/usr/include/trilinos/KokkosKernels_GaussSeidel_impl.hpp is in libtrilinos-kokkos-kernels-dev 12.12.1-5.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 | /*
//@HEADER
// ************************************************************************
//
// KokkosKernels 0.9: Linear Algebra and Graph Kernels
// Copyright 2017 Sandia Corporation
//
// Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
// the U.S. Government retains certain rights in this software.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// 1. Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// 3. Neither the name of the Corporation nor the names of the
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Questions? Contact Siva Rajamanickam (srajama@sandia.gov)
//
// ************************************************************************
//@HEADER
*/
#include "KokkosKernels_GraphColor.hpp"
#include "KokkosKernels_Utils.hpp"
#include <Kokkos_Core.hpp>
#include <Kokkos_Atomic.hpp>
#include <impl/Kokkos_Timer.hpp>
#include <Kokkos_Sort.hpp>
#include <Kokkos_MemoryTraits.hpp>
#ifndef _KOKKOSGSIMP_HPP
#define _KOKKOSGSIMP_HPP
namespace KokkosKernels{
namespace Experimental{
namespace Graph{
namespace Impl{
template <typename HandleType, typename lno_row_view_t_, typename lno_nnz_view_t_, typename scalar_nnz_view_t_>
class GaussSeidel{
public:
typedef lno_row_view_t_ in_lno_row_view_t;
typedef lno_nnz_view_t_ in_lno_nnz_view_t;
typedef scalar_nnz_view_t_ in_scalar_nnz_view_t;
typedef typename HandleType::HandleExecSpace MyExecSpace;
typedef typename HandleType::HandleTempMemorySpace MyTempMemorySpace;
typedef typename HandleType::HandlePersistentMemorySpace MyPersistentMemorySpace;
typedef typename in_lno_row_view_t::non_const_value_type row_lno_t;
typedef typename HandleType::size_type size_type;
typedef typename HandleType::nnz_lno_t nnz_lno_t;
typedef typename HandleType::nnz_scalar_t nnz_scalar_t;
typedef typename HandleType::const_lno_row_view_t const_lno_row_view_t;
typedef typename HandleType::non_const_lno_row_view_t non_const_lno_row_view_t;
typedef typename HandleType::const_lno_nnz_view_t const_lno_nnz_view_t;
typedef typename HandleType::non_const_lno_nnz_view_t non_const_lno_nnz_view_t;
typedef typename HandleType::const_scalar_nnz_view_t const_scalar_nnz_view_t;
typedef typename HandleType::non_const_scalar_nnz_view_t non_const_scalar_nnz_view_t;
typedef typename HandleType::row_lno_temp_work_view_t row_lno_temp_work_view_t;
typedef typename HandleType::row_lno_persistent_work_view_t row_lno_persistent_work_view_t;
typedef typename HandleType::row_lno_persistent_work_host_view_t row_lno_persistent_work_host_view_t; //Host view type
typedef typename HandleType::nnz_lno_temp_work_view_t nnz_lno_temp_work_view_t;
typedef typename HandleType::nnz_lno_persistent_work_view_t nnz_lno_persistent_work_view_t;
typedef typename HandleType::nnz_lno_persistent_work_host_view_t nnz_lno_persistent_work_host_view_t; //Host view type
typedef typename HandleType::scalar_temp_work_view_t scalar_temp_work_view_t;
typedef typename HandleType::scalar_persistent_work_view_t scalar_persistent_work_view_t;
typedef Kokkos::RangePolicy<MyExecSpace> my_exec_space;
typedef nnz_lno_t color_t;
typedef Kokkos::View<color_t *, MyTempMemorySpace> color_view_t;
typedef Kokkos::TeamPolicy<MyExecSpace> team_policy_t ;
typedef typename team_policy_t::member_type team_member_t ;
private:
HandleType *handle;
nnz_lno_t num_rows, num_cols;
const_lno_row_view_t row_map;
const_lno_nnz_view_t entries;
const_scalar_nnz_view_t values;
bool is_symmetric;
public:
struct PSGS{
row_lno_persistent_work_view_t _xadj;
nnz_lno_persistent_work_view_t _adj; // CSR storage of the graph.
scalar_persistent_work_view_t _adj_vals; // CSR storage of the graph.
