBodyContainer(const std::vector<Body::id_t>& ids2)
{
	shared_ptr<MPIBodyContainer> container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
	container->insertBodyList(ids2);
	return serializeMPIBodyContainer(container);
}

void Subdomain::clearSubdomainIds() { ids.clear(); }

void Subdomain::setIDstoSubdomain(boost::python::list& idList)
{ //Remark: probably no need for a function to assign a python list to vector<int>, boost::python does this very well.
	unsigned int listSize = boost::python::len(idList);
	for (unsigned int i = 0; i != listSize; ++i) {
		int b_id = boost::python::extract<int>(idList[i]);
		ids.push_back(b_id); // So it's not reset before filling?
	}
}

void Subdomain::getRankSize()
{
	if (!ranksSet) {
		MPI_Comm_rank(selfComm(), &subdomainRank);
		MPI_Comm_size(selfComm(), &commSize);
		ranksSet = true;
	} else {
		return;
	}
}

// driver function for merge operation // workers send bodies, master recieves, sets the bodies into bodycontainer, sets interactions in interactionContainer.

void Subdomain::mergeOp()
{
	getRankSize();
	sendAllBodiesToMaster();
	recvBodyContainersFromWorkers();
	if (subdomainRank == master) {
		Scene* scene           = Omega::instance().getScene().get();
		bool   ifMerge         = true;
		bool   overWriteBodies = true;
		processContainerStrings();
		setBodiesToBodyContainer(scene, recvdBodyContainers, ifMerge, overWriteBodies);
		recvdBodyContainers.clear();
		bodiesSet       = false; // reset flag for next merge op.
		containersRecvd = false;
	}
}

//unused at the moment, can be used to send and recv bodies between subdomains.
void Subdomain::setCommunicationContainers()
{
	//here, we setup the serialized MPIBodyContainer (MPIBodyContainer to string). it is std::vector<std::pair<container(string), sendRank/recvRank> >
	// fill the send container based on the ids from the intersection(local) map
	//Send container
	if (subdomainRank == master) { return; }
	recvRanks.clear();
	sendContainer.clear();
	unsigned int zero = 0;
	for (unsigned int i = 1; i != intersections.size(); ++i) {
		if ((intersections[i].size() == zero) || (i == (unsigned)subdomainRank)) { continue; } // exclude self or if no intersections with others)
		shared_ptr<MPIBodyContainer> container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
		container->subdomainRank    = subdomainRank; // used to identify the origin rank at the reciever side. (maybe not needed?)
		std::string containerString = fillContainerGetString(container, intersections[i]);
		sendContainer.push_back(std::make_pair(containerString, i));
	}
	//Recv container, here we just need the ranks for now.
	for (unsigned int i = 1; i != remoteCount.size(); ++i) {
		if ((static_cast<unsigned int>(subdomainRank) == i) || (!remoteCount[i])) { continue; }
		recvRanks.push_back(i);
	}
	commContainer = true; //flag to check if the communicationContainers are set.
}


void Subdomain::sendContainerString()
{
	//send the containers.
	if (subdomainRank == master) { return; }
	if (!commContainer) {
		LOG_ERROR("communication containers are not set!");
		return;
	}
	for (unsigned int i = 0; i != sendContainer.size(); ++i) {
		MPI_Request request;
		sendString(sendContainer[i].first, sendContainer[i].second, TAG_STRING + sendContainer[i].second, request);
		mpiReqs.push_back(request); //FIXME will not work since we access by index...
	}
}

void Subdomain::processContainerStrings()
{
	//convert the recieved string buffers to MPIBodyContainer.
	recvdBodyContainers.clear();
	if (!containersRecvd) {
		LOG_ERROR("containerStrings not recvd. Fail!");
		return;
	}
	for (unsigned int i = 0; i != recvdStringSizes.size(); ++i) {
		char* cbuf = recvdCharBuff[i];
		int   sz   = recvdStringSizes[i];
		cbuf[sz]   = '\0';
		shared_ptr<MPIBodyContainer> cntr(deSerializeMPIBodyContainer(cbuf, sz));
		recvdBodyContainers.push_back(cntr);
	}
	//free the pointers
	clearRecvdCharBuff(recvdCharBuff);
}


