Hu Chunming / vpt_ascend

opencl_kernels_features2d.cpp 27.6 KB
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 
// This file is auto-generated. Do not edit!

#include "opencv2/core.hpp"
#include "cvconfig.h"
#include "opencl_kernels_features2d.hpp"

#ifdef HAVE_OPENCL

namespace cv
{
namespace ocl
{
namespace features2d
{

static const char* const moduleName = "features2d";

struct cv::ocl::internal::ProgramEntry akaze_oclsrc={moduleName, "akaze",
"__kernel void\n"
"AKAZE_pm_g2(__global const float* lx, __global const float* ly, __global float* dst,\n"
"float k, int size)\n"
"{\n"
"int i = get_global_id(0);\n"
"if (!(i < size))\n"
"{\n"
"return;\n"
"}\n"
"const float k2inv = 1.0f / (k * k);\n"
"dst[i] = 1.0f / (1.0f + ((lx[i] * lx[i] + ly[i] * ly[i]) * k2inv));\n"
"}\n"
"__kernel void\n"
"AKAZE_nld_step_scalar(__global const float* lt, int lt_step, int lt_offset, int rows, int cols,\n"
"__global const float* lf, __global float* dst, float step_size)\n"
"{\n"
"int i = get_global_id(1);\n"
"int j = get_global_id(0);\n"
"if (!(i < rows && j < cols))\n"
"{\n"
"return;\n"
"}\n"
"int a = (i - 1) * cols;\n"
"int c = (i    ) * cols;\n"
"int b = (i + 1) * cols;\n"
"float res = 0.0f;\n"
"if (i == 0)\n"
"{\n"
"if (j == 0 || j == (cols - 1))\n"
"{\n"
"res = 0.0f;\n"
"} else\n"
"{\n"
"res = (lf[c + j] + lf[c + j + 1])*(lt[c + j + 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[c + j - 1])*(lt[c + j - 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[b + j    ])*(lt[b + j    ] - lt[c + j]);\n"
"}\n"
"} else if (i == (rows - 1))\n"
"{\n"
"if (j == 0 || j == (cols - 1))\n"
"{\n"
"res = 0.0f;\n"
"} else\n"
"{\n"
"res = (lf[c + j] + lf[c + j + 1])*(lt[c + j + 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[c + j - 1])*(lt[c + j - 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[a + j    ])*(lt[a + j    ] - lt[c + j]);\n"
"}\n"
"} else\n"
"{\n"
"if (j == 0)\n"
"{\n"
"res = (lf[c + 0] + lf[c + 1])*(lt[c + 1] - lt[c + 0]) +\n"
"(lf[c + 0] + lf[b + 0])*(lt[b + 0] - lt[c + 0]) +\n"
"(lf[c + 0] + lf[a + 0])*(lt[a + 0] - lt[c + 0]);\n"
"} else if (j == (cols - 1))\n"
"{\n"
"res = (lf[c + j] + lf[c + j - 1])*(lt[c + j - 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[b + j    ])*(lt[b + j    ] - lt[c + j]) +\n"
"(lf[c + j] + lf[a + j    ])*(lt[a + j    ] - lt[c + j]);\n"
"} else\n"
"{\n"
"res = (lf[c + j] + lf[c + j + 1])*(lt[c + j + 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[c + j - 1])*(lt[c + j - 1] - lt[c + j]) +\n"
"(lf[c + j] + lf[b + j    ])*(lt[b + j    ] - lt[c + j]) +\n"
"(lf[c + j] + lf[a + j    ])*(lt[a + j    ] - lt[c + j]);\n"
"}\n"
"}\n"
"dst[c + j] = res * step_size;\n"
"}\n"
"__kernel void\n"
"AKAZE_compute_determinant(__global const float* lxx, __global const float* lxy, __global const float* lyy,\n"
"__global float* dst, float sigma, int size)\n"
"{\n"
"int i = get_global_id(0);\n"
"if (!(i < size))\n"
"{\n"
"return;\n"
"}\n"
"dst[i] = (lxx[i] * lyy[i] - lxy[i] * lxy[i]) * sigma;\n"
"}\n"
, "80f6cd2f334b70062ed64a0a1a866593", NULL};
struct cv::ocl::internal::ProgramEntry brute_force_match_oclsrc={moduleName, "brute_force_match",
"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics:enable\n"
"#define MAX_FLOAT 3.40282e+038f\n"
"#ifndef T\n"
"#define T float\n"
"#endif\n"
"#ifndef BLOCK_SIZE\n"
"#define BLOCK_SIZE 16\n"
"#endif\n"
"#ifndef MAX_DESC_LEN\n"
"#define MAX_DESC_LEN 64\n"
"#endif\n"
"#define BLOCK_SIZE_ODD          (BLOCK_SIZE + 1)\n"
"#ifndef SHARED_MEM_SZ\n"