scalar_persistent_work_view_t _Xvector /*output*/;
scalar_persistent_work_view_t _Yvector;
scalar_persistent_work_view_t _permuted_diagonals;
PSGS(row_lno_persistent_work_view_t xadj_, nnz_lno_persistent_work_view_t adj_, scalar_persistent_work_view_t adj_vals_,
scalar_persistent_work_view_t Xvector_, scalar_persistent_work_view_t Yvector_, nnz_lno_persistent_work_view_t color_adj_,
scalar_persistent_work_view_t permuted_diagonals_):
_xadj( xadj_),
_adj( adj_),
_adj_vals( adj_vals_),
_Xvector( Xvector_),
_Yvector( Yvector_), _permuted_diagonals(permuted_diagonals_){}
KOKKOS_INLINE_FUNCTION
void operator()(const nnz_lno_t &ii) const {
size_type row_begin = _xadj[ii];
size_type row_end = _xadj[ii + 1];
nnz_scalar_t sum = _Yvector[ii];
for (size_type adjind = row_begin; adjind < row_end; ++adjind){
nnz_lno_t colIndex = _adj[adjind];
nnz_scalar_t val = _adj_vals[adjind];
sum -= val * _Xvector[colIndex];
}
nnz_scalar_t diagonalVal = _permuted_diagonals[ii];
_Xvector[ii] = (sum + diagonalVal * _Xvector[ii])/ diagonalVal;
}
};
struct Team_PSGS{
row_lno_persistent_work_view_t _xadj;
nnz_lno_persistent_work_view_t _adj; // CSR storage of the graph.
scalar_persistent_work_view_t _adj_vals; // CSR storage of the graph.
scalar_persistent_work_view_t _Xvector /*output*/;
scalar_persistent_work_view_t _Yvector;
nnz_lno_t _color_set_begin;
nnz_lno_t _color_set_end;
scalar_persistent_work_view_t _permuted_diagonals;
Team_PSGS(row_lno_persistent_work_view_t xadj_, nnz_lno_persistent_work_view_t adj_, scalar_persistent_work_view_t adj_vals_,
scalar_persistent_work_view_t Xvector_, scalar_persistent_work_view_t Yvector_,
nnz_lno_t color_set_begin, nnz_lno_t color_set_end,
scalar_persistent_work_view_t permuted_diagonals_):
_xadj( xadj_),
_adj( adj_),
_adj_vals( adj_vals_),
_Xvector( Xvector_),
_Yvector( Yvector_),
_color_set_begin(color_set_begin),
_color_set_end(color_set_end), _permuted_diagonals(permuted_diagonals_){}
KOKKOS_INLINE_FUNCTION
void operator()(const team_member_t & teamMember) const {
//idx ii = _color_adj[i];
//int ii = teamMember.league_rank() + _shift_index;
nnz_lno_t ii = teamMember.league_rank() * teamMember.team_size()+ teamMember.team_rank() + _color_set_begin;
//check ii is out of range. if it is, just return.
if (ii >= _color_set_end)
return;
size_type row_begin = _xadj[ii];
size_type row_end = _xadj[ii + 1];
//bool am_i_the_diagonal = false;
//nnz_scalar_t diagonal = 1;
nnz_scalar_t product = 0 ;
Kokkos::parallel_reduce(
Kokkos::ThreadVectorRange(teamMember, row_end - row_begin),
//Kokkos::TeamThreadRange(teamMember, row_end - row_begin),
[&] (size_type i, nnz_scalar_t & valueToUpdate) {
size_type adjind = i + row_begin;
nnz_lno_t colIndex = _adj[adjind];
nnz_scalar_t val = _adj_vals[adjind];
valueToUpdate += val * _Xvector[colIndex];
},
product);
Kokkos::single(Kokkos::PerThread(teamMember),[=] () {
nnz_scalar_t diagonalVal = _permuted_diagonals[ii];
_Xvector[ii] = (_Yvector[ii] - product + diagonalVal * _Xvector[ii])/ diagonalVal;
});
}
};
/**
* \brief constructor
*/
GaussSeidel(HandleType *handle_,
nnz_lno_t num_rows_,
nnz_lno_t num_cols_,
const_lno_row_view_t row_map_,
const_lno_nnz_view_t entries_,
const_scalar_nnz_view_t values_):
handle(handle_), num_rows(num_rows_), num_cols(num_cols_),
row_map(row_map_), entries(entries_), values(values_), is_symmetric(true){}
GaussSeidel(HandleType *handle_,
nnz_lno_t num_rows_,
