/usr/include/viennacl/linalg/matrix_operations.hpp is in libviennacl-dev 1.5.1-1.
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 | #ifndef VIENNACL_LINALG_MATRIX_OPERATIONS_HPP_
#define VIENNACL_LINALG_MATRIX_OPERATIONS_HPP_
/* =========================================================================
Copyright (c) 2010-2014, Institute for Microelectronics,
Institute for Analysis and Scientific Computing,
TU Wien.
Portions of this software are copyright by UChicago Argonne, LLC.
-----------------
ViennaCL - The Vienna Computing Library
-----------------
Project Head: Karl Rupp rupp@iue.tuwien.ac.at
(A list of authors and contributors can be found in the PDF manual)
License: MIT (X11), see file LICENSE in the base directory
============================================================================= */
/** @file viennacl/linalg/matrix_operations.hpp
@brief Implementations of dense matrix related operations including matrix-vector products.
*/
#include "viennacl/forwards.h"
#include "viennacl/scalar.hpp"
#include "viennacl/vector.hpp"
#include "viennacl/vector_proxy.hpp"
#include "viennacl/tools/tools.hpp"
#include "viennacl/meta/enable_if.hpp"
#include "viennacl/meta/predicate.hpp"
#include "viennacl/meta/result_of.hpp"
#include "viennacl/traits/size.hpp"
#include "viennacl/traits/start.hpp"
#include "viennacl/traits/handle.hpp"
#include "viennacl/traits/stride.hpp"
#include "viennacl/vector.hpp"
#include "viennacl/linalg/host_based/matrix_operations.hpp"
#ifdef VIENNACL_WITH_OPENCL
#include "viennacl/linalg/opencl/matrix_operations.hpp"
#endif
#ifdef VIENNACL_WITH_CUDA
#include "viennacl/linalg/cuda/matrix_operations.hpp"
#endif
namespace viennacl
{
namespace linalg
{
template <typename NumericT, typename F,
typename ScalarType1>
void am(matrix_base<NumericT, F> & mat1,
matrix_base<NumericT, F> const & mat2, ScalarType1 const & alpha, vcl_size_t len_alpha, bool reciprocal_alpha, bool flip_sign_alpha)
{
switch (viennacl::traits::handle(mat1).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::am(mat1, mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::am(mat1, mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::am(mat1, mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F,
typename ScalarType1, typename ScalarType2>
void ambm(matrix_base<NumericT, F> & mat1,
matrix_base<NumericT, F> const & mat2, ScalarType1 const & alpha, vcl_size_t len_alpha, bool reciprocal_alpha, bool flip_sign_alpha,
matrix_base<NumericT, F> const & mat3, ScalarType2 const & beta, vcl_size_t len_beta, bool reciprocal_beta, bool flip_sign_beta)
{
switch (viennacl::traits::handle(mat1).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::ambm(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::ambm(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::ambm(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F,
typename ScalarType1, typename ScalarType2>
void ambm_m(matrix_base<NumericT, F> & mat1,
matrix_base<NumericT, F> const & mat2, ScalarType1 const & alpha, vcl_size_t len_alpha, bool reciprocal_alpha, bool flip_sign_alpha,
matrix_base<NumericT, F> const & mat3, ScalarType2 const & beta, vcl_size_t len_beta, bool reciprocal_beta, bool flip_sign_beta)
{
switch (viennacl::traits::handle(mat1).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::ambm_m(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::ambm_m(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::ambm_m(mat1,
mat2, alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
mat3, beta, len_beta, reciprocal_beta, flip_sign_beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F>
void matrix_assign(matrix_base<NumericT, F> & mat, NumericT s, bool clear = false)
{
switch (viennacl::traits::handle(mat).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_assign(mat, s, clear);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_assign(mat, s, clear);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_assign(mat, s, clear);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F>
void matrix_diagonal_assign(matrix_base<NumericT, F> & mat, NumericT s)
{
switch (viennacl::traits::handle(mat).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_diagonal_assign(mat, s);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_diagonal_assign(mat, s);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_diagonal_assign(mat, s);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Dispatcher interface for A = diag(v, k) */
template <typename NumericT, typename F>
void matrix_diag_from_vector(const vector_base<NumericT> & v, int k, matrix_base<NumericT, F> & A)
{
