Reference documentation for deal.II version 9.3.3
\(\newcommand{\dealvcentcolon}{\mathrel{\mathop{:}}}\) \(\newcommand{\dealcoloneq}{\dealvcentcolon\mathrel{\mkern-1.2mu}=}\) \(\newcommand{\jump}[1]{\left[\!\left[ #1 \right]\!\right]}\) \(\newcommand{\average}[1]{\left\{\!\left\{ #1 \right\}\!\right\}}\)
task_info.cc
Go to the documentation of this file.
1// ---------------------------------------------------------------------
2//
3// Copyright (C) 2018 - 2021 by the deal.II authors
4//
5// This file is part of the deal.II library.
6//
7// The deal.II library is free software; you can use it, redistribute
8// it, and/or modify it under the terms of the GNU Lesser General
9// Public License as published by the Free Software Foundation; either
10// version 2.1 of the License, or (at your option) any later version.
11// The full text of the license can be found in the file LICENSE.md at
12// the top level directory of deal.II.
13//
14// ---------------------------------------------------------------------
15
16
17#include <deal.II/base/conditional_ostream.h>
18#include <deal.II/base/memory_consumption.h>
19#include <deal.II/base/mpi.h>
20#include <deal.II/base/multithread_info.h>
21#include <deal.II/base/parallel.h>
22#include <deal.II/base/utilities.h>
23
24#include <deal.II/matrix_free/task_info.h>
25
26
27#ifdef DEAL_II_WITH_TBB
28# include <tbb/blocked_range.h>
29# include <tbb/parallel_for.h>
30# include <tbb/task.h>
31# ifndef DEAL_II_TBB_WITH_ONEAPI
32# include <tbb/task_scheduler_init.h>
33# endif
34#endif
35
36#include <iostream>
37#include <set>
38
39//
40// TBB with oneAPI API has deprecated and removed the
41// <code>tbb::tasks</code> backend. With this it is no longer possible to
42// compile the following code that builds a directed acyclic graph (DAG) of
43// (thread parallel) tasks without a major porting effort. It turned out
44// that such a dynamic handling of dependencies and structures is not as
45// competitive as initially assumed. Consequently, this part of the matrix
46// free infrastructure has seen less attention than the rest over the last
47// years and is (presumably) not used that often.
48//
49// In case of detected oneAPI backend we simply disable threading in the
50// matrix free backend for now.
51//
52// Matthias Maier, Martin Kronbichler, 2021
53//
54
55DEAL_II_NAMESPACE_OPEN
56
57
58
59/*-------------------- Implementation of the matrix-free loop --------------*/
60namespace internal
61{
62 namespace MatrixFreeFunctions
63 {
64#if defined(DEAL_II_WITH_TBB) && !defined(DEAL_II_TBB_WITH_ONEAPI)
65
66 // This defines the TBB data structures that are needed to schedule the
67 // partition-partition variant
68
69 namespace partition
70 {
71 class ActualCellWork
72 {
73 public:
74 ActualCellWork(MFWorkerInterface **worker_pointer,
75 const unsigned int partition,
76 const TaskInfo & task_info)
77 : worker(nullptr)
78 , worker_pointer(worker_pointer)
79 , partition(partition)
80 , task_info(task_info)
81 {}
82
83 ActualCellWork(MFWorkerInterface &worker,
84 const unsigned int partition,
85 const TaskInfo & task_info)
86 : worker(&worker)
87 , worker_pointer(nullptr)
88 , partition(partition)
89 , task_info(task_info)
90 {}
91
92 void
93 operator()() const
94 {
95 MFWorkerInterface *used_worker =
96 worker != nullptr ? worker : *worker_pointer;
97 Assert(used_worker != nullptr, ExcInternalError());
98 used_worker->cell(partition);
99
100 if (task_info.face_partition_data.empty() == false)
101 {
102 used_worker->face(partition);
103 used_worker->boundary(partition);
104 }
105 }
106
107 private:
108 MFWorkerInterface * worker;
109 MFWorkerInterface **worker_pointer;
110 const unsigned int partition;
111 const TaskInfo & task_info;
112 };
113
114 class CellWork : public tbb::task
115 {
116 public:
117 CellWork(MFWorkerInterface &worker,
118 const unsigned int partition,
119 const TaskInfo & task_info,
120 const bool is_blocked)
121 : dummy(nullptr)
122 , work(worker, partition, task_info)
123 , is_blocked(is_blocked)
124 {}
125
126 tbb::task *
127 execute() override
128 {
129 work();
130
131 if (is_blocked == true)
132 tbb::empty_task::spawn(*dummy);
133 return nullptr;
134 }
135
136 tbb::empty_task *dummy;
137
138 private:
139 ActualCellWork work;
140 const bool is_blocked;
141 };
142
143
144
145 class PartitionWork : public tbb::task
146 {
147 public:
148 PartitionWork(MFWorkerInterface &function_in,
149 const unsigned int partition_in,
150 const TaskInfo & task_info_in,
151 const bool is_blocked_in = false)
152 : dummy(nullptr)
153 , function(function_in)
154 , partition(partition_in)
155 , task_info(task_info_in)
156 , is_blocked(is_blocked_in)
157 {}
158
159 tbb::task *
160 execute() override
161 {
162 tbb::empty_task *root =
163 new (tbb::task::allocate_root()) tbb::empty_task;
164 const unsigned int evens = task_info.partition_evens[partition];
165 const unsigned int odds = task_info.partition_odds[partition];
166 const unsigned int n_blocked_workers =
167 task_info.partition_n_blocked_workers[partition];
168 const unsigned int n_workers =
169 task_info.partition_n_workers[partition];
170 std::vector<CellWork *> worker(n_workers);
171 std::vector<CellWork *> blocked_worker(n_blocked_workers);
172
173 root->set_ref_count(evens + 1);
174 for (unsigned int j = 0; j < evens; j++)
175 {
176 worker[j] = new (root->allocate_child())
177 CellWork(function,
178 task_info.partition_row_index[partition] + 2 * j,
179 task_info,
180 false);
181 if (j > 0)
182 {
183 worker[j]->set_ref_count(2);
184 blocked_worker[j - 1]->dummy =
185 new (worker[j]->allocate_child()) tbb::empty_task;
186 tbb::task::spawn(*blocked_worker[j - 1]);
187 }
188 else
189 worker[j]->set_ref_count(1);
190 if (j < evens - 1)
191 {
192 blocked_worker[j] = new (worker[j]->allocate_child())
193 CellWork(function,
194 task_info.partition_row_index[partition] + 2 * j +
195 1,
196 task_info,
197 true);
198 }
199 else
200 {
201 if (odds == evens)
202 {
203 worker[evens] = new (worker[j]->allocate_child())
204 CellWork(function,
205 task_info.partition_row_index[partition] +
206 2 * j + 1,
207 task_info,
208 false);
209 tbb::task::spawn(*worker[evens]);
210 }
211 else
212 {
213 tbb::empty_task *child =
214 new (worker[j]->allocate_child()) tbb::empty_task();
215 tbb::task::spawn(*child);
216 }
217 }
218 }
219
220 root->wait_for_all();
221 root->destroy(*root);
222 if (is_blocked == true)
223 tbb::empty_task::spawn(*dummy);
224 return nullptr;
225 }
226
227 tbb::empty_task *dummy;
228
229 private:
230 MFWorkerInterface &function;
231 const unsigned int partition;
232 const TaskInfo & task_info;
233 const bool is_blocked;
234 };
235
236 } // end of namespace partition
237
238
239
240 namespace color
241 {
242 class CellWork
243 {
244 public:
245 CellWork(MFWorkerInterface &worker_in,
246 const TaskInfo & task_info_in,
247 const unsigned int partition_in)
248 : worker(worker_in)
249 , task_info(task_info_in)
250 , partition(partition_in)
251 {}
252
253 void
254 operator()(const tbb::blocked_range<unsigned int> &r) const
255 {
256 const unsigned int start_index =
257 task_info.cell_partition_data[partition] +
258 task_info.block_size * r.begin();
259 const unsigned int end_index =
260 std::min(start_index + task_info.block_size * (r.end() - r.begin()),
261 task_info.cell_partition_data[partition + 1]);
262 worker.cell(std::make_pair(start_index, end_index));
263
264 if (task_info.face_partition_data.empty() == false)
265 {
266 AssertThrow(false, ExcNotImplemented());
267 }
268 }
269
270 private:
271 MFWorkerInterface &worker;
272 const TaskInfo & task_info;
273 const unsigned int partition;
274 };
275
276
277
278 class PartitionWork : public tbb::task
279 {
280 public:
281 PartitionWork(MFWorkerInterface &worker_in,
282 const unsigned int partition_in,
283 const TaskInfo & task_info_in,
284 const bool is_blocked_in)
285 : dummy(nullptr)
286 , worker(worker_in)
287 , partition(partition_in)
288 , task_info(task_info_in)
289 , is_blocked(is_blocked_in)
290 {}
291
292 tbb::task *
293 execute() override
294 {
295 const unsigned int n_chunks =
296 (task_info.cell_partition_data[partition + 1] -
297 task_info.cell_partition_data[partition] + task_info.block_size -
298 1) /
299 task_info.block_size;
300 parallel_for(tbb::blocked_range<unsigned int>(0, n_chunks, 1),
301 CellWork(worker, task_info, partition));
302 if (is_blocked == true)
303 tbb::empty_task::spawn(*dummy);
304 return nullptr;
305 }
306
307 tbb::empty_task *dummy;
308
309 private:
310 MFWorkerInterface &worker;
311 const unsigned int partition;
312 const TaskInfo & task_info;
313 const bool is_blocked;
314 };
315
316 } // end of namespace color
317
318
319
320 class MPICommunication : public tbb::task
321 {
322 public:
323 MPICommunication(MFWorkerInterface &worker_in, const bool do_compress)
324 : worker(worker_in)
325 , do_compress(do_compress)
326 {}
327
328 tbb::task *
329 execute() override
330 {
331 if (do_compress == false)
332 worker.vector_update_ghosts_finish();
333 else
334 worker.vector_compress_start();
335 return nullptr;
336 }
337
338 private:
339 MFWorkerInterface &worker;
340 const bool do_compress;
341 };
342
343#endif // DEAL_II_WITH_TBB
344
345
346
347 void
348 TaskInfo::loop(MFWorkerInterface &funct) const
349 {
350 // If we use thread parallelism, we do not currently support to schedule
351 // pieces of updates within the loop, so this index will collect all
352 // calls in that case and work like a single complete loop over all
353 // cells
354 if (scheme != none)
355 funct.cell_loop_pre_range(numbers::invalid_unsigned_int);
356 else
357 funct.cell_loop_pre_range(
358 partition_row_index[partition_row_index.size() - 2]);
359
360 funct.vector_update_ghosts_start();
361
362#if defined(DEAL_II_WITH_TBB) && !defined(DEAL_II_TBB_WITH_ONEAPI)
363
364 if (scheme != none)
365 {
366 funct.zero_dst_vector_range(numbers::invalid_unsigned_int);
367 if (scheme == partition_partition && evens > 0)
368 {
369 tbb::empty_task *root =
370 new (tbb::task::allocate_root()) tbb::empty_task;
371 root->set_ref_count(evens + 1);
372 std::vector<partition::PartitionWork *> worker(n_workers);
373 std::vector<partition::PartitionWork *> blocked_worker(
374 n_blocked_workers);
375 MPICommunication *worker_compr =
376 new (root->allocate_child()) MPICommunication(funct, true);
377 worker_compr->set_ref_count(1);
378 for (unsigned int j = 0; j < evens; j++)
379 {
380 if (j > 0)
381 {
382 worker[j] = new (root->allocate_child())
383 partition::PartitionWork(funct, 2 * j, *this, false);
384 worker[j]->set_ref_count(2);
385 blocked_worker[j - 1]->dummy =
386 new (worker[j]->allocate_child()) tbb::empty_task;
387 tbb::task::spawn(*blocked_worker[j - 1]);
388 }
389 else
390 {
391 worker[j] = new (worker_compr->allocate_child())
392 partition::PartitionWork(funct, 2 * j, *this, false);
393 worker[j]->set_ref_count(2);
394 MPICommunication *worker_dist =
395 new (worker[j]->allocate_child())
396 MPICommunication(funct, false);
397 tbb::task::spawn(*worker_dist);
398 }
399 if (j < evens - 1)
400 {
401 blocked_worker[j] = new (worker[j]->allocate_child())
402 partition::PartitionWork(funct, 2 * j + 1, *this, true);
403 }
404 else
405 {
406 if (odds == evens)
407 {
408 worker[evens] = new (worker[j]->allocate_child())
409 partition::PartitionWork(funct,
410 2 * j + 1,
411 *this,
412 false);
413 tbb::task::spawn(*worker[evens]);
414 }
415 else
416 {
417 tbb::empty_task *child =
418 new (worker[j]->allocate_child()) tbb::empty_task();
419 tbb::task::spawn(*child);
420 }
421 }
422 }
423
424 root->wait_for_all();
425 root->destroy(*root);
426 }
427 else if (scheme == partition_partition)
428 {
429 // catch the case of empty partition list: we still need to call