"#  if (BLOCK_SIZE < MAX_DESC_LEN)\n"
"#    define SHARED_MEM_SZ      (kercn * (BLOCK_SIZE * MAX_DESC_LEN + BLOCK_SIZE * BLOCK_SIZE))\n"
"#  else\n"
"#    define SHARED_MEM_SZ      (kercn * 2 * BLOCK_SIZE_ODD * BLOCK_SIZE)\n"
"#  endif\n"
"#endif\n"
"#ifndef DIST_TYPE\n"
"#define DIST_TYPE 2\n"
"#endif\n"
"#if (DIST_TYPE == 2)\n"
"#   ifdef T_FLOAT\n"
"typedef float result_type;\n"
"#       if (8 == kercn)\n"
"typedef float8 value_type;\n"
"#           define DIST(x, y) {value_type d = fabs((x) - (y)); result += d.s0 + d.s1 + d.s2 + d.s3 + d.s4 + d.s5 + d.s6 + d.s7;}\n"
"#       elif (4 == kercn)\n"
"typedef float4 value_type;\n"
"#           define DIST(x, y) {value_type d = fabs((x) - (y)); result += d.s0 + d.s1 + d.s2 + d.s3;}\n"
"#       else\n"
"typedef float value_type;\n"
"#           define DIST(x, y) result += fabs((x) - (y))\n"
"#       endif\n"
"#   else\n"
"typedef int result_type;\n"
"#       if (8 == kercn)\n"
"typedef int8 value_type;\n"
"#           define DIST(x, y) {value_type d = abs((x) - (y)); result += d.s0 + d.s1 + d.s2 + d.s3 + d.s4 + d.s5 + d.s6 + d.s7;}\n"
"#       elif (4 == kercn)\n"
"typedef int4 value_type;\n"
"#           define DIST(x, y) {value_type d = abs((x) - (y)); result += d.s0 + d.s1 + d.s2 + d.s3;}\n"
"#       else\n"
"typedef int  value_type;\n"
"#           define DIST(x, y) result += abs((x) - (y))\n"
"#       endif\n"
"#   endif\n"
"#   define DIST_RES(x) (x)\n"
"#elif (DIST_TYPE == 4)\n"
"typedef float result_type;\n"
"#   if (8 == kercn)\n"
"typedef float8 value_type;\n"
"#       define DIST(x, y)   {value_type d = ((x) - (y)); result += dot(d.s0123, d.s0123) + dot(d.s4567, d.s4567);}\n"
"#   elif (4 == kercn)\n"
"typedef float4      value_type;\n"
"#       define DIST(x, y)   {value_type d = ((x) - (y)); result += dot(d, d);}\n"
"#   else\n"
"typedef float       value_type;\n"
"#       define DIST(x, y)   {value_type d = ((x) - (y)); result = mad(d, d, result);}\n"
"#   endif\n"
"#   define DIST_RES(x) sqrt(x)\n"
"#elif (DIST_TYPE == 6)\n"
"#   if (8 == kercn)\n"
"typedef int8 value_type;\n"
"#   elif (4 == kercn)\n"
"typedef int4 value_type;\n"
"#   else\n"
"typedef int value_type;\n"
"#   endif\n"
"typedef int result_type;\n"
"#   define DIST(x, y) result += popcount( (x) ^ (y) )\n"
"#   define DIST_RES(x) (x)\n"
"#endif\n"
"inline result_type reduce_block(\n"
"__local value_type *s_query,\n"
"__local value_type *s_train,\n"
"int lidx,\n"
"int lidy\n"
")\n"
"{\n"
"result_type result = 0;\n"
"#pragma unroll\n"
"for (int j = 0 ; j < BLOCK_SIZE ; j++)\n"
"{\n"
"DIST(s_query[lidy * BLOCK_SIZE_ODD + j], s_train[j * BLOCK_SIZE_ODD + lidx]);\n"
"}\n"
"return DIST_RES(result);\n"
"}\n"
"inline result_type reduce_block_match(\n"
"__local value_type *s_query,\n"
"__local value_type *s_train,\n"
"int lidx,\n"
"int lidy\n"
")\n"
"{\n"
"result_type result = 0;\n"
"#pragma unroll\n"
"for (int j = 0 ; j < BLOCK_SIZE ; j++)\n"
"{\n"
"DIST(s_query[lidy * BLOCK_SIZE_ODD + j], s_train[j * BLOCK_SIZE_ODD + lidx]);\n"
"}\n"
"return result;\n"
"}\n"
"inline result_type reduce_multi_block(\n"
"__local value_type *s_query,\n"
"__local value_type *s_train,\n"
"int block_index,\n"
"int lidx,\n"
"int lidy\n"
")\n"
"{\n"
"result_type result = 0;\n"
"#pragma unroll\n"
"for (int j = 0 ; j < BLOCK_SIZE ; j++)\n"
"{\n"
"DIST(s_query[lidy * MAX_DESC_LEN + block_index * BLOCK_SIZE + j], s_train[j * BLOCK_SIZE + lidx]);\n"
"}\n"
"return result;\n"
"}\n"
"__kernel void BruteForceMatch_Match(\n"
"__global T *query,\n"