switch (viennacl::traits::handle(v).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_diag_from_vector(v, k, A);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_diag_from_vector(v, k, A);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_diag_from_vector(v, k, A);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Dispatcher interface for v = diag(A, k) */
template <typename NumericT, typename F>
void matrix_diag_to_vector(const matrix_base<NumericT, F> & A, int k, vector_base<NumericT> & v)
{
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_diag_to_vector(A, k, v);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_diag_to_vector(A, k, v);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_diag_to_vector(A, k, v);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F>
void matrix_row(const matrix_base<NumericT, F> & A, unsigned int i, vector_base<NumericT> & v)
{
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_row(A, i, v);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_row(A, i, v);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_row(A, i, v);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
template <typename NumericT, typename F>
void matrix_column(const matrix_base<NumericT, F> & A, unsigned int j, vector_base<NumericT> & v)
{
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::matrix_column(A, j, v);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::matrix_column(A, j, v);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::matrix_column(A, j, v);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Computes the Frobenius norm of a matrix - dispatcher interface
*
* @param A The matrix
* @param result The result scalar
*/
template <typename T, typename F>
void norm_frobenius_impl(matrix_base<T, F> const & A,
scalar<T> & result)
{
typedef typename matrix_base<T, F>::handle_type HandleType;
viennacl::vector_base<T> temp(const_cast<HandleType &>(A.handle()), A.internal_size(), 0, 1);
norm_2_impl(temp, result);
}
/** @brief Computes the Frobenius norm of a vector with final reduction on the CPU
*
* @param A The matrix
* @param result The result scalar
*/
template <typename T, typename F>
void norm_frobenius_cpu(matrix_base<T, F> const & A,
T & result)
{
typedef typename matrix_base<T, F>::handle_type HandleType;
viennacl::vector_base<T> temp(const_cast<HandleType &>(A.handle()), A.internal_size(), 0, 1);
norm_2_cpu(temp, result);
}
//
///////////////////////// matrix-vector products /////////////////////////////////
//
// A * x
/** @brief Carries out matrix-vector multiplication
*
* Implementation of the convenience expression result = prod(mat, vec);
*
* @param mat The matrix
* @param vec The vector
* @param result The result vector
*/
template <typename NumericT, typename F>
void prod_impl(const matrix_base<NumericT, F> & mat,
const vector_base<NumericT> & vec,
vector_base<NumericT> & result)
{
assert( (viennacl::traits::size1(mat) == viennacl::traits::size(result)) && bool("Size check failed at v1 = prod(A, v2): size1(A) != size(v1)"));
assert( (viennacl::traits::size2(mat) == viennacl::traits::size(vec)) && bool("Size check failed at v1 = prod(A, v2): size2(A) != size(v2)"));
switch (viennacl::traits::handle(mat).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(mat, vec, result);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(mat, vec, result);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(mat, vec, result);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
// trans(A) * x
/** @brief Carries out matrix-vector multiplication with a transposed matrix
*
* Implementation of the convenience expression result = trans(mat) * vec;
*
* @param mat_trans The transposed matrix proxy
* @param vec The vector
* @param result The result vector
*/
template <typename NumericT, typename F>
void prod_impl(const matrix_expression< const matrix_base<NumericT, F>, const matrix_base<NumericT, F>, op_trans> & mat_trans,
const vector_base<NumericT> & vec,
vector_base<NumericT> & result)
{
assert( (viennacl::traits::size1(mat_trans.lhs()) == viennacl::traits::size(vec)) && bool("Size check failed at v1 = trans(A) * v2: size1(A) != size(v2)"));
assert( (viennacl::traits::size2(mat_trans.lhs()) == viennacl::traits::size(result)) && bool("Size check failed at v1 = trans(A) * v2: size2(A) != size(v1)"));
switch (viennacl::traits::handle(mat_trans.lhs()).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(mat_trans, vec, result);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(mat_trans, vec, result);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(mat_trans, vec, result);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
//
///////////////////////// matrix-matrix products /////////////////////////////////
//
/** @brief Carries out matrix-matrix multiplication
*
* Implementation of C = prod(A, B);
*
*/
template <typename NumericT, typename F1, typename F2, typename F3, typename ScalarType >