void Subdomain::sendAllBodiesToMaster()
{
	// send all bodies from this subdomain to the master. Can be used for merge.
	// (note to self: this can be improved based on the bisection decomposition.)
	if (subdomainRank == master) { return; }
	shared_ptr<MPIBodyContainer> container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
	std::string                  s = fillContainerGetString(container, ids);
	sendStringBlocking(s, master, TAG_BODY + master);
}

void Subdomain::sendBodies(const int receiver, const vector<Body::id_t>& idsToSend)
{
	shared_ptr<MPIBodyContainer> container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
	std::string                  s = idsToSerializedMPIBodyContainer(idsToSend);
	stringBuff[receiver]           = s;
	MPI_Request req;
	MPI_Isend(stringBuff[receiver].data(), s.size(), MPI_CHAR, receiver, TAG_BODY, selfComm(), &req);
	sendBodyReqs.push_back(req);
}

void Subdomain::receiveBodies(const int sender)
{
	// 	cout<<"receiving in "<<subdomainRank<<" from "<<sender<<" ";
	int   recv_size = probeIncomingBlocking(sender, TAG_BODY);
	char* cbuf      = new char[recv_size + 1];
	recvBuffBlocking(cbuf, recv_size, TAG_BODY, sender);
	cbuf[recv_size] = '\0';
	shared_ptr<MPIBodyContainer> mpiBC(deSerializeMPIBodyContainer(cbuf, recv_size));
	// 	cout<<mpiBC->bContainer.size()<<" bodies"<<endl;
	std::vector<shared_ptr<MPIBodyContainer>> mpiBCVect(1, mpiBC); //setBodiesToBodyContainer needs a vector of MPIBodyContainer, so create one of size 1.
	Scene*                                    scene = Omega::instance().getScene().get();
	setBodiesToBodyContainer(scene, mpiBCVect, false, /*overwriteBodies?*/ true);
	delete[] cbuf;
}

void Subdomain::completeSendBodies()
{
	processReqs(sendBodyReqs); // calls MPI_Wait on the reqs, cleans the vect of mpi Reqs
}

/********Functions exclusive to the master*************/

void Subdomain::initMasterContainer()
{
	if (subdomainRank != master) { return; }
	recvRanks.resize(commSize - 1);
	for (unsigned i = 0; i != recvRanks.size(); ++i) {
		recvRanks[i] = i + 1;
	}
	recvdStringSizes.resize(commSize - 1);
	recvdCharBuff.resize(commSize - 1);
	allocContainerMaster = true;
}

void Subdomain::recvBodyContainersFromWorkers()
{
	if (subdomainRank != master) {
		return;
	} else if (subdomainRank == master) {
		//if (! allocContainerMaster){ initMasterContainer(); std::cout << "init Done in  MASTER " << subdomainRank << std::endl;}
		for (int sourceRank = 1; sourceRank != commSize; ++sourceRank) {
			int sz                           = probeIncomingBlocking(sourceRank, TAG_BODY + subdomainRank);
			recvdStringSizes[sourceRank - 1] = sz;
			int   sztmp                      = sz + 1;
			char* cbuf                       = new char[sztmp];
			recvBuffBlocking(cbuf, sz, TAG_BODY + subdomainRank, sourceRank);
			recvdCharBuff[sourceRank - 1] = cbuf;
		}
		containersRecvd = true;
	}
}

//
// set all body properties from the recvd MPIBodyContainer
void Subdomain::setBodiesToBodyContainer(Scene* scene, std::vector<shared_ptr<MPIBodyContainer>>& containers, bool ifMerge, bool overwriteBodies)
{
	// to be used when deserializing a recieved container.
	shared_ptr<BodyContainer>& bodyContainer = scene->bodies;
	for (unsigned int i = 0; i != containers.size(); ++i) {
		shared_ptr<MPIBodyContainer>&  mpiContainer = containers[i];
		std::vector<shared_ptr<Body>>& bContainer   = mpiContainer->bContainer;
		for (auto bIter = bContainer.begin(); bIter != bContainer.end(); ++bIter) {
			const shared_ptr<Body>& newBody = *(bIter);
			// check if the body already exists in the existing bodycontainer
			const Body::id_t&                             idx        = newBody->id;
			std::map<Body::id_t, shared_ptr<Interaction>> intrsToSet = newBody->intrs;
			shared_ptr<Body>&                             b          = (*bodyContainer)[idx];
			if (!b) newBody->intrs.clear(); //we can clear here, interactions are stored in intrsToSet and will be reinserted
			else
				newBody->intrs = b->intrs; //keep interactions generated in current subdomain as they may not exist on the sender's side