430 // the vector communication routines to clean up and initiate
431 // things
432 funct.vector_update_ghosts_finish();
433 funct.vector_compress_start();
434 }
435 else // end of partition-partition, start of partition-color
436 {
437 // check whether there is only one partition. if not, build up the
438 // tree of partitions
439 if (odds > 0)
440 {
441 tbb::empty_task *root =
442 new (tbb::task::allocate_root()) tbb::empty_task;
443 root->set_ref_count(evens + 1);
444 const unsigned int n_blocked_workers =
445 odds - (odds + evens + 1) % 2;
446 const unsigned int n_workers =
447 cell_partition_data.size() - 1 - n_blocked_workers;
448 std::vector<color::PartitionWork *> worker(n_workers);
449 std::vector<color::PartitionWork *> blocked_worker(
450 n_blocked_workers);
451 unsigned int worker_index = 0, slice_index = 0;
452 int spawn_index_child = -2;
453 MPICommunication *worker_compr =
454 new (root->allocate_child()) MPICommunication(funct, true);
455 worker_compr->set_ref_count(1);
456 for (unsigned int part = 0;
457 part < partition_row_index.size() - 1;
458 part++)
459 {
460 if (part == 0)
461 worker[worker_index] =
462 new (worker_compr->allocate_child())
463 color::PartitionWork(funct,
464 slice_index,
465 *this,
466 false);
467 else
468 worker[worker_index] = new (root->allocate_child())
469 color::PartitionWork(funct,
470 slice_index,
471 *this,
472 false);
473 slice_index++;
474 for (; slice_index < partition_row_index[part + 1];
475 slice_index++)
476 {
477 worker[worker_index]->set_ref_count(1);
478 worker_index++;
479 worker[worker_index] =
480 new (worker[worker_index - 1]->allocate_child())
481 color::PartitionWork(funct,
482 slice_index,
483 *this,
484 false);
485 }
486 worker[worker_index]->set_ref_count(2);
487 if (part > 0)
488 {
489 blocked_worker[(part - 1) / 2]->dummy =
490 new (worker[worker_index]->allocate_child())
491 tbb::empty_task;
492 worker_index++;
493 if (spawn_index_child == -1)
494 tbb::task::spawn(*blocked_worker[(part - 1) / 2]);
495 else
496 {
497 Assert(spawn_index_child >= 0,
498 ExcInternalError());
499 tbb::task::spawn(*worker[spawn_index_child]);
500 }
501 spawn_index_child = -2;
502 }
503 else
504 {
505 MPICommunication *worker_dist =
506 new (worker[worker_index]->allocate_child())
507 MPICommunication(funct, false);
508 tbb::task::spawn(*worker_dist);
509 worker_index++;
510 }
511 part += 1;
512 if (part < partition_row_index.size() - 1)
513 {
514 if (part < partition_row_index.size() - 2)
515 {
516 blocked_worker[part / 2] =
517 new (worker[worker_index - 1]->allocate_child())
518 color::PartitionWork(funct,
519 slice_index,
520 *this,
521 true);
522 slice_index++;
523 if (slice_index < partition_row_index[part + 1])
524 {
525 blocked_worker[part / 2]->set_ref_count(1);
526 worker[worker_index] = new (
527 blocked_worker[part / 2]->allocate_child())
528 color::PartitionWork(funct,
529 slice_index,
530 *this,
531 false);
532 slice_index++;
533 }
534 else
535 {
536 spawn_index_child = -1;
537 continue;
538 }
539 }
540 for (; slice_index < partition_row_index[part + 1];
541 slice_index++)
542 {
543 if (slice_index > partition_row_index[part])
544 {
545 worker[worker_index]->set_ref_count(1);
546 worker_index++;
547 }
548 worker[worker_index] =
549 new (worker[worker_index - 1]->allocate_child())
550 color::PartitionWork(funct,
551 slice_index,
552 *this,
553 false);
554 }
555 spawn_index_child = worker_index;
556 worker_index++;
557 }
558 else
559 {
560 tbb::empty_task *final =
561 new (worker[worker_index - 1]->allocate_child())
562 tbb::empty_task;
563 tbb::task::spawn(*final);
564 spawn_index_child = worker_index - 1;
565 }
566 }
567 if (evens == odds)
568 {
569 Assert(spawn_index_child >= 0, ExcInternalError());
570 tbb::task::spawn(*worker[spawn_index_child]);
571 }
572 root->wait_for_all();
573 root->destroy(*root);
574 }
575 // case when we only have one partition: this is the usual
576 // coloring scheme, and we just schedule a parallel for loop for
577 // each color
578 else
579 {
580 Assert(evens <= 1, ExcInternalError());
581 funct.vector_update_ghosts_finish();
582
583 for (unsigned int color = 0; color < partition_row_index[1];
584 ++color)
585 {
586 tbb::empty_task *root =
587 new (tbb::task::allocate_root()) tbb::empty_task;
588 root->set_ref_count(2);
589 color::PartitionWork *worker =
590 new (root->allocate_child())
591 color::PartitionWork(funct, color, *this, false);
592 tbb::empty_task::spawn(*worker);
593 root->wait_for_all();
594 root->destroy(*root);
595 }
596
597 funct.vector_compress_start();
598 }
599 }
600 }
601 else
602#endif
603 // serial loop, go through up to three times and do the MPI transfer at
604 // the beginning/end of the second part
605 {
606 for (unsigned int part = 0; part < partition_row_index.size() - 2;
607 ++part)
608 {
609 if (part == 1)
610 funct.vector_update_ghosts_finish();
611
612 for (unsigned int i = partition_row_index[part];
613 i < partition_row_index[part + 1];
614 ++i)
615 {
616 funct.cell_loop_pre_range(i);
617 funct.zero_dst_vector_range(i);
618 AssertIndexRange(i + 1, cell_partition_data.size());
619 if (cell_partition_data[i + 1] > cell_partition_data[i])
620 {
621 funct.cell(i);
622 }
623
624 if (face_partition_data.empty() == false)
625 {
626 if (face_partition_data[i + 1] > face_partition_data[i])
627 funct.face(i);
628 if (boundary_partition_data[i + 1] >
629 boundary_partition_data[i])
630 funct.boundary(i);
631 }
632 funct.cell_loop_post_range(i);
633 }
634
635 if (part == 1)
636 funct.vector_compress_start();
637 }
638 }
639 funct.vector_compress_finish();
640
641 if (scheme != none)
642 funct.cell_loop_post_range(numbers::invalid_unsigned_int);
643 else
644 funct.cell_loop_post_range(
645 partition_row_index[partition_row_index.size() - 2]);
646 }
647
648
649
650 TaskInfo::TaskInfo()
651 {
652 clear();
653 }
654
655
656
657 void
658 TaskInfo::clear()
659 {
660 n_active_cells = 0;
661 n_ghost_cells = 0;
662 vectorization_length = 1;
663 block_size = 0;
664 n_blocks = 0;
665 scheme = none;
666 partition_row_index.clear();
667 partition_row_index.resize(2);
668 cell_partition_data.clear();
669 face_partition_data.clear();
670 boundary_partition_data.clear();
671 evens = 0;
672 odds = 0;
673 n_blocked_workers = 0;
674 n_workers = 0;
675 partition_evens.clear();
676 partition_odds.clear();
677 partition_n_blocked_workers.clear();
678 partition_n_workers.clear();
679 communicator = MPI_COMM_SELF;
680 my_pid = 0;
681 n_procs = 1;
682 }
683
684
685
686 template <typename StreamType>
687 void
688 TaskInfo::print_memory_statistics(StreamType & out,
689 const std::size_t data_length) const
690 {
691 Utilities::MPI::MinMaxAvg memory_c =
692 Utilities::MPI::min_max_avg(1e-6 * data_length, communicator);
693 if (n_procs < 2)
694 out << memory_c.min;
695 else
696 out << memory_c.min << "/" << memory_c.avg << "/" << memory_c.max;
697 out << " MB" << std::endl;
698 }
699
700
701
702 std::size_t
703 TaskInfo::memory_consumption() const
704 {
705 return (
706 sizeof(*this) +
707 MemoryConsumption::memory_consumption(partition_row_index) +
708 MemoryConsumption::memory_consumption(cell_partition_data) +
709 MemoryConsumption::memory_consumption(face_partition_data) +
710 MemoryConsumption::memory_consumption(boundary_partition_data) +
711 MemoryConsumption::memory_consumption(partition_evens) +
712 MemoryConsumption::memory_consumption(partition_odds) +
713 MemoryConsumption::memory_consumption(partition_n_blocked_workers) +
714 MemoryConsumption::memory_consumption(partition_n_workers));
715 }
716
717
718
719 void
720 TaskInfo::make_boundary_cells_divisible(
721 std::vector<unsigned int> &boundary_cells)
722 {
723 // try to make the number of boundary cells divisible by the number of
724 // vectors in vectorization
725 unsigned int fillup_needed =
726 (vectorization_length - boundary_cells.size() % vectorization_length) %
727 vectorization_length;
728 if (fillup_needed > 0 && boundary_cells.size() < n_active_cells)
729 {
730 // fill additional cells into the list of boundary cells to get a
731 // balanced number. Go through the indices successively until we
732 // found enough indices
733 std::vector<unsigned int> new_boundary_cells;
734 new_boundary_cells.reserve(boundary_cells.size());
735
736 unsigned int next_free_slot = 0, bound_index = 0;
737 while (fillup_needed > 0 && bound_index < boundary_cells.size())
738 {
739 if (next_free_slot < boundary_cells[bound_index])
740 {
741 // check if there are enough cells to fill with in the
742 // current slot
743 if (next_free_slot + fillup_needed <=
744 boundary_cells[bound_index])
745 {
746 for (unsigned int j =
747 boundary_cells[bound_index] - fillup_needed;
748 j < boundary_cells[bound_index];
749 ++j)
750 new_boundary_cells.push_back(j);
751 fillup_needed = 0;
752 }
753 // ok, not enough indices, so just take them all up to the
754 // next boundary cell
755 else
756 {
757 for (unsigned int j = next_free_slot;
758 j < boundary_cells[bound_index];
759 ++j)
760 new_boundary_cells.push_back(j);
761 fillup_needed -=
762 boundary_cells[bound_index] - next_free_slot;
763 }
764 }
765 new_boundary_cells.push_back(boundary_cells[bound_index]);
766 next_free_slot = boundary_cells[bound_index] + 1;
767 ++bound_index;
768 }
769 while (fillup_needed > 0 &&
770 (new_boundary_cells.size() == 0 ||
771 new_boundary_cells.back() < n_active_cells - 1))
772 new_boundary_cells.push_back(new_boundary_cells.back() + 1);
773 while (bound_index < boundary_cells.size())
774 new_boundary_cells.push_back(boundary_cells[bound_index++]);
775
776 boundary_cells.swap(new_boundary_cells);
777 }
778
779 // set the number of cells
780 std::sort(boundary_cells.begin(), boundary_cells.end());
781
782 // check that number of boundary cells is divisible by
783 // vectorization_length or that it contains all cells
784 Assert(boundary_cells.size() % vectorization_length == 0 ||
785 boundary_cells.size() == n_active_cells,
786 ExcInternalError());
787 }
788
789
790
791 void
792 TaskInfo::create_blocks_serial(
793 const std::vector<unsigned int> &cells_with_comm,
794 const unsigned int dofs_per_cell,
795 const bool categories_are_hp,
796 const std::vector<unsigned int> &cell_vectorization_categories,
797 const bool cell_vectorization_categories_strict,
798 const std::vector<unsigned int> &parent_relation,
799 std::vector<unsigned int> & renumbering,
800 std::vector<unsigned char> & incompletely_filled_vectorization)
801 {
802 Assert(dofs_per_cell > 0, ExcInternalError());
803 // This function is decomposed into several steps to determine a good
804 // ordering that satisfies the following constraints:
805 // a. Only cells belonging to the same category (or next higher if the
806 // cell_vectorization_categories_strict is false) can be grouped into
807 // the same SIMD batch
808 // b. hp-adaptive computations must form contiguous ranges for the same
809 // degree (category) in cell_partition_data
810 // c. We want to group the cells with the same parent in the same SIMD
811 // lane if possible
812 // d. The cell order should be similar to the initial one
813 // e. Form sets without MPI communication and those with to overlap
814 // communication with computation
815 //
816 // These constraints are satisfied by first grouping by the categories
817 // and, within the groups, to distinguish between cells with a parent
818 // and those without. All of this is set up with batches of cells (with
819 // padding if the size does not match). Then we define a vector of
820 // arrays where we define sorting criteria for the cell batches to
821 // satisfy the items b and d together, split by different parts to
822 // satisfy item e.