void prod_impl(const matrix_base<NumericT, F1> & A,
const matrix_base<NumericT, F2> & B,
matrix_base<NumericT, F3> & C,
ScalarType alpha,
ScalarType beta)
{
assert( (viennacl::traits::size1(A) == viennacl::traits::size1(C)) && bool("Size check failed at C = prod(A, B): size1(A) != size1(C)"));
assert( (viennacl::traits::size2(A) == viennacl::traits::size1(B)) && bool("Size check failed at C = prod(A, B): size2(A) != size1(B)"));
assert( (viennacl::traits::size2(B) == viennacl::traits::size2(C)) && bool("Size check failed at C = prod(A, B): size2(B) != size2(C)"));
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(A, B, C, alpha, beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(A, B, C, alpha, beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(A, B, C, alpha, beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Carries out matrix-matrix multiplication
*
* Implementation of C = prod(trans(A), B);
*
*/
template <typename NumericT, typename F1, typename F2, typename F3, typename ScalarType >
void prod_impl(const viennacl::matrix_expression< const matrix_base<NumericT, F1>,
const matrix_base<NumericT, F1>,
op_trans> & A,
const matrix_base<NumericT, F2> & B,
matrix_base<NumericT, F3> & C,
ScalarType alpha,
ScalarType beta)
{
assert(viennacl::traits::size2(A.lhs()) == viennacl::traits::size1(C) && bool("Size check failed at C = prod(trans(A), B): size2(A) != size1(C)"));
assert(viennacl::traits::size1(A.lhs()) == viennacl::traits::size1(B) && bool("Size check failed at C = prod(trans(A), B): size1(A) != size1(B)"));
assert(viennacl::traits::size2(B) == viennacl::traits::size2(C) && bool("Size check failed at C = prod(trans(A), B): size2(B) != size2(C)"));
switch (viennacl::traits::handle(A.lhs()).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(A, B, C, alpha, beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(A, B, C, alpha, beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(A, B, C, alpha, beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Carries out matrix-matrix multiplication
*
* Implementation of C = prod(A, trans(B));
*
*/
template <typename NumericT, typename F1, typename F2, typename F3, typename ScalarType >
void prod_impl(const matrix_base<NumericT, F1> & A,
const viennacl::matrix_expression< const matrix_base<NumericT, F2>, const matrix_base<NumericT, F2>, op_trans> & B,
matrix_base<NumericT, F3> & C,
ScalarType alpha,
ScalarType beta)
{
assert(viennacl::traits::size1(A) == viennacl::traits::size1(C) && bool("Size check failed at C = prod(A, trans(B)): size1(A) != size1(C)"));
assert(viennacl::traits::size2(A) == viennacl::traits::size2(B.lhs()) && bool("Size check failed at C = prod(A, trans(B)): size2(A) != size2(B)"));
assert(viennacl::traits::size1(B.lhs()) == viennacl::traits::size2(C) && bool("Size check failed at C = prod(A, trans(B)): size1(B) != size2(C)"));
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(A, B, C, alpha, beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(A, B, C, alpha, beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(A, B, C, alpha, beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
/** @brief Carries out matrix-matrix multiplication
*
* Implementation of C = prod(trans(A), trans(B));
*
*/
template <typename NumericT, typename F1, typename F2, typename F3, typename ScalarType >
void prod_impl(const viennacl::matrix_expression< const matrix_base<NumericT, F1>, const matrix_base<NumericT, F1>, op_trans> & A,
const viennacl::matrix_expression< const matrix_base<NumericT, F2>, const matrix_base<NumericT, F2>, op_trans> & B,
matrix_base<NumericT, F3> & C,
ScalarType alpha,
ScalarType beta)
{
assert(viennacl::traits::size2(A.lhs()) == viennacl::traits::size1(C) && bool("Size check failed at C = prod(trans(A), trans(B)): size2(A) != size1(C)"));
assert(viennacl::traits::size1(A.lhs()) == viennacl::traits::size2(B.lhs()) && bool("Size check failed at C = prod(trans(A), trans(B)): size1(A) != size2(B)"));
assert(viennacl::traits::size1(B.lhs()) == viennacl::traits::size2(C) && bool("Size check failed at C = prod(trans(A), trans(B)): size1(B) != size2(C)"));
switch (viennacl::traits::handle(A.lhs()).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::prod_impl(A, B, C, alpha, beta);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::prod_impl(A, B, C, alpha, beta);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::prod_impl(A, B, C, alpha, beta);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
///////////////////////// Elementwise operations /////////////
/** @brief Implementation of the element-wise operation A = B .* C and A = B ./ C for matrices (using MATLAB syntax). Don't use this function directly, use element_prod() and element_div().