			if (ifMerge) newBody->material = scene->materials[newBody->material->id];

			if (!b) bodyContainer->insertAtId(newBody, newBody->id);
			else if (overwriteBodies) {
				b                                = newBody;
				bodyContainer->dirty             = true;
				bodyContainer->checkedByCollider = false;
			}

			for (auto mapIter = intrsToSet.begin(); mapIter != intrsToSet.end(); ++mapIter) {
				const Body::id_t& id1 = mapIter->second->id1;
				const Body::id_t& id2 = mapIter->second->id2;
				if (ifMerge) {
					if ((*bodyContainer)[id1] and (*bodyContainer)[id2]) { // we will insert interactions only when both bodies are inserted
						/* FIXME: we should make really sure that we are not overwriting a live interaction with a deprecated one (possible solution: make all interactions between remote bodies virtual)*/
						scene->interactions->insertInteractionMPI(mapIter->second);
					}
				} else {
					if ((*bodyContainer)[id1] and (*bodyContainer)[id2]
					    and ((*bodyContainer)[id1]->subdomain == scene->subdomain
					         or (*bodyContainer)[id2]->subdomain == scene->subdomain)) {
						// we will insert interactions only when both bodies are inserted
						/* FIXME: we should make really sure that we are not overwriting a live interaction with a deprecated one (possible solution: make all interactions between remote bodies virtual)*/
						scene->interactions->insertInteractionMPI(mapIter->second);
					}
				}
			}
		}
	}
	containers.clear();
	bodiesSet = true;
}


void Subdomain::splitBodiesToWorkers(const bool& eraseWorkerIds)
{
	if (!eraseWorkerIds) { return; }
	shared_ptr<Scene>                    scene         = Omega::instance().getScene();
	shared_ptr<BodyContainer>            bodyContainer = scene->bodies;
	std::vector<std::vector<Body::id_t>> idsToSend;

	if (subdomainRank == 0) {
		idsToSend.resize(commSize - 1);
		for (const auto& b : bodyContainer->body) {
			if (!b->getIsSubdomain()) {
				if (b->subdomain != master) { idsToSend[b->subdomain - 1].push_back(b->id); }
				for (const auto& bIntrs : b->intrs) {
					Body::id_t              otherId;
					shared_ptr<Interaction> I = bIntrs.second;
					if (b->id == I->getId1()) {
						otherId = I->getId2();
					} else {
						otherId = I->getId1();
					}
					const shared_ptr<Body>& otherBody = (*bodyContainer)[otherId];
					if (otherBody->getIsSubdomain()) { idsToSend[otherBody->subdomain - 1].push_back(b->id); }
				}
			}
		}
	}

	if ((subdomainRank != master) && (eraseWorkerIds)) {
		for (const auto& b : bodyContainer->body) {
			if (!b) { continue; }
			if (b->subdomain == 0) { continue; }
			if (!b->getIsSubdomain()) { bodyContainer->erase(b->id, true); }
		}
		clearSubdomainIds();
	}

	if (subdomainRank == master) {
		for (unsigned rnk = 0; rnk != idsToSend.size(); ++rnk) {
			const std::vector<Body::id_t>& workerIds = idsToSend[rnk];
			shared_ptr<MPIBodyContainer>   container(shared_ptr<MPIBodyContainer>(new MPIBodyContainer()));
			std::string                    s = fillContainerGetString(container, workerIds);
			sendStringBlocking(s, rnk + 1, TAG_BODY);
		}
	}


	if (subdomainRank != master) {
		receiveBodies(master);
		for (const auto& b : bodyContainer->body) {
			if (!b) { continue; }
			if (!b->getIsSubdomain()) {
				if (b->subdomain == subdomainRank) { ids.push_back(b->id); }
			}
		}
	}
}