823
824 // Give the compiler a chance to detect that vectorization_length is a
825 // power of two, which allows it to replace integer divisions by shifts
826 unsigned int vectorization_length_bits = 0;
827 unsigned int my_length = vectorization_length;
828 while (my_length >>= 1)
829 ++vectorization_length_bits;
830 const unsigned int n_lanes = 1 << vectorization_length_bits;
831
832 // Step 1: find tight map of categories for not taking exceeding amounts
833 // of memory below. Sort the new categories by the numbers in the
834 // old one to ensure we respect the given rules
835 unsigned int n_categories = 1;
836 std::vector<unsigned int> tight_category_map(n_active_cells, 0);
837 if (cell_vectorization_categories.empty() == false)
838 {
839 AssertDimension(cell_vectorization_categories.size(),
840 n_active_cells + n_ghost_cells);
841
842 std::set<unsigned int> used_categories;
843 for (unsigned int i = 0; i < n_active_cells; ++i)
844 used_categories.insert(cell_vectorization_categories[i]);
845 std::vector<unsigned int> used_categories_vector(
846 used_categories.size());
847 n_categories = 0;
848 for (const auto &it : used_categories)
849 used_categories_vector[n_categories++] = it;
850 for (unsigned int i = 0; i < n_active_cells; ++i)
851 {
852 const unsigned int index =
853 std::lower_bound(used_categories_vector.begin(),
854 used_categories_vector.end(),
855 cell_vectorization_categories[i]) -
856 used_categories_vector.begin();
857 AssertIndexRange(index, used_categories_vector.size());
858 tight_category_map[i] = index;
859 }
860 }
861
862 // Step 2: Sort the cells by the category. If we want to fill up the
863 // ranges in vectorization, promote some of the cells to a higher
864 // category
865 std::vector<std::vector<unsigned int>> renumbering_category(n_categories);
866 for (unsigned int i = 0; i < n_active_cells; ++i)
867 renumbering_category[tight_category_map[i]].push_back(i);
868
869 if (cell_vectorization_categories_strict == false && n_categories > 1)
870 for (unsigned int j = n_categories - 1; j > 0; --j)
871 {
872 unsigned int lower_index = j - 1;
873 while (renumbering_category[j].size() % n_lanes)
874 {
875 while (renumbering_category[j].size() % n_lanes &&
876 !renumbering_category[lower_index].empty())
877 {
878 renumbering_category[j].push_back(
879 renumbering_category[lower_index].back());
880 renumbering_category[lower_index].pop_back();
881 }
882 if (lower_index == 0)
883 break;
884 else
885 --lower_index;
886 }
887 }
888
889 // Step 3: Use the parent relation to find a good grouping of cells. To
890 // do this, we first put cells of each category defined above into two
891 // bins, those which we know can be grouped together by the given parent
892 // relation and those which cannot
893 std::vector<unsigned int> temporary_numbering;
894 temporary_numbering.reserve(n_active_cells +
895 (n_lanes - 1) * n_categories);
896 const unsigned int n_cells_per_parent =
897 std::count(parent_relation.begin(), parent_relation.end(), 0);
898 std::vector<unsigned int> category_size;
899 for (unsigned int j = 0; j < n_categories; ++j)
900 {
901 std::vector<std::pair<unsigned int, unsigned int>> grouped_cells;
902 std::vector<unsigned int> other_cells;
903 for (const unsigned int cell : renumbering_category[j])
904 if (parent_relation.empty() ||
905 parent_relation[cell] == numbers::invalid_unsigned_int)
906 other_cells.push_back(cell);
907 else
908 grouped_cells.emplace_back(parent_relation[cell], cell);
909
910 // Compute the number of cells per group
911 std::sort(grouped_cells.begin(), grouped_cells.end());
912 std::vector<unsigned int> n_cells_per_group;
913 unsigned int length = 0;
914 for (unsigned int i = 0; i < grouped_cells.size(); ++i, ++length)
915 if (i > 0 && grouped_cells[i].first != grouped_cells[i - 1].first)
916 {
917 n_cells_per_group.push_back(length);
918 length = 0;
919 }
920 if (length > 0)
921 n_cells_per_group.push_back(length);
922
923 // Move groups that do not have the complete size (due to
924 // categories) to the 'other_cells'. The cells with correct group
925 // size are immediately appended to the temporary cell numbering
926 auto group_it = grouped_cells.begin();
927 for (unsigned int length : n_cells_per_group)
928 if (length < n_cells_per_parent)
929 for (unsigned int j = 0; j < length; ++j)
930 other_cells.push_back((group_it++)->second);
931 else
932 {
933 // we should not have more cells in a group than in the first
934 // check we did above
935 AssertDimension(length, n_cells_per_parent);
936 for (unsigned int j = 0; j < length; ++j)
937 temporary_numbering.push_back((group_it++)->second);
938 }
939
940 // Sort the remaining cells and append them as well
941 std::sort(other_cells.begin(), other_cells.end());
942 temporary_numbering.insert(temporary_numbering.end(),
943 other_cells.begin(),
944 other_cells.end());
945
946 while (temporary_numbering.size() % n_lanes != 0)
947 temporary_numbering.push_back(numbers::invalid_unsigned_int);
948
949 category_size.push_back(temporary_numbering.size());
950 }
951
952 // Step 4: Identify the batches with cells marked as "comm"
953 std::vector<bool> batch_with_comm(temporary_numbering.size() / n_lanes,
954 false);
955 std::vector<unsigned int> temporary_numbering_inverse(n_active_cells);
956 for (unsigned int i = 0; i < temporary_numbering.size(); ++i)
957 if (temporary_numbering[i] != numbers::invalid_unsigned_int)
958 temporary_numbering_inverse[temporary_numbering[i]] = i;
959 for (const unsigned int cell : cells_with_comm)
960 batch_with_comm[temporary_numbering_inverse[cell] / n_lanes] = true;
961
962 // Step 5: Sort the batches of cells by their last cell index to get
963 // good locality, assuming that the initial cell order is of good
964 // locality. In case we have hp-calculations with categories, we need to
965 // sort also by the category.
966 std::vector<std::array<unsigned int, 3>> batch_order;
967 std::vector<std::array<unsigned int, 3>> batch_order_comm;
968 for (unsigned int i = 0; i < temporary_numbering.size(); i += n_lanes)
969 {
970 unsigned int max_index = 0;
971 for (unsigned int j = 0; j < n_lanes; ++j)
972 if (temporary_numbering[i + j] < numbers::invalid_unsigned_int)
973 max_index = std::max(temporary_numbering[i + j], max_index);
974 const unsigned int category_hp =
975 categories_are_hp ?
976 std::upper_bound(category_size.begin(), category_size.end(), i) -
977 category_size.begin() :
978 0;
979 const std::array<unsigned int, 3> next{{category_hp, max_index, i}};
980 if (batch_with_comm[i / n_lanes])
981 batch_order_comm.emplace_back(next);
982 else
983 batch_order.emplace_back(next);
984 }
985
986 std::sort(batch_order.begin(), batch_order.end());
987 std::sort(batch_order_comm.begin(), batch_order_comm.end());
988
989 // Step 6: Put the cells with communication in the middle of the cell
990 // range. For the MPI case, we need three groups to enable overlap for
991 // communication and computation (part before comm, part with comm, part
992 // after comm), whereas we need one for the other case. And in each
993 // case, we allow for a slot of "ghosted" cells.
994 std::vector<unsigned int> blocks;
995 if (n_procs == 1)
996 {
997 if (batch_order.empty())
998 std::swap(batch_order_comm, batch_order);
999 Assert(batch_order_comm.empty(), ExcInternalError());
1000 partition_row_index.resize(3);
1001 blocks = {0, static_cast<unsigned int>(batch_order.size())};
1002 }
1003 else
1004 {
1005 partition_row_index.resize(5);
1006 const unsigned int comm_begin = batch_order.size() / 2;
1007 batch_order.insert(batch_order.begin() + comm_begin,
1008 batch_order_comm.begin(),
1009 batch_order_comm.end());
1010 const unsigned int comm_end = comm_begin + batch_order_comm.size();
1011 const unsigned int end = batch_order.size();
1012 blocks = {0, comm_begin, comm_end, end};
1013 }
1014
1015 // Step 7: Fill in the data by batches for the locally owned cells.