*
* @param A The result matrix (or -range, or -slice)
* @param proxy The proxy object holding B, C, and the operation
*/
template <typename T, typename F, typename OP>
void element_op(matrix_base<T, F> & A,
matrix_expression<const matrix_base<T, F>, const matrix_base<T, F>, OP> const & proxy)
{
assert( (viennacl::traits::size1(A) == viennacl::traits::size1(proxy)) && bool("Size check failed at A = element_op(B): size1(A) != size1(B)"));
assert( (viennacl::traits::size2(A) == viennacl::traits::size2(proxy)) && bool("Size check failed at A = element_op(B): size2(A) != size2(B)"));
switch (viennacl::traits::handle(A).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::element_op(A, proxy);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::element_op(A, proxy);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::element_op(A, proxy);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
#define VIENNACL_MAKE_BINARY_OP(OPNAME)\
template <typename T, typename F>\
viennacl::matrix_expression<const matrix_base<T, F>, const matrix_base<T, F>, op_element_binary<op_##OPNAME> >\
element_##OPNAME(matrix_base<T, F> const & A, matrix_base<T, F> const & B)\
{\
return viennacl::matrix_expression<const matrix_base<T, F>, const matrix_base<T, F>, op_element_binary<op_##OPNAME> >(A, B);\
}\
\
template <typename M1, typename M2, typename OP, typename T, typename F>\
viennacl::matrix_expression<const matrix_expression<const M1, const M2, OP>,\
const matrix_base<T, F>,\
op_element_binary<op_##OPNAME> >\
element_##OPNAME(matrix_expression<const M1, const M2, OP> const & proxy, matrix_base<T, F> const & B)\
{\
return viennacl::matrix_expression<const matrix_expression<const M1, const M2, OP>,\
const matrix_base<T, F>,\
op_element_binary<op_##OPNAME> >(proxy, B);\
}\
\
template <typename T, typename F, typename M2, typename M3, typename OP>\
viennacl::matrix_expression<const matrix_base<T, F>,\
const matrix_expression<const M2, const M3, OP>,\
op_element_binary<op_##OPNAME> >\
element_##OPNAME(matrix_base<T, F> const & A, matrix_expression<const M2, const M3, OP> const & proxy)\
{\
return viennacl::matrix_expression<const matrix_base<T, F>,\
const matrix_expression<const M2, const M3, OP>,\
op_element_binary<op_##OPNAME> >(A, proxy);\
}\
\
template <typename M1, typename M2, typename OP1,\
typename M3, typename M4, typename OP2>\
viennacl::matrix_expression<const matrix_expression<const M1, const M2, OP1>,\
const matrix_expression<const M3, const M4, OP2>,\
op_element_binary<op_##OPNAME> >\
element_##OPNAME(matrix_expression<const M1, const M2, OP1> const & proxy1,\
matrix_expression<const M3, const M4, OP2> const & proxy2)\
{\
return viennacl::matrix_expression<const matrix_expression<const M1, const M2, OP1>,\
const matrix_expression<const M3, const M4, OP2>,\
op_element_binary<op_##OPNAME> >(proxy1, proxy2);\
}
VIENNACL_MAKE_BINARY_OP(prod)
VIENNACL_MAKE_BINARY_OP(div)
VIENNACL_MAKE_BINARY_OP(pow)
#undef VIENNACL_GENERATE_BINARY_OP_OVERLOADS
#define VIENNACL_MAKE_UNARY_ELEMENT_OP(funcname) \
template <typename T, typename F> \
viennacl::matrix_expression<const matrix_base<T, F>, const matrix_base<T, F>, op_element_unary<op_##funcname> > \
element_##funcname(matrix_base<T, F> const & A) \
{ \
return viennacl::matrix_expression<const matrix_base<T, F>, const matrix_base<T, F>, op_element_unary<op_##funcname> >(A, A); \
} \
template <typename LHS, typename RHS, typename OP> \
viennacl::matrix_expression<const matrix_expression<const LHS, const RHS, OP>, \
const matrix_expression<const LHS, const RHS, OP>, \
op_element_unary<op_##funcname> > \
element_##funcname(matrix_expression<const LHS, const RHS, OP> const & proxy) \
{ \
return viennacl::matrix_expression<const matrix_expression<const LHS, const RHS, OP>, \
const matrix_expression<const LHS, const RHS, OP>, \
op_element_unary<op_##funcname> >(proxy, proxy); \
} \
VIENNACL_MAKE_UNARY_ELEMENT_OP(abs)
VIENNACL_MAKE_UNARY_ELEMENT_OP(acos)
VIENNACL_MAKE_UNARY_ELEMENT_OP(asin)
VIENNACL_MAKE_UNARY_ELEMENT_OP(atan)
VIENNACL_MAKE_UNARY_ELEMENT_OP(ceil)