nnz_lno_t num_cols_,
const_lno_row_view_t row_map_,
const_lno_nnz_view_t entries_,
bool is_symmetric_ = true):
handle(handle_),
num_rows(num_rows_), num_cols(num_cols_),
row_map(row_map_),
entries(entries_),
values(), is_symmetric(is_symmetric_){}
/**
* \brief constructor
*/
GaussSeidel(HandleType *handle_,
nnz_lno_t num_rows_,
nnz_lno_t num_cols_,
const_lno_row_view_t row_map_,
const_lno_nnz_view_t entries_,
const_scalar_nnz_view_t values_,
bool is_symmetric_):
handle(handle_),
num_rows(num_rows_), num_cols(num_cols_),
row_map(row_map_), entries(entries_), values(values_), is_symmetric(is_symmetric_){}
void initialize_symbolic(){
//std::cout << std::endl<< std::endl<< std::endl<< std::endl<< std::endl<< std::endl;
typename HandleType::GraphColoringHandleType *gchandle = this->handle->get_graph_coloring_handle();
if (gchandle == NULL){
this->handle->create_graph_coloring_handle();
//this->handle->create_gs_handle();
this->handle->get_gs_handle()->set_owner_of_coloring();
gchandle = this->handle->get_graph_coloring_handle();
}
const_lno_row_view_t xadj = this->row_map;
const_lno_nnz_view_t adj = this->entries;
size_type nnz = adj.dimension_0();
#ifdef KOKKOSKERNELS_TIME_REVERSE
Kokkos::Impl::Timer timer;
#endif
{
if (!is_symmetric){
if (gchandle->get_coloring_algo_type() == KokkosKernels::Experimental::Graph::COLORING_EB){
gchandle->symmetrize_and_calculate_lower_diagonal_edge_list(num_rows, xadj, adj);
graph_color_symbolic <HandleType, const_lno_row_view_t, const_lno_nnz_view_t>
(this->handle, num_rows, num_rows, xadj , adj);
}
else {
row_lno_temp_work_view_t tmp_xadj;
nnz_lno_temp_work_view_t tmp_adj;
KokkosKernels::Experimental::Util::symmetrize_graph_symbolic_hashmap
< const_lno_row_view_t, const_lno_nnz_view_t,
row_lno_temp_work_view_t, nnz_lno_temp_work_view_t,
MyExecSpace>
(num_rows, xadj, adj, tmp_xadj, tmp_adj );
graph_color_symbolic <HandleType, row_lno_temp_work_view_t, nnz_lno_temp_work_view_t> (this->handle, num_rows, num_rows, tmp_xadj , tmp_adj);
}
}
else {
graph_color_symbolic <HandleType, const_lno_row_view_t, const_lno_nnz_view_t> (this->handle, num_rows, num_rows, xadj , adj);
}
}
color_t numColors = gchandle->get_num_colors();
//std::cout << "numCol:" << numColors << " numRows:" << num_rows << " cols:" << num_cols << " nnz:" << adj.dimension_0() << std::endl;
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "COLORING_TIME:" << timer.seconds() << std::endl;
#endif
typename HandleType::GraphColoringHandleType::color_view_t colors = gchandle->get_vertex_colors();
nnz_lno_persistent_work_view_t color_xadj;
nnz_lno_persistent_work_view_t color_adj;
#ifdef KOKKOSKERNELS_TIME_REVERSE
timer.reset();
#endif
KokkosKernels::Experimental::Util::create_reverse_map
<typename HandleType::GraphColoringHandleType::color_view_t,
nnz_lno_persistent_work_view_t, MyExecSpace>
(num_rows, numColors, colors, color_xadj, color_adj);
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "CREATE_REVERSE_MAP:" << timer.seconds() << std::endl;
timer.reset();
#endif
nnz_lno_persistent_work_host_view_t h_color_xadj = Kokkos::create_mirror_view (color_xadj);
Kokkos::deep_copy (h_color_xadj , color_xadj);
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "DEEP_COPY:" << timer.seconds() << std::endl;
timer.reset();
#endif
#if defined( KOKKOS_HAVE_CUDA )
if (Kokkos::Impl::is_same<Kokkos::Cuda, MyExecSpace >::value){
for (nnz_lno_t i = 0; i < numColors; ++i){