/*********************communication functions**************/
//blocking  send and recv

void Subdomain::sendStringBlocking(std::string& s, int destRank, int tag) { MPI_Send(s.data(), s.size(), MPI_CHAR, destRank, tag, selfComm()); }

int Subdomain::probeIncomingBlocking(int sourceRank, int tag)
{
	MPI_Status status;
	MPI_Probe(sourceRank, tag, selfComm(), &status);
	int sz;
	MPI_Get_count(&status, MPI_CHAR, &sz);
	return sz;
}

void Subdomain::recvBuffBlocking(char* cbuf, int cbufSz, int tag, int sourceRank)
{
	MPI_Status status;
	MPI_Recv(cbuf, cbufSz, MPI_CHAR, sourceRank, tag, selfComm(), &status);
}


//non-blocking calls  --> Isend, Iprobe, Irecv, Waitall();  (we will use mpi_isend + mpi_wait and then mpi_probe followed by mpi_recv)

void Subdomain::sendString(std::string& s, int destRank, int tag, MPI_Request& request)
{
	int len = s.size();
	MPI_Isend(s.data(), len, MPI_CHAR, destRank, tag, selfComm(), &request); //
}

int Subdomain::probeIncoming(int sourceRank, int tag)
{
	int        flag = 0;
	MPI_Status status;
	while (!flag) {
		MPI_Iprobe(sourceRank, tag, selfComm(), &flag, &status);
	}
	int sz;
	MPI_Get_count(&status, MPI_CHAR, &sz);
	return sz;
}

void Subdomain::recvBuff(char* cbuf, int cbufsZ, int sourceRank, MPI_Request& request)
{
	MPI_Irecv(cbuf, cbufsZ, MPI_CHAR, sourceRank, TAG_STRING + subdomainRank, selfComm(), &request);
}

void Subdomain::processReqs(std::vector<MPI_Request>& mpiReqs2)
{
	// mpiReqs shadows a member yade::Subdomain::mpiReqs
	if (!mpiReqs2.size()) { return; }
	for (unsigned int i = 0; i != mpiReqs2.size(); ++i) {
		MPI_Status status;
		MPI_Wait(&mpiReqs2[i], &status);
	}

	resetReqs(mpiReqs2);
}

void Subdomain::resetReqs(std::vector<MPI_Request>& mpiReqs2) { mpiReqs2.clear(); }

void Subdomain::processReqsAll(std::vector<MPI_Request>& mpiReqs2, std::vector<MPI_Status>& mpiStats)
{
	for (unsigned int i = 0; i != mpiReqs2.size(); ++i) {
		//MPI_Status status;
		MPI_Waitall(1, &mpiReqs2[i], &mpiStats[i]);
	}
	mpiStats.clear();
	resetReqs(mpiReqs2);
}


void Subdomain::clearRecvdCharBuff(std::vector<char*>& rcharBuff)
{
	for (std::vector<char*>::iterator cIter = rcharBuff.begin(); cIter != rcharBuff.end(); ++cIter) {
		delete[](*cIter);
	}
	if (subdomainRank != master) { rcharBuff.clear(); } // assuming master alwasy recieves from workers, hence the size of this vector for master is fixed.
}

void Subdomain::getMirrorIntersections()
{
	/* warning : local intersections have to be generated first. */

	std::vector<MPI_Request> interReqs;
	mirrorIntersections.clear();
	mirrorIntersections.resize(commSize);

	//workers exchange their intersections.
	if (subdomainRank != master) {
		assert(intersections[subdomainRank].size());
		//get procs to communicate.
		const auto& interProcs = intersections[subdomainRank];
		for (const auto& proc : interProcs) {
			if (proc == master) continue;
			const auto& interVec = intersections[proc];
			MPI_Request req;
			//send the intersections
			MPI_Isend(&interVec.front(), (int)interVec.size(), MPI_INT, proc, TAG_INTERSECTIONS, selfComm(), &req);
			interReqs.push_back(req);
		}

		// probe size of incoming intersections :
		for (const auto& proc : interProcs) {
			if (proc == master) continue;
			MPI_Status status;
			MPI_Probe(proc, TAG_INTERSECTIONS, selfComm(), &status);
			int sz;
			MPI_Get_count(&status, MPI_INT, &sz);
			auto& mirrorVec = mirrorIntersections[proc];
			mirrorVec.resize(sz);
		}