"__global T *train,\n"
"__global int *bestTrainIdx,\n"
"__global float *bestDistance,\n"
"int query_rows,\n"
"int query_cols,\n"
"int train_rows,\n"
"int train_cols,\n"
"int step\n"
")\n"
"{\n"
"const int lidx = get_local_id(0);\n"
"const int lidy = get_local_id(1);\n"
"const int groupidx = get_group_id(0);\n"
"const int queryIdx = mad24(BLOCK_SIZE, groupidx, lidy);\n"
"const int queryOffset = min(queryIdx, query_rows - 1) * step;\n"
"__global TN *query_vec = (__global TN *)(query + queryOffset);\n"
"query_cols /= kercn;\n"
"__local float sharebuffer[SHARED_MEM_SZ];\n"
"__local value_type *s_query = (__local value_type *)sharebuffer;\n"
"#if 0 < MAX_DESC_LEN\n"
"__local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * MAX_DESC_LEN;\n"
"#pragma unroll\n"
"for (int i = 0; i < MAX_DESC_LEN / BLOCK_SIZE; i++)\n"
"{\n"
"const int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"s_query[mad24(MAX_DESC_LEN, lidy, loadx)] = loadx < query_cols ? query_vec[loadx] : 0;\n"
"}\n"
"#else\n"
"__local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE_ODD * BLOCK_SIZE;\n"
"const int s_query_i = mad24(BLOCK_SIZE_ODD, lidy, lidx);\n"
"const int s_train_i = mad24(BLOCK_SIZE_ODD, lidx, lidy);\n"
"#endif\n"
"float myBestDistance = MAX_FLOAT;\n"
"int myBestTrainIdx = -1;\n"
"for (int t = 0, endt = (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt; t++)\n"
"{\n"
"result_type result = 0;\n"
"const int trainOffset = min(mad24(BLOCK_SIZE, t, lidy), train_rows - 1) * step;\n"
"__global TN *train_vec = (__global TN *)(train + trainOffset);\n"
"#if 0 < MAX_DESC_LEN\n"
"#pragma unroll\n"
"for (int i = 0; i < MAX_DESC_LEN / BLOCK_SIZE; i++)\n"
"{\n"
"const int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"s_train[mad24(BLOCK_SIZE, lidx, lidy)] = loadx < train_cols ? train_vec[loadx] : 0;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"result += reduce_multi_block(s_query, s_train, i, lidx, lidy);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"}\n"
"#else\n"
"for (int i = 0, endq = (query_cols + BLOCK_SIZE - 1) / BLOCK_SIZE; i < endq; i++)\n"
"{\n"
"const int loadx = mad24(i, BLOCK_SIZE, lidx);\n"
"if (loadx < query_cols)\n"
"{\n"
"s_query[s_query_i] = query_vec[loadx];\n"
"s_train[s_train_i] = train_vec[loadx];\n"
"}\n"
"else\n"
"{\n"
"s_query[s_query_i] = 0;\n"
"s_train[s_train_i] = 0;\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"result += reduce_block_match(s_query, s_train, lidx, lidy);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"}\n"
"#endif\n"
"result = DIST_RES(result);\n"
"const int trainIdx = mad24(BLOCK_SIZE, t, lidx);\n"
"if (queryIdx < query_rows && trainIdx < train_rows && result < myBestDistance )\n"
"{\n"
"myBestDistance = result;\n"
"myBestTrainIdx = trainIdx;\n"
"}\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"__local float *s_distance = (__local float *)sharebuffer;\n"
"__local int *s_trainIdx = (__local int *)(sharebuffer + BLOCK_SIZE_ODD * BLOCK_SIZE);\n"
"s_distance += lidy * BLOCK_SIZE_ODD;\n"
"s_trainIdx += lidy * BLOCK_SIZE_ODD;\n"
"s_distance[lidx] = myBestDistance;\n"
"s_trainIdx[lidx] = myBestTrainIdx;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"#pragma unroll\n"
"for (int k = 0 ; k < BLOCK_SIZE; k++)\n"
"{\n"
"if (myBestDistance > s_distance[k])\n"
"{\n"
"myBestDistance = s_distance[k];\n"
"myBestTrainIdx = s_trainIdx[k];\n"
"}\n"
"}\n"
"if (queryIdx < query_rows && lidx == 0)\n"
"{\n"
"bestTrainIdx[queryIdx] = myBestTrainIdx;\n"
"bestDistance[queryIdx] = myBestDistance;\n"