nnz_lno_t color_index_begin = h_color_xadj(i);
nnz_lno_t color_index_end = h_color_xadj(i + 1);
if (color_index_begin + 1 >= color_index_end ) continue;
auto colorsubset =
subview(color_adj, Kokkos::pair<row_lno_t, row_lno_t> (color_index_begin, color_index_end));
Kokkos::sort (colorsubset);
}
}
#endif
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "SORT_TIME:" << timer.seconds() << std::endl;
timer.reset();
//std::cout << "sort" << std::endl;
#endif
row_lno_persistent_work_view_t permuted_xadj ("new xadj", num_rows + 1);
nnz_lno_persistent_work_view_t old_to_new_map ("old_to_new_index_", num_rows );
nnz_lno_persistent_work_view_t permuted_adj ("newadj_", nnz );
Kokkos::parallel_for( my_exec_space(0,num_rows),
create_permuted_xadj(
color_adj,
xadj,
permuted_xadj,
old_to_new_map));
//std::cout << "create_permuted_xadj" << std::endl;
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "CREATE_PERMUTED_XADJ:" << timer.seconds() << std::endl;
timer.reset();
#endif
KokkosKernels::Experimental::Util::inclusive_parallel_prefix_sum
<row_lno_persistent_work_view_t, MyExecSpace>
(num_rows + 1, permuted_xadj);
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "INCLUSIVE_PPS:" << timer.seconds() << std::endl;
timer.reset();
#endif
Kokkos::parallel_for( my_exec_space(0,num_rows),
fill_matrix_symbolic(
num_rows,
color_adj,
xadj,
adj,
//adj_vals,
permuted_xadj,
permuted_adj,
//newvals_,
old_to_new_map));
MyExecSpace::fence();
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "SYMBOLIC_FILL:" << timer.seconds() << std::endl;
timer.reset();
#endif
typename HandleType::GaussSeidelHandleType *gsHandler = this->handle->get_gs_handle();
gsHandler->set_color_set_xadj(h_color_xadj);
gsHandler->set_color_set_adj(color_adj);
gsHandler->set_num_colors(numColors);
gsHandler->set_new_xadj(permuted_xadj);
gsHandler->set_new_adj(permuted_adj);
//gsHandler->set_new_adj_val(newvals_);
gsHandler->set_old_to_new_map(old_to_new_map);
if (this->handle->get_gs_handle()->is_owner_of_coloring()){
this->handle->destroy_graph_coloring_handle();
this->handle->get_gs_handle()->set_owner_of_coloring(false);
}
this->handle->get_gs_handle()->set_call_symbolic(true);
this->handle->get_gs_handle()->allocate_x_y_vectors(this->num_rows, this->num_cols);
//std::cout << "all end" << std::endl;
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "ALLOC:" << timer.seconds() << std::endl;
#endif
}
struct create_permuted_xadj{
nnz_lno_persistent_work_view_t color_adj;
const_lno_row_view_t oldxadj;
row_lno_persistent_work_view_t newxadj;
nnz_lno_persistent_work_view_t old_to_new_index;
create_permuted_xadj(
nnz_lno_persistent_work_view_t color_adj_,
const_lno_row_view_t oldxadj_,
row_lno_persistent_work_view_t newxadj_,
nnz_lno_persistent_work_view_t old_to_new_index_):
color_adj(color_adj_), oldxadj(oldxadj_),
newxadj(newxadj_),old_to_new_index(old_to_new_index_){}
KOKKOS_INLINE_FUNCTION
void operator()(const nnz_lno_t &i) const{
nnz_lno_t index = color_adj(i);
newxadj(i + 1) = oldxadj[index + 1] - oldxadj[index];
old_to_new_index[index] = i;
}
};
struct fill_matrix_symbolic{
nnz_lno_t num_rows;
nnz_lno_persistent_work_view_t color_adj;
const_lno_row_view_t oldxadj;
const_lno_nnz_view_t oldadj;
//value_array_type oldadjvals;
row_lno_persistent_work_view_t newxadj;
nnz_lno_persistent_work_view_t newadj;
//value_persistent_work_array_type newadjvals;
nnz_lno_persistent_work_view_t old_to_new_index;