1016 const unsigned int n_cell_batches = batch_order.size();
1017 const unsigned int n_ghost_batches =
1018 (n_ghost_cells + n_lanes - 1) / n_lanes;
1019 incompletely_filled_vectorization.resize(n_cell_batches +
1020 n_ghost_batches);
1021
1022 cell_partition_data.clear();
1023 cell_partition_data.resize(1, 0);
1024
1025 renumbering.clear();
1026 renumbering.resize(n_active_cells + n_ghost_cells,
1027 numbers::invalid_unsigned_int);
1028
1029 unsigned int counter = 0;
1030 for (unsigned int block = 0; block < blocks.size() - 1; ++block)
1031 {
1032 const unsigned int grain_size =
1033 std::max((2048U / dofs_per_cell) / 8 * 4, 2U);
1034 for (unsigned int k = blocks[block]; k < blocks[block + 1];
1035 k += grain_size)
1036 cell_partition_data.push_back(
1037 std::min(k + grain_size, blocks[block + 1]));
1038 partition_row_index[block + 1] = cell_partition_data.size() - 1;
1039
1040 // Set the numbering according to the reordered temporary one
1041 for (unsigned int k = blocks[block]; k < blocks[block + 1]; ++k)
1042 {
1043 const unsigned int pos = batch_order[k][2];
1044 unsigned int j = 0;
1045 for (; j < n_lanes && temporary_numbering[pos + j] !=
1046 numbers::invalid_unsigned_int;
1047 ++j)
1048 renumbering[counter++] = temporary_numbering[pos + j];
1049 if (j < n_lanes)
1050 incompletely_filled_vectorization[k] = j;
1051 }
1052 }
1053 AssertDimension(counter, n_active_cells);
1054
1055 // Step 8: Treat the ghost cells
1056 for (unsigned int cell = n_active_cells;
1057 cell < n_active_cells + n_ghost_cells;
1058 ++cell)
1059 {
1060 if (!cell_vectorization_categories.empty())
1061 AssertDimension(cell_vectorization_categories[cell],
1062 cell_vectorization_categories[n_active_cells]);
1063 renumbering[cell] = cell;
1064 }
1065 if (n_ghost_cells % n_lanes)
1066 incompletely_filled_vectorization.back() = n_ghost_cells % n_lanes;
1067 cell_partition_data.push_back(n_cell_batches + n_ghost_batches);
1068 partition_row_index.back() = cell_partition_data.size() - 1;
1069
1070#ifdef DEBUG
1071 std::vector<unsigned int> renumber_cpy(renumbering);
1072 std::sort(renumber_cpy.begin(), renumber_cpy.end());
1073 for (unsigned int i = 0; i < renumber_cpy.size(); ++i)
1074 AssertDimension(i, renumber_cpy[i]);
1075#endif
1076 }
1077
1078
1079
1080 void
1081 TaskInfo::initial_setup_blocks_tasks(
1082 const std::vector<unsigned int> &boundary_cells,
1083 std::vector<unsigned int> & renumbering,
1084 std::vector<unsigned char> & incompletely_filled_vectorization)
1085 {
1086 const unsigned int n_cell_batches =
1087 (n_active_cells + vectorization_length - 1) / vectorization_length;
1088 const unsigned int n_ghost_slots =
1089 (n_ghost_cells + vectorization_length - 1) / vectorization_length;
1090 incompletely_filled_vectorization.resize(n_cell_batches + n_ghost_slots);
1091 if (n_cell_batches * vectorization_length > n_active_cells)
1092 incompletely_filled_vectorization[n_cell_batches - 1] =
1093 vectorization_length -
1094 (n_cell_batches * vectorization_length - n_active_cells);
1095 if (n_ghost_slots * vectorization_length > n_ghost_cells)
1096 incompletely_filled_vectorization[n_cell_batches + n_ghost_slots - 1] =
1097 vectorization_length -
1098 (n_ghost_slots * vectorization_length - n_ghost_cells);
1099
1100 std::vector<unsigned int> reverse_numbering(
1101 n_active_cells, numbers::invalid_unsigned_int);
1102 for (unsigned int j = 0; j < boundary_cells.size(); ++j)
1103 reverse_numbering[boundary_cells[j]] = j;
1104 unsigned int counter = boundary_cells.size();
1105 for (unsigned int j = 0; j < n_active_cells; ++j)
1106 if (reverse_numbering[j] == numbers::invalid_unsigned_int)
1107 reverse_numbering[j] = counter++;
1108
1109 AssertDimension(counter, n_active_cells);
1110 renumbering = Utilities::invert_permutation(reverse_numbering);
1111
1112 for (unsigned int j = n_active_cells; j < n_active_cells + n_ghost_cells;
1113 ++j)
1114 renumbering.push_back(j);
1115
1116 // TODO: might be able to simplify this code by not relying on the cell
1117 // partition data while computing the thread graph
1118 cell_partition_data.clear();
1119 cell_partition_data.push_back(0);
1120 if (n_procs > 1)
1121 {
1122 const unsigned int n_macro_boundary_cells =
1123 (boundary_cells.size() + vectorization_length - 1) /
1124 vectorization_length;
1125 cell_partition_data.push_back(
1126 (n_cell_batches - n_macro_boundary_cells) / 2);
1127 cell_partition_data.push_back(cell_partition_data[1] +
1128 n_macro_boundary_cells);
1129 }
1130 else
1131 AssertDimension(boundary_cells.size(), 0);
1132 cell_partition_data.push_back(n_cell_batches);
1133 cell_partition_data.push_back(cell_partition_data.back() + n_ghost_slots);
1134 partition_row_index.resize(n_procs > 1 ? 4 : 2);
1135 partition_row_index[0] = 0;
1136 partition_row_index[1] = 1;
1137 if (n_procs > 1)
1138 {
1139 partition_row_index[2] = 2;
1140 partition_row_index[3] = 3;
1141 }
1142 }
1143
1144
1145
1146 void
1147 TaskInfo::guess_block_size(const unsigned int dofs_per_cell)
1148 {
1149 // user did not say a positive number, so we have to guess
1150 if (block_size == 0)
1151 {
1152 // we would like to have enough work to do, so as first guess, try
1153 // to get 16 times as many chunks as we have threads on the system.
1154 block_size = n_active_cells / (MultithreadInfo::n_threads() * 16 *
1155 vectorization_length);
1156
1157 // if there are too few degrees of freedom per cell, need to
1158 // increase the block size
1159 const unsigned int minimum_parallel_grain_size = 200;
1160 if (dofs_per_cell * block_size < minimum_parallel_grain_size)
1161 block_size = (minimum_parallel_grain_size / dofs_per_cell + 1);
1162 if (dofs_per_cell * block_size > 10000)
1163 block_size /= 4;
1164
1165 block_size =
1166 1 << static_cast<unsigned int>(std::log2(block_size + 1));
1167 }
1168 if (block_size > n_active_cells)
1169 block_size = std::max(1U, n_active_cells);
1170 }
1171
1172
1173
1174 void
1175 TaskInfo::make_thread_graph_partition_color(
1176 DynamicSparsityPattern & connectivity_large,
1177 std::vector<unsigned int> & renumbering,
1178 std::vector<unsigned char> &irregular_cells,
1179 const bool)
1180 {
1181 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
1182 if (n_cell_batches == 0)
1183 return;
1184
1185 Assert(vectorization_length > 0, ExcInternalError());
1186
1187 unsigned int partition = 0, counter = 0;
1188
1189 // Create connectivity graph for blocks based on connectivity graph for
1190 // cells.
1191 DynamicSparsityPattern connectivity(n_blocks, n_blocks);
1192 make_connectivity_cells_to_blocks(irregular_cells,
1193 connectivity_large,
1194 connectivity);
1195
1196 // Create cell-block partitioning.
1197
1198 // For each block of cells, this variable saves to which partitions the
1199 // block belongs. Initialize all to -1 to mark them as not yet assigned
1200 // a partition.
1201 std::vector<unsigned int> cell_partition(n_blocks,
1202 numbers::invalid_unsigned_int);
1203
1204 // In element j of this variable, one puts the old number of the block
1205 // that should be the jth block in the new numeration.
1206 std::vector<unsigned int> partition_list(n_blocks, 0);
1207 std::vector<unsigned int> partition_color_list(n_blocks, 0);
1208
1209 // This vector points to the start of each partition.
1210 std::vector<unsigned int> partition_size(2, 0);
1211
1212 // blocking_connectivity = true;
1213
1214 // The cluster_size in make_partitioning defines that the no. of cells
1215 // in each partition should be a multiple of cluster_size.
1216 unsigned int cluster_size = 1;
1217
1218 // Make the partitioning of the first layer of the blocks of cells.
1219 make_partitioning(connectivity,
1220 cluster_size,
1221 cell_partition,
1222 partition_list,
1223 partition_size,
1224 partition);
1225
1226 // Color the cells within each partition
1227 make_coloring_within_partitions_pre_blocked(connectivity,
1228 partition,
1229 cell_partition,
1230 partition_list,
1231 partition_size,
1232 partition_color_list);
1233
1234 partition_list = renumbering;
1235
1236#ifdef DEBUG
1237 // in debug mode, check that the partition color list is one-to-one
1238 {
1239 std::vector<unsigned int> sorted_pc_list(partition_color_list);
1240 std::sort(sorted_pc_list.begin(), sorted_pc_list.end());
1241 for (unsigned int i = 0; i < sorted_pc_list.size(); ++i)
1242 Assert(sorted_pc_list[i] == i, ExcInternalError());
1243 }
1244#endif
1245
1246 // set the start list for each block and compute the renumbering of
1247 // cells
1248 std::vector<unsigned int> block_start(n_cell_batches + 1);
1249 std::vector<unsigned char> irregular(n_cell_batches);
1250
1251 unsigned int mcell_start = 0;
1252 block_start[0] = 0;
1253 for (unsigned int block = 0; block < n_blocks; block++)
1254 {
1255 block_start[block + 1] = block_start[block];
1256 for (unsigned int mcell = mcell_start;
1257 mcell < std::min(mcell_start + block_size, n_cell_batches);
1258 ++mcell)
1259 {
1260 unsigned int n_comp = (irregular_cells[mcell] > 0) ?
1261 irregular_cells[mcell] :
1262 vectorization_length;
1263 block_start[block + 1] += n_comp;
1264 ++counter;
1265 }
1266 mcell_start += block_size;
1267 }
1268 counter = 0;
1269 unsigned int counter_macro = 0;
1270 unsigned int block_size_last =
1271 n_cell_batches - block_size * (n_blocks - 1);
1272 if (block_size_last == 0)
1273 block_size_last = block_size;
1274
1275 unsigned int tick = 0;
1276 for (unsigned int block = 0; block < n_blocks; block++)
1277 {
1278 unsigned int present_block = partition_color_list[block];
1279 for (unsigned int cell = block_start[present_block];
1280 cell < block_start[present_block + 1];
1281 ++cell)
1282 renumbering[counter++] = partition_list[cell];
1283 unsigned int this_block_size =
1284 (present_block == n_blocks - 1) ? block_size_last : block_size;
1285
1286 // Also re-compute the content of cell_partition_data to
1287 // contain the numbers of cells, not blocks
1288 if (cell_partition_data[tick] == block)
1289 cell_partition_data[tick++] = counter_macro;
1290
1291 for (unsigned int j = 0; j < this_block_size; j++)
1292 irregular[counter_macro++] =
1293 irregular_cells[present_block * block_size + j];
1294 }
1295 AssertDimension(tick + 1, cell_partition_data.size());
1296 cell_partition_data.back() = counter_macro;
1297
1298 irregular_cells.swap(irregular);
1299 AssertDimension(counter, n_active_cells);
1300 AssertDimension(counter_macro, n_cell_batches);
1301
1302 // check that the renumbering is one-to-one
1303#ifdef DEBUG
1304 {
1305 std::vector<unsigned int> sorted_renumbering(renumbering);
1306 std::sort(sorted_renumbering.begin(), sorted_renumbering.end());
1307 for (unsigned int i = 0; i < sorted_renumbering.size(); ++i)
1308 Assert(sorted_renumbering[i] == i, ExcInternalError());
1309 }
1310#endif
1311
1312
1313 update_task_info(
1314 partition); // Actually sets too much for partition color case
1315
1316 AssertDimension(cell_partition_data.back(), n_cell_batches);
1317 }
1318
1319
1320
1321 void
1322 TaskInfo::make_thread_graph(
1323 const std::vector<unsigned int> &cell_active_fe_index,
1324 DynamicSparsityPattern & connectivity,
1325 std::vector<unsigned int> & renumbering,
1326 std::vector<unsigned char> & irregular_cells,
1327 const bool hp_bool)
1328 {
1329 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
1330 if (n_cell_batches == 0)
1331 return;
1332
1333 Assert(vectorization_length > 0, ExcInternalError());
1334
1335 // if we want to block before partitioning, create connectivity graph
1336 // for blocks based on connectivity graph for cells.