VIENNACL_MAKE_UNARY_ELEMENT_OP(cos)
VIENNACL_MAKE_UNARY_ELEMENT_OP(cosh)
VIENNACL_MAKE_UNARY_ELEMENT_OP(exp)
VIENNACL_MAKE_UNARY_ELEMENT_OP(fabs)
VIENNACL_MAKE_UNARY_ELEMENT_OP(floor)
VIENNACL_MAKE_UNARY_ELEMENT_OP(log)
VIENNACL_MAKE_UNARY_ELEMENT_OP(log10)
VIENNACL_MAKE_UNARY_ELEMENT_OP(sin)
VIENNACL_MAKE_UNARY_ELEMENT_OP(sinh)
VIENNACL_MAKE_UNARY_ELEMENT_OP(sqrt)
VIENNACL_MAKE_UNARY_ELEMENT_OP(tan)
VIENNACL_MAKE_UNARY_ELEMENT_OP(tanh)
#undef VIENNACL_MAKE_UNARY_ELEMENT_OP
//
///////////////////////// miscellaneous operations /////////////////////////////////
//
/** @brief Returns a proxy class for the operation mat += vec1 * vec2^T, i.e. a rank 1 update
*
* @param vec1 The first vector
* @param vec2 The second vector
*/
template <typename NumericT>
viennacl::matrix_expression<const vector_base<NumericT>, const vector_base<NumericT>, op_prod>
outer_prod(const vector_base<NumericT> & vec1, const vector_base<NumericT> & vec2)
{
return viennacl::matrix_expression< const vector_base<NumericT>, const vector_base<NumericT>, op_prod>(vec1, vec2);
}
/** @brief The implementation of the operation mat += alpha * vec1 * vec2^T, i.e. a scaled rank 1 update
*
* Implementation of the convenience expression result += alpha * outer_prod(vec1, vec2);
*
* @param mat1 The matrix to be updated
* @param alpha The scaling factor (either a viennacl::scalar<>, float, or double)
* @param len_alpha Length of the buffer for an eventual final reduction step (currently always '1')
* @param reciprocal_alpha Use 1/alpha instead of alpha
* @param flip_sign_alpha Use -alpha instead of alpha
* @param vec1 The first vector
* @param vec2 The second vector
*/
template <typename NumericT, typename F, typename S1>
void scaled_rank_1_update(matrix_base<NumericT, F> & mat1,
S1 const & alpha, vcl_size_t len_alpha, bool reciprocal_alpha, bool flip_sign_alpha,
const vector_base<NumericT> & vec1,
const vector_base<NumericT> & vec2)
{
switch (viennacl::traits::handle(mat1).get_active_handle_id())
{
case viennacl::MAIN_MEMORY:
viennacl::linalg::host_based::scaled_rank_1_update(mat1,
alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
vec1, vec2);
break;
#ifdef VIENNACL_WITH_OPENCL
case viennacl::OPENCL_MEMORY:
viennacl::linalg::opencl::scaled_rank_1_update(mat1,
alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
vec1, vec2);
break;
#endif
#ifdef VIENNACL_WITH_CUDA
case viennacl::CUDA_MEMORY:
viennacl::linalg::cuda::scaled_rank_1_update(mat1,
alpha, len_alpha, reciprocal_alpha, flip_sign_alpha,
vec1, vec2);
break;
#endif
case viennacl::MEMORY_NOT_INITIALIZED:
throw memory_exception("not initialised!");
default:
throw memory_exception("not implemented");
}
}
} //namespace linalg
//
///////////////////////// Operator overloads /////////////////////////////////
//
//v += A * x
/** @brief Implementation of the operation v1 += A * v2, where A is a matrix
*
* @param v1 The result vector v1 where A * v2 is added to
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator+=(vector_base<NumericT> & v1,
const viennacl::vector_expression< const matrix_base<NumericT, F>, const vector_base<NumericT>, viennacl::op_prod> & proxy)
{
assert(viennacl::traits::size1(proxy.lhs()) == v1.size() && bool("Size check failed for v1 += A * v2: size1(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size1(proxy.lhs()));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
v1 += result;
return v1;
}
/** @brief Implementation of the operation v1 -= A * v2, where A is a matrix
*
* @param v1 The result vector v1 where A * v2 is subtracted from
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator-=(vector_base<NumericT> & v1,
const viennacl::vector_expression< const matrix_base<NumericT, F>, const vector_base<NumericT>, viennacl::op_prod> & proxy)
{
assert(viennacl::traits::size1(proxy.lhs()) == v1.size() && bool("Size check failed for v1 -= A * v2: size1(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size1(proxy.lhs()));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
v1 -= result;
return v1;
}
//free functions:
/** @brief Implementation of the operation 'result = v1 + A * v2', where A is a matrix