fill_matrix_symbolic(
nnz_lno_t num_rows_,
nnz_lno_persistent_work_view_t color_adj_,
const_lno_row_view_t oldxadj_,
const_lno_nnz_view_t oldadj_,
//value_array_type oldadjvals_,
row_lno_persistent_work_view_t newxadj_,
nnz_lno_persistent_work_view_t newadj_,
//value_persistent_work_array_type newadjvals_,
nnz_lno_persistent_work_view_t old_to_new_index_):
num_rows(num_rows_),
color_adj(color_adj_), oldxadj(oldxadj_), oldadj(oldadj_), //oldadjvals(oldadjvals_),
newxadj(newxadj_), newadj(newadj_), //newadjvals(newadjvals_),
old_to_new_index(old_to_new_index_){}
KOKKOS_INLINE_FUNCTION
void operator()(const nnz_lno_t &i) const{
nnz_lno_t index = color_adj(i);
size_type xadj_begin = newxadj(i);
size_type old_xadj_end = oldxadj[index + 1];
for (size_type j = oldxadj[index]; j < old_xadj_end; ++j){
nnz_lno_t neighbor = oldadj[j];
if(neighbor < num_rows) neighbor = old_to_new_index[neighbor];
newadj[xadj_begin++] = neighbor;
//newadjvals[xadj_begin++] = oldadjvals[j];
}
}
};
struct fill_matrix_numeric{
nnz_lno_persistent_work_view_t color_adj;
const_lno_row_view_t oldxadj;
const_scalar_nnz_view_t oldadjvals;
row_lno_persistent_work_view_t newxadj;
scalar_persistent_work_view_t newadjvals;
fill_matrix_numeric(
nnz_lno_persistent_work_view_t color_adj_,
const_lno_row_view_t oldxadj_,
const_scalar_nnz_view_t oldadjvals_,
row_lno_persistent_work_view_t newxadj_,
scalar_persistent_work_view_t newadjvals_):
color_adj(color_adj_), oldxadj(oldxadj_), oldadjvals(oldadjvals_),
newxadj(newxadj_), newadjvals(newadjvals_){}
KOKKOS_INLINE_FUNCTION
void operator()(const nnz_lno_t &i) const{
nnz_lno_t index = color_adj(i);
size_type xadj_begin = newxadj(i);
size_type old_xadj_end = oldxadj[index + 1];
for (size_type j = oldxadj[index]; j < old_xadj_end; ++j){
newadjvals[xadj_begin++] = oldadjvals[j];
}
}
};
struct Get_Matrix_Diagonals{
row_lno_persistent_work_view_t _xadj;
nnz_lno_persistent_work_view_t _adj; // CSR storage of the graph.
scalar_persistent_work_view_t _adj_vals; // CSR storage of the graph.
scalar_persistent_work_view_t _diagonals;
size_type nr;
Get_Matrix_Diagonals(
row_lno_persistent_work_view_t xadj_,
nnz_lno_persistent_work_view_t adj_,
scalar_persistent_work_view_t adj_vals_,
scalar_persistent_work_view_t diagonals_):
_xadj( xadj_),
_adj( adj_),
_adj_vals( adj_vals_), _diagonals(diagonals_),
nr(xadj_.dimension_0() - 1){}
KOKKOS_INLINE_FUNCTION
void operator()(const nnz_lno_t & ii) const {
size_type row_begin = _xadj[ii];
size_type row_end = _xadj[ii + 1];
for (size_type c = row_begin; c < row_end; ++c){
nnz_lno_t colIndex = _adj[c];
if (colIndex == ii){
nnz_scalar_t val = _adj_vals[c];
_diagonals[ii] = val;
}
}
}
};
void initialize_numeric(){
if (this->handle->get_gs_handle()->is_symbolic_called() == false){
this->initialize_symbolic();
}
//else
#ifdef KOKKOSKERNELS_TIME_REVERSE
Kokkos::Impl::Timer timer;
#endif
{
const_lno_row_view_t xadj = this->row_map;
const_lno_nnz_view_t adj = this->entries;
size_type nnz = adj.dimension_0();
const_scalar_nnz_view_t adj_vals = this->values;
typename HandleType::GaussSeidelHandleType *gsHandler = this->handle->get_gs_handle();
row_lno_persistent_work_view_t newxadj_ = gsHandler->get_new_xadj();
nnz_lno_persistent_work_view_t old_to_new_map = gsHandler->get_old_to_new_map();
nnz_lno_persistent_work_view_t newadj_ = gsHandler->get_new_adj();
nnz_lno_persistent_work_view_t color_adj = gsHandler->get_color_adj();