1337 DynamicSparsityPattern connectivity_blocks(n_blocks, n_blocks);
1338 make_connectivity_cells_to_blocks(irregular_cells,
1339 connectivity,
1340 connectivity_blocks);
1341
1342 unsigned int n_blocks = 0;
1343 if (scheme == partition_color ||
1344 scheme == color) // blocking_connectivity == true
1345 n_blocks = this->n_blocks;
1346 else
1347 n_blocks = n_active_cells;
1348
1349 // For each block of cells, this variable saves to which partitions the
1350 // block belongs. Initialize all to -1 to mark them as not yet assigned
1351 // a partition.
1352 std::vector<unsigned int> cell_partition(n_blocks,
1353 numbers::invalid_unsigned_int);
1354
1355 // In element j of this variable, one puts the old number (but after
1356 // renumbering according to the input renumbering) of the block that
1357 // should be the jth block in the new numeration.
1358 std::vector<unsigned int> partition_list(n_blocks, 0);
1359 std::vector<unsigned int> partition_2layers_list(n_blocks, 0);
1360
1361 // This vector points to the start of each partition.
1362 std::vector<unsigned int> partition_size(2, 0);
1363
1364 unsigned int partition = 0;
1365
1366 // Within the partitions we want to be able to block for the case that
1367 // we do not block already in the connectivity. The cluster_size in
1368 // make_partitioning defines that the no. of cells in each partition
1369 // should be a multiple of cluster_size.
1370 unsigned int cluster_size = 1;
1371 if (scheme == partition_partition)
1372 cluster_size = block_size * vectorization_length;
1373
1374 // Make the partitioning of the first layer of the blocks of cells.
1375 if (scheme == partition_color || scheme == color)
1376 make_partitioning(connectivity_blocks,
1377 cluster_size,
1378 cell_partition,
1379 partition_list,
1380 partition_size,
1381 partition);
1382 else
1383 make_partitioning(connectivity,
1384 cluster_size,
1385 cell_partition,
1386 partition_list,
1387 partition_size,
1388 partition);
1389
1390 // Partition or color second layer
1391 if (scheme == partition_partition)
1392
1393 {
1394 // Partition within partitions.
1395 make_partitioning_within_partitions_post_blocked(
1396 connectivity,
1397 cell_active_fe_index,
1398 partition,
1399 cluster_size,
1400 hp_bool,
1401 cell_partition,
1402 partition_list,
1403 partition_size,
1404 partition_2layers_list,
1405 irregular_cells);
1406 }
1407 else if (scheme == partition_color || scheme == color)
1408 {
1409 make_coloring_within_partitions_pre_blocked(connectivity_blocks,
1410 partition,
1411 cell_partition,
1412 partition_list,
1413 partition_size,
1414 partition_2layers_list);
1415 }
1416
1417 // in debug mode, check that the partition_2layers_list is one-to-one
1418#ifdef DEBUG
1419 {
1420 std::vector<unsigned int> sorted_pc_list(partition_2layers_list);
1421 std::sort(sorted_pc_list.begin(), sorted_pc_list.end());
1422 for (unsigned int i = 0; i < sorted_pc_list.size(); ++i)
1423 Assert(sorted_pc_list[i] == i, ExcInternalError());
1424 }
1425#endif
1426
1427 // Set the new renumbering
1428 std::vector<unsigned int> renumbering_in(n_active_cells, 0);
1429 renumbering_in.swap(renumbering);
1430 if (scheme == partition_partition) // blocking_connectivity == false
1431 {
1432 // This is the simple case. The renumbering is just a combination of
1433 // the renumbering that we were given as an input and the
1434 // renumbering of partition/coloring given in partition_2layers_list
1435 for (unsigned int j = 0; j < renumbering.size(); j++)
1436 renumbering[j] = renumbering_in[partition_2layers_list[j]];
1437 // Account for the ghost cells, finally.
1438 for (unsigned int i = 0; i < n_ghost_cells; ++i)
1439 renumbering.push_back(i + n_active_cells);
1440 }
1441 else
1442 {
1443 // set the start list for each block and compute the renumbering of
1444 // cells
1445 std::vector<unsigned int> block_start(n_cell_batches + 1);
1446 std::vector<unsigned char> irregular(n_cell_batches);
1447
1448 unsigned int counter = 0;
1449 unsigned int mcell_start = 0;
1450 block_start[0] = 0;
1451 for (unsigned int block = 0; block < n_blocks; block++)
1452 {
1453 block_start[block + 1] = block_start[block];
1454 for (unsigned int mcell = mcell_start;
1455 mcell < std::min(mcell_start + block_size, n_cell_batches);
1456 ++mcell)
1457 {
1458 unsigned int n_comp = (irregular_cells[mcell] > 0) ?
1459 irregular_cells[mcell] :
1460 vectorization_length;
1461 block_start[block + 1] += n_comp;
1462 ++counter;
1463 }
1464 mcell_start += block_size;
1465 }
1466 counter = 0;
1467 unsigned int counter_macro = 0;
1468 unsigned int block_size_last =
1469 n_cell_batches - block_size * (n_blocks - 1);
1470 if (block_size_last == 0)
1471 block_size_last = block_size;
1472
1473 unsigned int tick = 0;
1474 for (unsigned int block = 0; block < n_blocks; block++)
1475 {
1476 unsigned int present_block = partition_2layers_list[block];
1477 for (unsigned int cell = block_start[present_block];
1478 cell < block_start[present_block + 1];
1479 ++cell)
1480 renumbering[counter++] = renumbering_in[cell];
1481 unsigned int this_block_size =
1482 (present_block == n_blocks - 1) ? block_size_last : block_size;
1483
1484 // Also re-compute the content of cell_partition_data to
1485 // contain the numbers of cells, not blocks
1486 if (cell_partition_data[tick] == block)
1487 cell_partition_data[tick++] = counter_macro;
1488
1489 for (unsigned int j = 0; j < this_block_size; j++)
1490 irregular[counter_macro++] =
1491 irregular_cells[present_block * block_size + j];
1492 }
1493 AssertDimension(tick + 1, cell_partition_data.size());
1494 cell_partition_data.back() = counter_macro;
1495
1496 irregular_cells.swap(irregular);
1497 AssertDimension(counter, n_active_cells);
1498 AssertDimension(counter_macro, n_cell_batches);
1499 // check that the renumbering is one-to-one
1500#ifdef DEBUG
1501 {
1502 std::vector<unsigned int> sorted_renumbering(renumbering);
1503 std::sort(sorted_renumbering.begin(), sorted_renumbering.end());
1504 for (unsigned int i = 0; i < sorted_renumbering.size(); ++i)
1505 Assert(sorted_renumbering[i] == i, ExcInternalError());
1506 }
1507#endif
1508 }
1509
1510 // Update the task_info with the more information for the thread graph.
1511 update_task_info(partition);
1512 }
1513
1514
1515
1516 void
1517 TaskInfo::make_thread_graph_partition_partition(
1518 const std::vector<unsigned int> &cell_active_fe_index,
1519 DynamicSparsityPattern & connectivity,
1520 std::vector<unsigned int> & renumbering,
1521 std::vector<unsigned char> & irregular_cells,
1522 const bool hp_bool)
1523 {
1524 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
1525 if (n_cell_batches == 0)
1526 return;
1527
1528 const unsigned int cluster_size = block_size * vectorization_length;
1529
1530 // Create cell-block partitioning.
1531
1532 // For each block of cells, this variable saves to which partitions the
1533 // block belongs. Initialize all to n_cell_batches to mark them as not
1534 // yet assigned a partition.
1535 std::vector<unsigned int> cell_partition(n_active_cells,
1536 numbers::invalid_unsigned_int);
1537
1538
1539 // In element j of this variable, one puts the old number of the block
1540 // that should be the jth block in the new numeration.
1541 std::vector<unsigned int> partition_list(n_active_cells, 0);
1542 std::vector<unsigned int> partition_partition_list(n_active_cells, 0);
1543
1544 // This vector points to the start of each partition.
1545 std::vector<unsigned int> partition_size(2, 0);
1546
1547 unsigned int partition = 0;
1548 // Here, we do not block inside the connectivity graph
1549 // blocking_connectivity = false;
1550
1551 // Make the partitioning of the first layer of the blocks of cells.
1552 make_partitioning(connectivity,
1553 cluster_size,
1554 cell_partition,
1555 partition_list,
1556 partition_size,
1557 partition);
1558
1559 // Partition within partitions.
1560 make_partitioning_within_partitions_post_blocked(connectivity,
1561 cell_active_fe_index,
1562 partition,
1563 cluster_size,
1564 hp_bool,
1565 cell_partition,
1566 partition_list,
1567 partition_size,
1568 partition_partition_list,
1569 irregular_cells);
1570
1571 partition_list.swap(renumbering);
1572
1573 for (unsigned int j = 0; j < renumbering.size(); j++)
1574 renumbering[j] = partition_list[partition_partition_list[j]];
1575
1576 for (unsigned int i = 0; i < n_ghost_cells; ++i)
1577 renumbering.push_back(i + n_active_cells);
1578
1579 update_task_info(partition);
1580 }
1581
1582
1583
1584 void
1585 TaskInfo::make_connectivity_cells_to_blocks(
1586 const std::vector<unsigned char> &irregular_cells,
1587 const DynamicSparsityPattern & connectivity_cells,
1588 DynamicSparsityPattern & connectivity_blocks) const
1589 {
1590 std::vector<std::vector<unsigned int>> cell_blocks(n_blocks);
1591 std::vector<unsigned int> touched_cells(n_active_cells);
1592 unsigned int cell = 0;
1593 for (unsigned int i = 0, mcell = 0; i < n_blocks; ++i)
1594 {
1595 for (unsigned int c = 0;
1596 c < block_size && mcell < *(cell_partition_data.end() - 2);
1597 ++c, ++mcell)
1598 {
1599 unsigned int ncomp = (irregular_cells[mcell] > 0) ?
1600 irregular_cells[mcell] :
1601 vectorization_length;
1602 for (unsigned int c = 0; c < ncomp; ++c, ++cell)
1603 {
1604 cell_blocks[i].push_back(cell);
1605 touched_cells[cell] = i;
1606 }
1607 }
1608 }
1609 AssertDimension(cell, n_active_cells);
1610 for (unsigned int i = 0; i < cell_blocks.size(); ++i)
1611 for (unsigned int col = 0; col < cell_blocks[i].size(); ++col)
1612 {
1613 for (DynamicSparsityPattern::iterator it =
1614 connectivity_cells.begin(cell_blocks[i][col]);
1615 it != connectivity_cells.end(cell_blocks[i][col]);
1616 ++it)
1617 {
1618 if (touched_cells[it->column()] != i)
1619 connectivity_blocks.add(i, touched_cells[it->column()]);
1620 }
1621 }
1622 }
1623
1624
1625
1626 // Function to create partitioning on the second layer within each
1627 // partition. Version without preblocking.