*
* @param v1 The addend vector.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
viennacl::vector<NumericT>
operator+(const vector_base<NumericT> & v1,
const vector_expression< const matrix_base<NumericT, F>, const vector_base<NumericT>, op_prod> & proxy)
{
assert(viennacl::traits::size1(proxy.lhs()) == viennacl::traits::size(v1) && bool("Size check failed for v1 + A * v2: size1(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size(v1));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
result += v1;
return result;
}
/** @brief Implementation of the operation 'result = v1 - A * v2', where A is a matrix
*
* @param v1 The addend vector.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
viennacl::vector<NumericT>
operator-(const vector_base<NumericT> & v1,
const vector_expression< const matrix_base<NumericT, F>, const vector_base<NumericT>, op_prod> & proxy)
{
assert(viennacl::traits::size1(proxy.lhs()) == viennacl::traits::size(v1) && bool("Size check failed for v1 - A * v2: size1(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size(v1));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
result = v1 - result;
return result;
}
////////// transposed_matrix_proxy
//v += A^T * x
/** @brief Implementation of the operation v1 += A * v2, where A is a matrix
*
* @param v1 The addend vector where the result is written to.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator+=(vector_base<NumericT> & v1,
const vector_expression< const matrix_expression<const matrix_base<NumericT, F>, const matrix_base<NumericT, F>, op_trans>,
const vector_base<NumericT>,
op_prod> & proxy)
{
assert(viennacl::traits::size2(proxy.lhs()) == v1.size() && bool("Size check failed in v1 += trans(A) * v2: size2(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size2(proxy.lhs()));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
v1 += result;
return v1;
}
//v -= A^T * x
/** @brief Implementation of the operation v1 -= A * v2, where A is a matrix
*
* @param v1 The addend vector where the result is written to.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator-=(vector_base<NumericT> & v1,
const vector_expression< const matrix_expression<const matrix_base<NumericT, F>, const matrix_base<NumericT, F>, op_trans>,
const vector_base<NumericT>,
op_prod> & proxy)
{
assert(viennacl::traits::size2(proxy.lhs()) == v1.size() && bool("Size check failed in v1 += trans(A) * v2: size2(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size2(proxy.lhs()));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
v1 -= result;
return v1;
}
//free functions:
/** @brief Implementation of the operation 'result = v1 + A * v2', where A is a matrix
*
* @param v1 The addend vector.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator+(const vector_base<NumericT> & v1,
const vector_expression< const matrix_expression<const matrix_base<NumericT, F>, const matrix_base<NumericT, F>, op_trans>,
const vector_base<NumericT>,
op_prod> & proxy)
{
assert(viennacl::traits::size2(proxy.lhs()) == viennacl::traits::size(v1) && bool("Size check failed in v1 + trans(A) * v2: size2(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size(v1));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
result += v1;
return result;
}
/** @brief Implementation of the operation 'result = v1 - A * v2', where A is a matrix
*
* @param v1 The addend vector.
* @param proxy An expression template proxy class.
*/
template <typename NumericT, typename F>
vector<NumericT>
operator-(const vector_base<NumericT> & v1,
const vector_expression< const matrix_expression<const matrix_base<NumericT, F>, const matrix_base<NumericT, F>, op_trans>,
const vector_base<NumericT>,
op_prod> & proxy)
{
assert(viennacl::traits::size2(proxy.lhs()) == viennacl::traits::size(v1) && bool("Size check failed in v1 - trans(A) * v2: size2(A) != size(v1)"));
vector<NumericT> result(viennacl::traits::size(v1));
viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), result);
result = v1 - result;
return result;
}
} //namespace viennacl
#endif
|