scalar_persistent_work_view_t permuted_adj_vals (Kokkos::ViewAllocateWithoutInitializing("newvals_"), nnz );
Kokkos::parallel_for( my_exec_space(0,num_rows),
fill_matrix_numeric(
color_adj,
xadj,
//adj,
adj_vals,
newxadj_,
//newadj_,
permuted_adj_vals
//,old_to_new_map
));
MyExecSpace::fence();
gsHandler->set_new_adj_val(permuted_adj_vals);
scalar_persistent_work_view_t permuted_diagonals (Kokkos::ViewAllocateWithoutInitializing("permuted_diagonals"), num_rows );
Get_Matrix_Diagonals gmd(newxadj_, newadj_, permuted_adj_vals, permuted_diagonals);
/*
int teamSizeMax = 0;
int vector_size = 0;
int max_allowed_team_size = team_policy_t::team_size_max(gmd);
this->handle->get_gs_handle()->vector_team_size(max_allowed_team_size, vector_size, teamSizeMax, num_rows, nnz);
Kokkos::parallel_for(
team_policy_t(num_rows / teamSizeMax + 1 , teamSizeMax, vector_size),
gmd );
*/
Kokkos::parallel_for(
my_exec_space(0,num_rows),
gmd );
MyExecSpace::fence();
this->handle->get_gs_handle()->set_permuted_diagonals(permuted_diagonals);
this->handle->get_gs_handle()->set_call_numeric(true);
}
#ifdef KOKKOSKERNELS_TIME_REVERSE
std::cout << "NUMERIC:" << timer.seconds() << std::endl;
#endif
}
template <typename x_value_array_type, typename y_value_array_type>
void apply(
x_value_array_type x_lhs_output_vec,
y_value_array_type y_rhs_input_vec,
bool init_zero_x_vector = false,
int numIter = 1,
bool apply_forward = true,
bool apply_backward = true,
bool update_y_vector = true){
if (this->handle->get_gs_handle()->is_numeric_called() == false){
this->initialize_numeric();
}
typename HandleType::GaussSeidelHandleType *gsHandler = this->handle->get_gs_handle();
scalar_persistent_work_view_t Permuted_Yvector = gsHandler->get_permuted_y_vector();
scalar_persistent_work_view_t Permuted_Xvector = gsHandler->get_permuted_x_vector();
row_lno_persistent_work_view_t newxadj_ = gsHandler->get_new_xadj();
nnz_lno_persistent_work_view_t old_to_new_map = gsHandler->get_old_to_new_map();
nnz_lno_persistent_work_view_t newadj_ = gsHandler->get_new_adj();
nnz_lno_persistent_work_view_t color_adj = gsHandler->get_color_adj();
color_t numColors = gsHandler->get_num_colors();
if (update_y_vector){
KokkosKernels::Experimental::Util::permute_vector
<y_value_array_type,
scalar_persistent_work_view_t,
nnz_lno_persistent_work_view_t, MyExecSpace>(
num_rows,
old_to_new_map,
y_rhs_input_vec,
Permuted_Yvector
);
}
MyExecSpace::fence();
if(init_zero_x_vector){
KokkosKernels::Experimental::Util::zero_vector<scalar_persistent_work_view_t, MyExecSpace>(num_cols, Permuted_Xvector);
}
else{
KokkosKernels::Experimental::Util::permute_vector
<x_value_array_type, scalar_persistent_work_view_t, nnz_lno_persistent_work_view_t, MyExecSpace>(
num_cols,
old_to_new_map,
x_lhs_output_vec,
Permuted_Xvector
);
}
MyExecSpace::fence();
row_lno_persistent_work_view_t permuted_xadj = gsHandler->get_new_xadj();
nnz_lno_persistent_work_view_t permuted_adj = gsHandler->get_new_adj();
scalar_persistent_work_view_t permuted_adj_vals = gsHandler->get_new_adj_val();
scalar_persistent_work_view_t permuted_diagonals = gsHandler->get_permuted_diagonals();
nnz_lno_persistent_work_host_view_t h_color_xadj = gsHandler->get_color_xadj();
if (gsHandler->get_algorithm_type()== GS_PERMUTED){
PSGS gs(permuted_xadj, permuted_adj, permuted_adj_vals,
Permuted_Xvector, Permuted_Yvector, color_adj, permuted_diagonals);