1628 void
1629 TaskInfo::make_partitioning_within_partitions_post_blocked(
1630 const DynamicSparsityPattern & connectivity,
1631 const std::vector<unsigned int> &cell_active_fe_index,
1632 const unsigned int partition,
1633 const unsigned int cluster_size,
1634 const bool hp_bool,
1635 const std::vector<unsigned int> &cell_partition,
1636 const std::vector<unsigned int> &partition_list,
1637 const std::vector<unsigned int> &partition_size,
1638 std::vector<unsigned int> & partition_partition_list,
1639 std::vector<unsigned char> & irregular_cells)
1640 {
1641 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
1642 const unsigned int n_ghost_slots =
1643 *(cell_partition_data.end() - 1) - n_cell_batches;
1644
1645 // List of cells in previous partition
1646 std::vector<unsigned int> neighbor_list;
1647 // List of cells in current partition for use as neighbors in next
1648 // partition
1649 std::vector<unsigned int> neighbor_neighbor_list;
1650
1651 std::vector<unsigned int> renumbering(n_active_cells);
1652
1653 irregular_cells.back() = 0;
1654 irregular_cells.resize(n_active_cells + n_ghost_slots);
1655
1656 unsigned int max_fe_index = 0;
1657 for (const unsigned int fe_index : cell_active_fe_index)
1658 max_fe_index = std::max(fe_index, max_fe_index);
1659
1660 Assert(!hp_bool || cell_active_fe_index.size() == n_active_cells,
1661 ExcInternalError());
1662
1663 {
1664 unsigned int n_cell_batches_before = 0;
1665 // Create partitioning within partitions.
1666
1667 // For each block of cells, this variable saves to which partitions
1668 // the block belongs. Initialize all to n_cell_batches to mark them as
1669 // not yet assigned a partition.
1670 std::vector<unsigned int> cell_partition_l2(
1671 n_active_cells, numbers::invalid_unsigned_int);
1672 partition_row_index.clear();
1673 partition_row_index.resize(partition + 1, 0);
1674 cell_partition_data.resize(1, 0);
1675
1676 unsigned int counter = 0;
1677 unsigned int missing_macros;
1678 for (unsigned int part = 0; part < partition; ++part)
1679 {
1680 neighbor_neighbor_list.resize(0);
1681 neighbor_list.resize(0);
1682 bool work = true;
1683 unsigned int partition_l2 = 0;
1684 unsigned int start_up = partition_size[part];
1685 unsigned int partition_counter = 0;
1686 while (work)
1687 {
1688 if (neighbor_list.size() == 0)
1689 {
1690 work = false;
1691 partition_counter = 0;
1692 for (unsigned int j = start_up;
1693 j < partition_size[part + 1];
1694 ++j)
1695 if (cell_partition[partition_list[j]] == part &&
1696 cell_partition_l2[partition_list[j]] ==
1697 numbers::invalid_unsigned_int)
1698 {
1699 start_up = j;
1700 work = true;
1701 partition_counter = 1;
1702 // To start up, set the start_up cell to partition
1703 // and list all its neighbors.
1704 AssertIndexRange(start_up, partition_size[part + 1]);
1705 cell_partition_l2[partition_list[start_up]] =
1706 partition_l2;
1707 neighbor_neighbor_list.push_back(
1708 partition_list[start_up]);
1709 partition_partition_list[counter++] =
1710 partition_list[start_up];
1711 start_up++;
1712 break;
1713 }
1714 }
1715 else
1716 {
1717 partition_counter = 0;
1718 for (const unsigned int neighbor : neighbor_list)
1719 {
1720 Assert(cell_partition[neighbor] == part,
1721 ExcInternalError());
1722 Assert(cell_partition_l2[neighbor] == partition_l2 - 1,
1723 ExcInternalError());
1724 auto neighbor_it = connectivity.begin(neighbor);
1725 const auto end_it = connectivity.end(neighbor);
1726 for (; neighbor_it != end_it; ++neighbor_it)
1727 {
1728 if (cell_partition[neighbor_it->column()] == part &&
1729 cell_partition_l2[neighbor_it->column()] ==
1730 numbers::invalid_unsigned_int)
1731 {
1732 cell_partition_l2[neighbor_it->column()] =
1733 partition_l2;
1734 neighbor_neighbor_list.push_back(
1735 neighbor_it->column());
1736 partition_partition_list[counter++] =
1737 neighbor_it->column();
1738 partition_counter++;
1739 }
1740 }
1741 }
1742 }
1743 if (partition_counter > 0)
1744 {
1745 int index_before = neighbor_neighbor_list.size(),
1746 index = index_before;
1747 {
1748 // put the cells into separate lists for each FE index
1749 // within one partition-partition
1750 missing_macros = 0;
1751 std::vector<unsigned int> remaining_per_cell_batch(
1752 max_fe_index + 1);
1753 std::vector<std::vector<unsigned int>>
1754 renumbering_fe_index;
1755 unsigned int cell;
1756 bool filled = true;
1757 if (hp_bool == true)
1758 {
1759 renumbering_fe_index.resize(max_fe_index + 1);
1760 for (cell = counter - partition_counter;
1761 cell < counter;
1762 ++cell)
1763 {
1764 renumbering_fe_index
1765 [cell_active_fe_index.empty() ?
1766 0 :
1767 cell_active_fe_index
1768 [partition_partition_list[cell]]]
1769 .push_back(partition_partition_list[cell]);
1770 }
1771 // check how many more cells are needed in the lists
1772 for (unsigned int j = 0; j < max_fe_index + 1; j++)
1773 {
1774 remaining_per_cell_batch[j] =
1775 renumbering_fe_index[j].size() %
1776 vectorization_length;
1777 if (remaining_per_cell_batch[j] != 0)
1778 filled = false;
1779 missing_macros +=
1780 ((renumbering_fe_index[j].size() +
1781 vectorization_length - 1) /
1782 vectorization_length);
1783 }
1784 }
1785 else
1786 {
1787 remaining_per_cell_batch.resize(1);
1788 remaining_per_cell_batch[0] =
1789 partition_counter % vectorization_length;
1790 missing_macros =
1791 partition_counter / vectorization_length;
1792 if (remaining_per_cell_batch[0] != 0)
1793 {
1794 filled = false;
1795 missing_macros++;
1796 }
1797 }
1798 missing_macros =
1799 cluster_size - (missing_macros % cluster_size);
1800
1801 // now we realized that there are some cells missing.
1802 while (missing_macros > 0 || filled == false)
1803 {
1804 if (index == 0)
1805 {
1806 index = neighbor_neighbor_list.size();
1807 if (index == index_before)
1808 {
1809 if (missing_macros != 0)
1810 {
1811 neighbor_neighbor_list.resize(0);
1812 }
1813 start_up--;
1814 break; // not connected - start again
1815 }
1816 index_before = index;
1817 }
1818 index--;
1819 unsigned int additional =
1820 neighbor_neighbor_list[index];
1821
1822 // go through the neighbors of the last cell in the
1823 // current partition and check if we find some to
1824 // fill up with.
1825 DynamicSparsityPattern::iterator neighbor =
1826 connectivity.begin(
1827 additional),
1828 end =
1829 connectivity.end(
1830 additional);
1831 for (; neighbor != end; ++neighbor)
1832 {
1833 if (cell_partition[neighbor->column()] == part &&
1834 cell_partition_l2[neighbor->column()] ==
1835 numbers::invalid_unsigned_int)
1836 {
1837 unsigned int this_index = 0;
1838 if (hp_bool == true)
1839 this_index =
1840 cell_active_fe_index.empty() ?
1841 0 :
1842 cell_active_fe_index[neighbor
1843 ->column()];
1844
1845 // Only add this cell if we need more macro
1846 // cells in the current block or if there is
1847 // a macro cell with the FE index that is
1848 // not yet fully populated
1849 if (missing_macros > 0 ||
1850 remaining_per_cell_batch[this_index] > 0)
1851 {
1852 cell_partition_l2[neighbor->column()] =
1853 partition_l2;
1854 neighbor_neighbor_list.push_back(
1855 neighbor->column());
1856 if (hp_bool == true)
1857 renumbering_fe_index[this_index]
1858 .push_back(neighbor->column());
1859 partition_partition_list[counter] =
1860 neighbor->column();
1861 counter++;
1862 partition_counter++;
1863 if (remaining_per_cell_batch
1864 [this_index] == 0 &&
1865 missing_macros > 0)
1866 missing_macros--;
1867 remaining_per_cell_batch[this_index]++;
1868 if (remaining_per_cell_batch
1869 [this_index] ==
1870 vectorization_length)
1871 {
1872 remaining_per_cell_batch[this_index] =
1873 0;
1874 }
1875 if (missing_macros == 0)
1876 {
1877 filled = true;
1878 for (unsigned int fe_ind = 0;
1879 fe_ind < max_fe_index + 1;
1880 ++fe_ind)
1881 if (remaining_per_cell_batch
1882 [fe_ind] != 0)
1883 filled = false;
1884 }
1885 if (filled == true)
1886 break;
1887 }
1888 }
1889 }
1890 }
1891 if (hp_bool == true)
1892 {
1893 // set the renumbering according to their active FE
1894 // index within one partition-partition which was
1895 // implicitly assumed above
1896 cell = counter - partition_counter;
1897 for (unsigned int j = 0; j < max_fe_index + 1; j++)
1898 {
1899 for (const unsigned int jj :
1900 renumbering_fe_index[j])
1901 renumbering[cell++] = jj;
1902 if (renumbering_fe_index[j].size() %
1903 vectorization_length !=
1904 0)
1905 irregular_cells[renumbering_fe_index[j].size() /
1906 vectorization_length +
1907 n_cell_batches_before] =
1908 renumbering_fe_index[j].size() %
1909 vectorization_length;
1910 n_cell_batches_before +=
1911 (renumbering_fe_index[j].size() +
1912 vectorization_length - 1) /
1913 vectorization_length;
1914 renumbering_fe_index[j].resize(0);
1915 }
1916 }
1917 else
1918 {
1919 n_cell_batches_before +=
1920 partition_counter / vectorization_length;
1921 if (partition_counter % vectorization_length != 0)
1922 {
1923 irregular_cells[n_cell_batches_before] =
1924 partition_counter % vectorization_length;
1925 n_cell_batches_before++;
1926 }
1927 }
1928 }
1929 cell_partition_data.push_back(n_cell_batches_before);
1930 partition_l2++;
1931 }
1932 neighbor_list = neighbor_neighbor_list;
1933 neighbor_neighbor_list.resize(0);
1934 }
1935 partition_row_index[part + 1] =
1936 partition_row_index[part] + partition_l2;
1937 }
1938 }
1939 if (hp_bool == true)
1940 {
1941 partition_partition_list.swap(renumbering);
1942 }
1943 }
1944
1945
1946
1947 // Function to create coloring on the second layer within each partition.
1948 // Version assumes preblocking.