		// recv intersections..
		for (const auto& proc : interProcs) {
			if (proc == master) continue;
			auto&      mirrorVec = mirrorIntersections[proc];
			MPI_Status stat;
			MPI_Recv(&mirrorVec.front(), (int)mirrorVec.size(), MPI_INT, proc, TAG_INTERSECTIONS, selfComm(), &stat);
		}
		//complete the interactive send.
		processReqs(interReqs);
	}

	//get intesections from master
	std::vector<int> intrSzMaster;
	if (subdomainRank == master) {
		for (const auto& vec : intersections) {
			intrSzMaster.push_back((int)vec.size());
		}
	} else {
		intrSzMaster.resize(commSize);
	}

	//master bcasts it's size of intersections
	MPI_Bcast(&intrSzMaster.front(), commSize, MPI_INT, master, selfComm());

	// master sends intersections to those procs with size
	if (subdomainRank == master) {
		interReqs.clear();
		int prc = 0;
		for (auto& inters : intersections) {
			if (inters.size() and prc != subdomainRank) {
				MPI_Request req;
				MPI_Isend(&inters.front(), (int)inters.size(), MPI_INT, prc, TAG_INTERSECTIONS, selfComm(), &req);
				interReqs.push_back(req);
			}
			++prc;
		}
	}
	// workers with intersection with master receives.
	if (subdomainRank != master) {
		if (intrSzMaster[subdomainRank] > 0) {
			const auto& it = std::find(intersections[subdomainRank].begin(), intersections[subdomainRank].end(), master);
			if (it == intersections[subdomainRank].end()) intersections[subdomainRank].push_back(master);
			auto& vecMaster = mirrorIntersections[0];
			vecMaster.clear();
			vecMaster.resize(intrSzMaster[subdomainRank]);
			// recv
			MPI_Status stat;
			MPI_Recv(&vecMaster.front(), (int)vecMaster.size(), MPI_INT, master, TAG_INTERSECTIONS, selfComm(), &stat);
		}
	}
	//complete the interactive send in master's side.
	if (subdomainRank == master) { processReqs(interReqs); }
}


/* Migrate bodies, translation of python functions  */

Real Subdomain::boundOnAxis(Bound& b, const Vector3r& direction, bool min)
        const //return projected extremum of an AABB in a particular direction (in the the '+' or '-' sign depending on 'min' )
{
	Vector3r size     = b.max - b.min;
	Real     extremum = 0;
	for (unsigned k = 0; k < 3; k++)
		extremum += math::abs(size[k] * direction[k]); // this is equivalent to taking the vertex maximizing projected length
	if (min) extremum = -extremum;
	extremum
	        += (b.max + b.min)
	                   .dot(direction); // should be *0.5 to be center of the box, but since we use 'size' instead of half-size everything is doubled, neutral in terms of ordering the boxes
	return 0.5 * extremum;
}

std::vector<projectedBoundElem> Subdomain::projectedBoundsCPP(int otherSD, const Vector3r& otherSubDCM, const Vector3r& subDCM, bool useAABB)
{
	std::vector<projectedBoundElem> pos;

	const shared_ptr<Scene>& scene              = Omega::instance().getScene();
	const shared_ptr<Body>&  otherSubdomainBody = (*scene->bodies)[subdomains[otherSD - 1]];
	if (!otherSubdomainBody) {
		LOG_ERROR("invalid subdomain id, perhaps not in intersection?, other subd =  " << otherSD);
		return pos;
	}
	//const shared_ptr<Subdomain>& otherSubD = YADE_PTR_CAST<Subdomain>(otherSubdomainBody->shape);
	Vector3r pt1, pt2, axis;

	if (useAABB) {
		const auto& otherSubDBound = otherSubdomainBody->bound;
		const auto& thisSubDBound  = (*scene->bodies)[subdomains[subdomainRank - 1]]->bound;
		pt1                        = 0.5 * (thisSubDBound->min + thisSubDBound->max);
		pt2                        = 0.5 * (otherSubDBound->min + otherSubDBound->max);
	} else {
		pt1 = subDCM;
		pt2 = otherSubDCM;
	}

	axis = pt2 - pt1;
	axis.normalize();