this->IterativePSGS(
gs,
numColors,
h_color_xadj,
numIter,
apply_forward,
apply_backward);
}
else{
Team_PSGS gs(permuted_xadj, permuted_adj, permuted_adj_vals,
Permuted_Xvector, Permuted_Yvector,0,0, permuted_diagonals);
this->IterativePSGS(
gs,
numColors,
h_color_xadj,
numIter,
apply_forward,
apply_backward);
}
//Kokkos::parallel_for( my_exec_space(0,nr), PermuteVector(x_lhs_output_vec, Permuted_Xvector, color_adj));
KokkosKernels::Experimental::Util::permute_vector
<scalar_persistent_work_view_t,x_value_array_type, nnz_lno_persistent_work_view_t, MyExecSpace>(
num_cols,
color_adj,
Permuted_Xvector,
x_lhs_output_vec
);
MyExecSpace::fence();
}
void IterativePSGS(
Team_PSGS &gs,
color_t numColors,
nnz_lno_persistent_work_host_view_t h_color_xadj,
int num_iteration,
bool apply_forward,
bool apply_backward){
for (int i = 0; i < num_iteration; ++i){
this->DoPSGS(gs, numColors, h_color_xadj, apply_forward, apply_backward);
}
}
void DoPSGS(Team_PSGS &gs, color_t numColors, nnz_lno_persistent_work_host_view_t h_color_xadj,
bool apply_forward,
bool apply_backward){
int teamSizeMax = 0;
int vector_size = 0;
int max_allowed_team_size = team_policy_t::team_size_max(gs);
size_type nnz = this->entries.dimension_0();
this->handle->get_gs_handle()->vector_team_size(max_allowed_team_size, vector_size, teamSizeMax, num_rows, nnz);
/*std::cout
<< "max_allowed_team_size" << max_allowed_team_size
<< " vector_size:" << vector_size
<< " teamSizeMax:" << teamSizeMax << std::endl;
*/
if (apply_forward){
for (color_t i = 0; i < numColors; ++i){
nnz_lno_t color_index_begin = h_color_xadj(i);
nnz_lno_t color_index_end = h_color_xadj(i + 1);
int overall_work = color_index_end - color_index_begin;// /256 + 1;
gs._color_set_begin = color_index_begin;
gs._color_set_end = color_index_end;
Kokkos::parallel_for(
team_policy_t(overall_work / teamSizeMax + 1 , teamSizeMax, vector_size),
gs );
MyExecSpace::fence();
}
}
if (apply_backward){
if (numColors > 0)
for (color_t i = numColors - 1; ; --i){
nnz_lno_t color_index_begin = h_color_xadj(i);
nnz_lno_t color_index_end = h_color_xadj(i + 1);
nnz_lno_t numberOfTeams = color_index_end - color_index_begin;// /256 + 1;
gs._color_set_begin = color_index_begin;
gs._color_set_end = color_index_end;
Kokkos::parallel_for(
team_policy_t(numberOfTeams / teamSizeMax + 1 , teamSizeMax, vector_size),
gs );
MyExecSpace::fence();
if (i == 0){
break;
}
}
}
}
void IterativePSGS(
PSGS &gs,
color_t numColors,
nnz_lno_persistent_work_host_view_t h_color_xadj,
int num_iteration,
bool apply_forward,
bool apply_backward){
for (int i = 0; i < num_iteration; ++i){
this->DoPSGS(gs, numColors, h_color_xadj, apply_forward, apply_backward);
}
}
void DoPSGS(PSGS &gs, color_t numColors, nnz_lno_persistent_work_host_view_t h_color_xadj,
bool apply_forward,
bool apply_backward){
if (apply_forward){
for (color_t i = 0; i < numColors; ++i){
nnz_lno_t color_index_begin = h_color_xadj(i);
nnz_lno_t color_index_end = h_color_xadj(i + 1);
Kokkos::parallel_for (my_exec_space (color_index_begin, color_index_end) , gs);
MyExecSpace::fence();
}
}
if (apply_backward && numColors){
for (size_type i = numColors - 1; ; --i){
nnz_lno_t color_index_begin = h_color_xadj(i);
nnz_lno_t color_index_end = h_color_xadj(i + 1);
Kokkos::parallel_for (my_exec_space (color_index_begin, color_index_end) , gs);
MyExecSpace::fence();
if (i == 0){
break;
}
}
}
}
};
}
}
}
}
#endif
|