1949 void
1950 TaskInfo::make_coloring_within_partitions_pre_blocked(
1951 const DynamicSparsityPattern & connectivity,
1952 const unsigned int partition,
1953 const std::vector<unsigned int> &cell_partition,
1954 const std::vector<unsigned int> &partition_list,
1955 const std::vector<unsigned int> &partition_size,
1956 std::vector<unsigned int> & partition_color_list)
1957 {
1958 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
1959 std::vector<unsigned int> cell_color(n_blocks, n_cell_batches);
1960 std::vector<bool> color_finder;
1961
1962 partition_row_index.resize(partition + 1);
1963 cell_partition_data.clear();
1964 unsigned int color_counter = 0, index_counter = 0;
1965 for (unsigned int part = 0; part < partition; part++)
1966 {
1967 partition_row_index[part] = index_counter;
1968 unsigned int max_color = 0;
1969 for (unsigned int k = partition_size[part];
1970 k < partition_size[part + 1];
1971 k++)
1972 {
1973 unsigned int cell = partition_list[k];
1974 unsigned int n_neighbors = connectivity.row_length(cell);
1975
1976 // In the worst case, each neighbor has a different color. So we
1977 // find at least one available color between 0 and n_neighbors.
1978 color_finder.resize(n_neighbors + 1);
1979 for (unsigned int j = 0; j <= n_neighbors; ++j)
1980 color_finder[j] = true;
1981 DynamicSparsityPattern::iterator neighbor =
1982 connectivity.begin(cell),
1983 end = connectivity.end(cell);
1984 for (; neighbor != end; ++neighbor)
1985 {
1986 // Mark the color that a neighbor within the partition has
1987 // as taken
1988 if (cell_partition[neighbor->column()] == part &&
1989 cell_color[neighbor->column()] <= n_neighbors)
1990 color_finder[cell_color[neighbor->column()]] = false;
1991 }
1992 // Choose the smallest color that is not taken for the block
1993 cell_color[cell] = 0;
1994 while (color_finder[cell_color[cell]] == false)
1995 cell_color[cell]++;
1996 if (cell_color[cell] > max_color)
1997 max_color = cell_color[cell];
1998 }
1999 // Reorder within partition: First, all blocks that belong the 0 and
2000 // then so on until those with color max (Note that the smaller the
2001 // number the larger the partition)
2002 for (unsigned int color = 0; color <= max_color; color++)
2003 {
2004 cell_partition_data.push_back(color_counter);
2005 index_counter++;
2006 for (unsigned int k = partition_size[part];
2007 k < partition_size[part + 1];
2008 k++)
2009 {
2010 unsigned int cell = partition_list[k];
2011 if (cell_color[cell] == color)
2012 {
2013 partition_color_list[color_counter++] = cell;
2014 }
2015 }
2016 }
2017 }
2018 cell_partition_data.push_back(n_blocks);
2019 partition_row_index[partition] = index_counter;
2020 AssertDimension(color_counter, n_blocks);
2021 }
2022
2023
2024 // Function to create partitioning on the first layer.
2025 void
2026 TaskInfo::make_partitioning(const DynamicSparsityPattern &connectivity,
2027 const unsigned int cluster_size,
2028 std::vector<unsigned int> & cell_partition,
2029 std::vector<unsigned int> & partition_list,
2030 std::vector<unsigned int> & partition_size,
2031 unsigned int & partition) const
2032
2033 {
2034 // For each block of cells, this variable saves to which partitions the
2035 // block belongs. Initialize all to n_cell_batches to mark them as not
2036 // yet assigned a partition.
2037 // std::vector<unsigned int> cell_partition (n_active_cells,
2038 // numbers::invalid_unsigned_int);
2039 // List of cells in previous partition
2040 std::vector<unsigned int> neighbor_list;
2041 // List of cells in current partition for use as neighbors in next
2042 // partition
2043 std::vector<unsigned int> neighbor_neighbor_list;
2044
2045 // In element j of this variable, one puts the old number of the block
2046 // that should be the jth block in the new numeration.
2047 // std::vector<unsigned int> partition_list(n_active_cells,0);
2048
2049 // This vector points to the start of each partition.
2050 // std::vector<unsigned int> partition_size(2,0);
2051
2052 partition = 0;
2053 unsigned int counter = 0;
2054 unsigned int start_nonboundary =
2055 cell_partition_data.size() == 5 ?
2056 vectorization_length *
2057 (cell_partition_data[2] - cell_partition_data[1]) :
2058 0;
2059
2060 const unsigned int n_cell_batches = *(cell_partition_data.end() - 2);
2061 if (n_cell_batches == 0)
2062 return;
2063 if (scheme == color)
2064 start_nonboundary = n_cell_batches;
2065 if (scheme == partition_color ||
2066 scheme == color) // blocking_connectivity == true
2067 start_nonboundary = ((start_nonboundary + block_size - 1) / block_size);
2068 unsigned int n_blocks;
2069 if (scheme == partition_color ||
2070 scheme == color) // blocking_connectivity == true
2071 n_blocks = this->n_blocks;
2072 else
2073 n_blocks = n_active_cells;
2074
2075 if (start_nonboundary > n_blocks)
2076 start_nonboundary = n_blocks;
2077
2078
2079 unsigned int start_up = 0;
2080 bool work = true;
2081 unsigned int remainder = cluster_size;
2082
2083 // this performs a classical breath-first search in the connectivity
2084 // graph of the cells under the restriction that the size of the
2085 // partitions should be a multiple of the given block size
2086 while (work)
2087 {
2088 // put the cells with neighbors on remote MPI processes up front
2089 if (start_nonboundary > 0)
2090 {
2091 for (unsigned int cell = 0; cell < start_nonboundary; ++cell)
2092 {
2093 const unsigned int cell_nn = cell;
2094 cell_partition[cell_nn] = partition;
2095 neighbor_list.push_back(cell_nn);
2096 partition_list[counter++] = cell_nn;
2097 partition_size.back()++;
2098 }
2099 start_nonboundary = 0;
2100 remainder -= (start_nonboundary % cluster_size);
2101 if (remainder == cluster_size)
2102 remainder = 0;
2103 }
2104 else
2105 {
2106 // To start up, set the start_up cell to partition and list all
2107 // its neighbors.
2108 cell_partition[start_up] = partition;
2109 neighbor_list.push_back(start_up);
2110 partition_list[counter++] = start_up;
2111 partition_size.back()++;
2112 start_up++;
2113 remainder--;
2114 if (remainder == cluster_size)
2115 remainder = 0;
2116 }
2117 int index_before = neighbor_list.size(), index = index_before,
2118 index_stop = 0;
2119 while (remainder > 0)
2120 {
2121 if (index == index_stop)
2122 {
2123 index = neighbor_list.size();
2124 if (index == index_before)
2125 {
2126 neighbor_list.resize(0);
2127 goto not_connect;
2128 }
2129 index_stop = index_before;
2130 index_before = index;
2131 }
2132 index--;
2133 unsigned int additional = neighbor_list[index];
2134 DynamicSparsityPattern::iterator neighbor =
2135 connectivity.begin(additional),
2136 end =
2137 connectivity.end(additional);
2138 for (; neighbor != end; ++neighbor)
2139 {
2140 if (cell_partition[neighbor->column()] ==
2141 numbers::invalid_unsigned_int)
2142 {
2143 partition_size.back()++;
2144 cell_partition[neighbor->column()] = partition;
2145 neighbor_list.push_back(neighbor->column());
2146 partition_list[counter++] = neighbor->column();
2147 remainder--;
2148 if (remainder == 0)
2149 break;
2150 }
2151 }
2152 }
2153
2154 while (neighbor_list.size() > 0)
2155 {
2156 partition++;
2157
2158 // counter for number of cells so far in current partition
2159 unsigned int partition_counter = 0;
2160
2161 // Mark the start of the new partition
2162 partition_size.push_back(partition_size.back());
2163
2164 // Loop through the list of cells in previous partition and put
2165 // all their neighbors in current partition
2166 for (const unsigned int cell : neighbor_list)
2167 {
2168 Assert(cell_partition[cell] == partition - 1,
2169 ExcInternalError());
2170 auto neighbor = connectivity.begin(cell);
2171 const auto end = connectivity.end(cell);
2172 for (; neighbor != end; ++neighbor)
2173 {
2174 if (cell_partition[neighbor->column()] ==
2175 numbers::invalid_unsigned_int)
2176 {
2177 partition_size.back()++;
2178 cell_partition[neighbor->column()] = partition;
2179
2180 // collect the cells of the current partition for
2181 // use as neighbors in next partition
2182 neighbor_neighbor_list.push_back(neighbor->column());
2183 partition_list[counter++] = neighbor->column();
2184 partition_counter++;
2185 }
2186 }
2187 }
2188 remainder = cluster_size - (partition_counter % cluster_size);
2189 if (remainder == cluster_size)
2190 remainder = 0;
2191 int index_stop = 0;
2192 int index_before = neighbor_neighbor_list.size(),
2193 index = index_before;
2194 while (remainder > 0)
2195 {
2196 if (index == index_stop)
2197 {
2198 index = neighbor_neighbor_list.size();
2199 if (index == index_before)
2200 {
2201 neighbor_neighbor_list.resize(0);
2202 break;
2203 }
2204 index_stop = index_before;
2205 index_before = index;
2206 }
2207 index--;
2208 unsigned int additional = neighbor_neighbor_list[index];
2209 DynamicSparsityPattern::iterator neighbor =
2210 connectivity.begin(
2211 additional),
2212 end = connectivity.end(
2213 additional);
2214 for (; neighbor != end; ++neighbor)
2215 {
2216 if (cell_partition[neighbor->column()] ==
2217 numbers::invalid_unsigned_int)
2218 {
2219 partition_size.back()++;
2220 cell_partition[neighbor->column()] = partition;
2221 neighbor_neighbor_list.push_back(neighbor->column());
2222 partition_list[counter++] = neighbor->column();
2223 remainder--;
2224 if (remainder == 0)
2225 break;
2226 }
2227 }
2228 }
2229
2230 neighbor_list = neighbor_neighbor_list;
2231 neighbor_neighbor_list.resize(0);
2232 }
2233 not_connect:
2234 // One has to check if the graph is not connected so we have to find
2235 // another partition.
2236 work = false;
2237 for (unsigned int j = start_up; j < n_blocks; ++j)
2238 if (cell_partition[j] == numbers::invalid_unsigned_int)
2239 {
2240 start_up = j;
2241 work = true;
2242 if (remainder == 0)
2243 remainder = cluster_size;
2244 break;
2245 }
2246 }
2247 if (remainder != 0)
2248 partition++;
2249
2250 AssertDimension(partition_size[partition], n_blocks);
2251 }
2252
2253
2254 void
2255 TaskInfo::update_task_info(const unsigned int partition)
2256 {
2257 evens = (partition + 1) / 2;
2258 odds = partition / 2;
2259 n_blocked_workers = odds - (odds + evens + 1) % 2;
2260 n_workers = evens + odds - n_blocked_workers;
2261 // From here only used for partition partition option.