	// from intersections (bodies in this subdomain which has intersections with the other sd)
	for (const auto& bId : intersections[otherSD]) {
		const shared_ptr<Body>& b = (*scene->bodies)[bId];
		if (!b or b->getIsSubdomain()) { continue; }
		Real               ps = boundOnAxis((*b->bound), axis, true);
		projectedBoundElem pElem(ps, std::make_pair(subdomainRank, bId));
		pos.push_back(pElem);
	}

	// from mirror intersections (bodies from other subdomain which has intersections with this sd)
	for (const auto& bId : mirrorIntersections[otherSD]) {
		const shared_ptr<Body>& b = (*scene->bodies)[bId];
		if (!b or b->getIsSubdomain()) { continue; }
		Real               ps = boundOnAxis((*b->bound), axis, false);
		projectedBoundElem pElem(ps, std::make_pair(otherSD, bId));
		pos.push_back(pElem);
	}

	// sort
	std::sort(pos.begin(), pos.end(), [](const auto& p1, const auto& p2) { return p1.first < p2.first; });
	return pos;
}

std::vector<Body::id_t> Subdomain::medianFilterCPP(int otherSD, const Vector3r& otherSubDCM, const Vector3r& subDCM, int giveAway, bool useAABB)
{
	std::vector<Body::id_t>         idsToSend;
	std::vector<projectedBoundElem> pos = projectedBoundsCPP(otherSD, otherSubDCM, subDCM, useAABB);
	if (!pos.size()) LOG_ERROR("ERROR IN CALCULATING PROJECTED BOUNDS WITH SUBDOMAIN = " << otherSD << "  from Subdomain = " << subdomainRank);
	int firstJ = pos.size();
	int lastI  = 0;
	for (int n = 0; n < (int)pos.size(); n++) {
		if (pos[n].second.first != subdomainRank and n < firstJ) firstJ = n;
		if (pos[n].second.first == subdomainRank) lastI = n;
	}
	int finalSize = std::max(0, (int)intersections[otherSD].size() - giveAway); // The desired final number on this side
	if (finalSize > lastI) finalSize = lastI + 1;
	if (finalSize < firstJ) finalSize = firstJ + 1;
	for (int x = finalSize; x < (int)pos.size(); x++) // whatever is on the other side is given away
		if (pos[x].second.first == subdomainRank) idsToSend.push_back(pos[x].second.second);
	return idsToSend;
}


void Subdomain::migrateBodiesSend(const std::vector<Body::id_t>& sendIds, int destination)
{
	const shared_ptr<Scene>& scene    = Omega::instance().getScene();
	const auto&              thisSubd = subdomains[subdomainRank - 1];
	for (const auto& bId : sendIds) {
		const shared_ptr<Body>& bdy = (*scene->bodies)[bId];
		if (!bdy) { LOG_ERROR("reassignBodies failed " << bId << "  is not in subdomain " << subdomainRank << std::endl); }
		bdy->subdomain = destination;
		InteractionLoop::createExplicitInteraction(thisSubd, bId, false, true);
	}
	sendBodies(destination, sendIds);
}

// void Subdomain::migrateBodiesRecv()

void Subdomain::updateLocalIds(bool eraseRemoteMaster)
{
	/* in case of the master proc and not eraseRemoteMaster  the worker ids are updated */
	if (subdomainRank != master) {
		const shared_ptr<Scene>& scene = Omega::instance().getScene();
		ids.clear();
		for (const auto& b : (*scene->bodies)) {
			if (!b) { continue; }
			if ((b->subdomain == subdomainRank) && (!(b->getIsSubdomain()))) { ids.push_back(b->id); }
		}
	}
	if (!eraseRemoteMaster) {
		MPI_Status  iSendstat;
		MPI_Request iSendReq;
		if (subdomainRank != master) { MPI_Isend(&ids.front(), (int)ids.size(), MPI_INT, master, 500, selfComm(), &iSendReq); }

		if (subdomainRank == master) {
			std::vector<std::vector<Body::id_t>> workerIdsVec;
			workerIdsVec.resize(commSize - 1);
			int worker = 1;
			for (auto& workerId : workerIdsVec) {
				MPI_Status status;
				MPI_Probe(worker, 500, selfComm(), &status);
				int sz;
				MPI_Get_count(&status, MPI_INT, &sz);
				workerId.resize(sz);
				MPI_Recv(&workerId.front(), sz, MPI_INT, worker, 500, selfComm(), &status);
				++worker;
			}
			// in master
			const shared_ptr<Scene>& scene = Omega::instance().getScene();
			worker                         = 1;
			for (const auto& workerIds : workerIdsVec) {
				for (const auto& bId : workerIds) {
					(*scene->bodies)[bId]->subdomain = worker;
				}
				const auto& workerSubD = YADE_PTR_CAST<Subdomain>((*scene->bodies)[subdomains[worker - 1]]->shape);
				workerSubD->ids        = workerIds;
				++worker;
			}
		}

		if (subdomainRank != master) { MPI_Wait(&iSendReq, &iSendstat); }
	}
}

void Subdomain::cleanIntersections(int otherDomain)
{
	std::vector<Body::id_t>  ints;
	const shared_ptr<Scene>& scene = Omega::instance().getScene();
	for (const auto& bId : intersections[otherDomain]) {
		const shared_ptr<Body>& b = (*scene->bodies)[bId];
		if (b && (b->subdomain == subdomainRank)) ints.push_back(bId);
	}
	intersections[otherDomain] = ints;
}

void Subdomain::updateNewMirrorIntrs(int otherDomain, const std::vector<Body::id_t>& newMirror) { mirrorIntersections[otherDomain] = newMirror; }


// Count interactions of a body with given subdomain
unsigned Subdomain::countIntsWith(Body::id_t body, Body::id_t someSubD, const shared_ptr<Scene>& scene) const
{
	if (not Body::byId(body, scene)) {
		LOG_WARN("invalid body id: " << body << " vs. sd " << someSubD << ", compared in " << subdomainRank);
		return 0;
	}
	const auto& intrs = Body::byId(body, scene)->intrs;
	return std::count_if(intrs.begin(), intrs.end(), [&](auto i) {
		assert(scene->bodies->exists(i.first));
		return (Body::byId(i.first, scene)->subdomain == someSubD and not Body::byId(i.first, scene)->getIsSubdomain());
	});
}

vector<Body::id_t> Subdomain::filteredInts(Body::id_t someSubD, bool mirror) const
{
	auto&                    intrs = mirror ? mirrorIntersections[someSubD] : intersections[someSubD];
	std::vector<Body::id_t>  filtered;
	const shared_ptr<Scene>& scene = Omega::instance().getScene();
	std::copy_if(intrs.begin(), intrs.end(), std::back_inserter(filtered), [&](auto i) {
		return (this->countIntsWith(i, mirror ? scene->subdomain : someSubD, scene) > 0);
	});
	return filtered;
}

double Subdomain::filterIntersections()
{
	// we don't touch intersections with zero yet, unsure it would work directly as it's a bit special
	assert(intersections.size() == mirrorIntersections.size());
	const shared_ptr<Scene>& scene = Omega::instance().getScene();
	assert(scene->subdomain > 0); // this function should not be called by master
	unsigned oldNum(0), newNum(0);
	for (Body::id_t subd = 1; unsigned(subd) < intersections.size(); subd++)
		if (subd != scene->subdomain) {
			// PLEASE DON'T REMOVE THOSE COMMENTED DEBUG MESSAGES
			// 			unsigned oldI(intersections[subd].size()), oldM(mirrorIntersections[subd].size());
			oldNum += intersections[subd].size();
			if (mirrorIntersections[subd].size() > 0) mirrorIntersections[subd] = filteredInts(subd, true);
			if (intersections[subd].size() > 0) intersections[subd] = filteredInts(subd, false);
			newNum += intersections[subd].size();
			// 			LOG_WARN("SubD "<<scene->subdomain<<" suppressed "<<(oldI-intersections[subd].size())<<" / "<<oldI<<"  vs. "<<subd);
			// 			LOG_WARN("SubD "<<scene->subdomain<<" suppressed "<<(oldM-mirrorIntersections[subd].size())<<" / "<<oldM<<" from "<<subd);
		}
	// 		LOG_WARN("SubD "<<scene->subdomain<<" suppressed "<<oldNum-newNum<<" / "<<oldNum);
	return (oldNum ? (oldNum - newNum) / double(oldNum) : 0); //return overall ratio of removed elements (low means useless)
}


} // namespace yade

#endif