2262 partition_evens.resize(partition);
2263 partition_odds.resize(partition);
2264 partition_n_blocked_workers.resize(partition);
2265 partition_n_workers.resize(partition);
2266 for (unsigned int part = 0; part < partition; part++)
2267 {
2268 partition_evens[part] =
2269 (partition_row_index[part + 1] - partition_row_index[part] + 1) / 2;
2270 partition_odds[part] =
2271 (partition_row_index[part + 1] - partition_row_index[part]) / 2;
2272 partition_n_blocked_workers[part] =
2273 partition_odds[part] -
2274 (partition_odds[part] + partition_evens[part] + 1) % 2;
2275 partition_n_workers[part] = partition_evens[part] +
2276 partition_odds[part] -
2277 partition_n_blocked_workers[part];
2278 }
2279 }
2280 } // namespace MatrixFreeFunctions
2281} // namespace internal
2282
2283
2284
2285// explicit instantiations of template functions
2286template void
2287internal::MatrixFreeFunctions::TaskInfo::print_memory_statistics<std::ostream>(
2288 std::ostream &,
2289 const std::size_t) const;
2290template void
2291internal::MatrixFreeFunctions::TaskInfo::print_memory_statistics<
2292 ConditionalOStream>(ConditionalOStream &, const std::size_t) const;
2293
2294
2295DEAL_II_NAMESPACE_CLOSE
MultithreadInfo::n_threads
static unsigned int n_threads()
Definition: multithread_info.cc:109
conditional_ostream.h
DEAL_II_NAMESPACE_OPEN
#define DEAL_II_NAMESPACE_OPEN
Definition: config.h:402
DEAL_II_NAMESPACE_CLOSE
#define DEAL_II_NAMESPACE_CLOSE
Definition: config.h:403
first
Point< 2 > first
Definition: grid_out.cc:4587
StandardExceptions::ExcNotImplemented
static ::ExceptionBase & ExcNotImplemented()
Assert
#define Assert(cond, exc)
Definition: exceptions.h:1465
AssertDimension
#define AssertDimension(dim1, dim2)
Definition: exceptions.h:1622
AssertIndexRange
#define AssertIndexRange(index, range)
Definition: exceptions.h:1690
StandardExceptions::ExcInternalError
static ::ExceptionBase & ExcInternalError()
AssertThrow
#define AssertThrow(cond, exc)
Definition: exceptions.h:1575
DynamicSparsityPatternIterators::Accessor::column
size_type column() const
Definition: dynamic_sparsity_pattern.h:807
DynamicSparsityPattern::begin
iterator begin() const
Definition: dynamic_sparsity_pattern.h:1100
DynamicSparsityPattern::row_length
size_type row_length(const size_type row) const
Definition: dynamic_sparsity_pattern.h:1067
DynamicSparsityPattern::end
iterator end() const
Definition: dynamic_sparsity_pattern.h:1110
DynamicSparsityPattern::add
void add(const size_type i, const size_type j)
Definition: dynamic_sparsity_pattern.h:1027
memory_consumption.h
multithread_info.h
MemoryConsumption::memory_consumption
std::enable_if< std::is_fundamental< T >::value, std::size_t >::type memory_consumption(const T &t)
Definition: memory_consumption.h:266
MemorySpace::swap
void swap(MemorySpaceData< Number, MemorySpace > &, MemorySpaceData< Number, MemorySpace > &)
Definition: memory_space_data.h:92
Physics::Elasticity::Kinematics::e
SymmetricTensor< 2, dim, Number > e(const Tensor< 2, dim, Number > &F)
SparsityTools::partition
void partition(const SparsityPattern &sparsity_pattern, const unsigned int n_partitions, std::vector< unsigned int > &partition_indices, const Partitioner partitioner=Partitioner::metis)
Definition: sparsity_tools.cc:399
TrilinosWrappers::internal::end
VectorType::value_type * end(VectorType &V)
Definition: trilinos_sparse_matrix.cc:67
Utilities::MPI::min_max_avg
MinMaxAvg min_max_avg(const double my_value, const MPI_Comm &mpi_communicator)
Definition: mpi.cc:91
Utilities::lower_bound
Iterator lower_bound(Iterator first, Iterator last, const T &val)
Definition: utilities.h:1111
Utilities::invert_permutation
std::vector< Integer > invert_permutation(const std::vector< Integer > &permutation)
Definition: utilities.h:1482
internal::VectorImplementation::minimum_parallel_grain_size
unsigned int minimum_parallel_grain_size
Definition: parallel.cc:34
numbers::invalid_unsigned_int
static const unsigned int invalid_unsigned_int
Definition: types.h:196
parallel::internal::parallel_for
void parallel_for(Iterator x_begin, Iterator x_end, const Functor &functor, const unsigned int grainsize)
Definition: parallel.h:158
std::min
::VectorizedArray< Number, width > min(const ::VectorizedArray< Number, width > &, const ::VectorizedArray< Number, width > &)
Definition: vectorization.h:5501
std::max
::VectorizedArray< Number, width > max(const ::VectorizedArray< Number, width > &, const ::VectorizedArray< Number, width > &)
Definition: vectorization.h:5484
internal::MFWorkerInterface::face
virtual void face(const unsigned int range_index)=0
internal::MFWorkerInterface::zero_dst_vector_range
virtual void zero_dst_vector_range(const unsigned int range_index)=0
internal::MFWorkerInterface::cell
virtual void cell(const std::pair< unsigned int, unsigned int > &cell_range)=0
internal::MFWorkerInterface::boundary
virtual void boundary(const unsigned int range_index)=0
internal::MFWorkerInterface::vector_compress_start
virtual void vector_compress_start()=0
Starts the communication for the vector compress operation.
internal::MFWorkerInterface::cell_loop_post_range
virtual void cell_loop_post_range(const unsigned int range_index)=0
internal::MFWorkerInterface::vector_update_ghosts_start
virtual void vector_update_ghosts_start()=0
Starts the communication for the update ghost values operation.
internal::MFWorkerInterface::cell_loop_pre_range
virtual void cell_loop_pre_range(const unsigned int range_index)=0
internal::MFWorkerInterface::vector_update_ghosts_finish
virtual void vector_update_ghosts_finish()=0
Finishes the communication for the update ghost values operation.
internal::MFWorkerInterface::vector_compress_finish
virtual void vector_compress_finish()=0
Finishes the communication for the vector compress operation.
internal::MatrixFreeFunctions::TaskInfo::memory_consumption
std::size_t memory_consumption() const
Definition: task_info.cc:703
internal::MatrixFreeFunctions::TaskInfo::boundary_partition_data
std::vector< unsigned int > boundary_partition_data
Definition: task_info.h:518
internal::MatrixFreeFunctions::TaskInfo::loop
void loop(MFWorkerInterface &worker) const
Definition: task_info.cc:348
internal::MatrixFreeFunctions::TaskInfo::make_thread_graph_partition_color
void make_thread_graph_partition_color(DynamicSparsityPattern &connectivity, std::vector< unsigned int > &renumbering, std::vector< unsigned char > &irregular_cells, const bool hp_bool)
Definition: task_info.cc:1175
internal::MatrixFreeFunctions::TaskInfo::partition_n_workers
std::vector< unsigned int > partition_n_workers
Definition: task_info.h:576
internal::MatrixFreeFunctions::TaskInfo::create_blocks_serial
void create_blocks_serial(const std::vector< unsigned int > &cells_with_comm, const unsigned int dofs_per_cell, const bool categories_are_hp, const std::vector< unsigned int > &cell_vectorization_categories, const bool cell_vectorization_categories_strict, const std::vector< unsigned int > &parent_relation, std::vector< unsigned int > &renumbering, std::vector< unsigned char > &incompletely_filled_vectorization)
Definition: task_info.cc:792
internal::MatrixFreeFunctions::TaskInfo::make_connectivity_cells_to_blocks
void make_connectivity_cells_to_blocks(const std::vector< unsigned char > &irregular_cells, const DynamicSparsityPattern &connectivity_cells, DynamicSparsityPattern &connectivity_blocks) const
Definition: task_info.cc:1585
internal::MatrixFreeFunctions::TaskInfo::make_partitioning_within_partitions_post_blocked
void make_partitioning_within_partitions_post_blocked(const DynamicSparsityPattern &connectivity, const std::vector< unsigned int > &cell_active_fe_index, const unsigned int partition, const unsigned int cluster_size, const bool hp_bool, const std::vector< unsigned int > &cell_partition, const std::vector< unsigned int > &partition_list, const std::vector< unsigned int > &partition_size, std::vector< unsigned int > &partition_partition_list, std::vector< unsigned char > &irregular_cells)
Definition: task_info.cc:1629
internal::MatrixFreeFunctions::TaskInfo::initial_setup_blocks_tasks
void initial_setup_blocks_tasks(const std::vector< unsigned int > &boundary_cells, std::vector< unsigned int > &renumbering, std::vector< unsigned char > &incompletely_filled_vectorization)
Definition: task_info.cc:1081
internal::MatrixFreeFunctions::TaskInfo::print_memory_statistics
void print_memory_statistics(StreamType &out, std::size_t data_length) const
Definition: task_info.cc:688
internal::MatrixFreeFunctions::TaskInfo::make_coloring_within_partitions_pre_blocked
void make_coloring_within_partitions_pre_blocked(const DynamicSparsityPattern &connectivity, const unsigned int partition, const std::vector< unsigned int > &cell_partition, const std::vector< unsigned int > &partition_list, const std::vector< unsigned int > &partition_size, std::vector< unsigned int > &partition_color_list)
Definition: task_info.cc:1950
internal::MatrixFreeFunctions::TaskInfo::partition_row_index
std::vector< unsigned int > partition_row_index
Definition: task_info.h:466
internal::MatrixFreeFunctions::TaskInfo::make_thread_graph_partition_partition
void make_thread_graph_partition_partition(const std::vector< unsigned int > &cell_active_fe_index, DynamicSparsityPattern &connectivity, std::vector< unsigned int > &renumbering, std::vector< unsigned char > &irregular_cells, const bool hp_bool)
Definition: task_info.cc:1517
internal::MatrixFreeFunctions::TaskInfo::partition_evens
std::vector< unsigned int > partition_evens
Definition: task_info.h:558
internal::MatrixFreeFunctions::TaskInfo::make_thread_graph
void make_thread_graph(const std::vector< unsigned int > &cell_active_fe_index, DynamicSparsityPattern &connectivity, std::vector< unsigned int > &renumbering, std::vector< unsigned char > &irregular_cells, const bool hp_bool)
Definition: task_info.cc:1322
internal::MatrixFreeFunctions::TaskInfo::partition_n_blocked_workers
std::vector< unsigned int > partition_n_blocked_workers
Definition: task_info.h:570
internal::MatrixFreeFunctions::TaskInfo::cell_partition_data
std::vector< unsigned int > cell_partition_data
Definition: task_info.h:474
internal::MatrixFreeFunctions::TaskInfo::make_partitioning
void make_partitioning(const DynamicSparsityPattern &connectivity, const unsigned int cluster_size, std::vector< unsigned int > &cell_partition, std::vector< unsigned int > &partition_list, std::vector< unsigned int > &partition_size, unsigned int &partition) const
Definition: task_info.cc:2026
internal::MatrixFreeFunctions::TaskInfo::update_task_info
void update_task_info(const unsigned int partition)
Definition: task_info.cc:2255
internal::MatrixFreeFunctions::TaskInfo::make_boundary_cells_divisible
void make_boundary_cells_divisible(std::vector< unsigned int > &boundary_cells)
Definition: task_info.cc:720
internal::MatrixFreeFunctions::TaskInfo::guess_block_size
void guess_block_size(const unsigned int dofs_per_cell)
Definition: task_info.cc:1147
internal::MatrixFreeFunctions::TaskInfo::partition_odds
std::vector< unsigned int > partition_odds
Definition: task_info.h:564
internal::MatrixFreeFunctions::TaskInfo::face_partition_data
std::vector< unsigned int > face_partition_data
Definition: task_info.h:496
task_info.h
Generated by doxygen 1.9.4