"}\n"
"}\n"
"__kernel void BruteForceMatch_RadiusMatch(\n"
"__global T *query,\n"
"__global T *train,\n"
"float maxDistance,\n"
"__global int *bestTrainIdx,\n"
"__global float *bestDistance,\n"
"__global int *nMatches,\n"
"int query_rows,\n"
"int query_cols,\n"
"int train_rows,\n"
"int train_cols,\n"
"int bestTrainIdx_cols,\n"
"int step,\n"
"int ostep\n"
")\n"
"{\n"
"const int lidx = get_local_id(0);\n"
"const int lidy = get_local_id(1);\n"
"const int groupidx = get_group_id(0);\n"
"const int groupidy = get_group_id(1);\n"
"const int queryIdx = mad24(BLOCK_SIZE, groupidy, lidy);\n"
"const int queryOffset = min(queryIdx, query_rows - 1) * step;\n"
"__global TN *query_vec = (__global TN *)(query + queryOffset);\n"
"const int trainIdx = mad24(BLOCK_SIZE, groupidx, lidx);\n"
"const int trainOffset = min(mad24(BLOCK_SIZE, groupidx, lidy), train_rows - 1) * step;\n"
"__global TN *train_vec = (__global TN *)(train + trainOffset);\n"
"query_cols /= kercn;\n"
"__local float sharebuffer[SHARED_MEM_SZ];\n"
"__local value_type *s_query = (__local value_type *)sharebuffer;\n"
"__local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE_ODD * BLOCK_SIZE;\n"
"result_type result = 0;\n"
"const int s_query_i = mad24(BLOCK_SIZE_ODD, lidy, lidx);\n"
"const int s_train_i = mad24(BLOCK_SIZE_ODD, lidx, lidy);\n"
"for (int i = 0 ; i < (query_cols + BLOCK_SIZE - 1) / BLOCK_SIZE ; ++i)\n"
"{\n"
"const int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"if (loadx < query_cols)\n"
"{\n"
"s_query[s_query_i] = query_vec[loadx];\n"
"s_train[s_train_i] = train_vec[loadx];\n"
"}\n"
"else\n"
"{\n"
"s_query[s_query_i] = 0;\n"
"s_train[s_train_i] = 0;\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"result += reduce_block(s_query, s_train, lidx, lidy);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"}\n"
"if (queryIdx < query_rows && trainIdx < train_rows && convert_float(result) < maxDistance)\n"
"{\n"
"int ind = atom_inc(nMatches + queryIdx);\n"
"if(ind < bestTrainIdx_cols)\n"
"{\n"
"bestTrainIdx[mad24(queryIdx, ostep, ind)] = trainIdx;\n"
"bestDistance[mad24(queryIdx, ostep, ind)] = result;\n"
"}\n"
"}\n"
"}\n"
"__kernel void BruteForceMatch_knnMatch(\n"
"__global T *query,\n"
"__global T *train,\n"
"__global int2 *bestTrainIdx,\n"
"__global float2 *bestDistance,\n"
"int query_rows,\n"
"int query_cols,\n"
"int train_rows,\n"
"int train_cols,\n"
"int step\n"
")\n"
"{\n"
"const int lidx = get_local_id(0);\n"
"const int lidy = get_local_id(1);\n"
"const int groupidx = get_group_id(0);\n"
"const int queryIdx = mad24(BLOCK_SIZE, groupidx, lidy);\n"
"const int queryOffset = min(queryIdx, query_rows - 1) * step;\n"
"__global TN *query_vec = (__global TN *)(query + queryOffset);\n"
"query_cols /= kercn;\n"
"__local float sharebuffer[SHARED_MEM_SZ];\n"
"__local value_type *s_query = (__local value_type *)sharebuffer;\n"
"#if 0 < MAX_DESC_LEN\n"
"__local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE * MAX_DESC_LEN;\n"
"#pragma unroll\n"
"for (int i = 0 ;  i <  MAX_DESC_LEN / BLOCK_SIZE; i ++)\n"
"{\n"
"int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"s_query[mad24(MAX_DESC_LEN, lidy, loadx)] = loadx < query_cols ? query_vec[loadx] : 0;\n"
"}\n"
"#else\n"
"__local value_type *s_train = (__local value_type *)sharebuffer + BLOCK_SIZE_ODD * BLOCK_SIZE;\n"
"const int s_query_i = mad24(BLOCK_SIZE_ODD, lidy, lidx);\n"
"const int s_train_i = mad24(BLOCK_SIZE_ODD, lidx, lidy);\n"
"#endif\n"
"float myBestDistance1 = MAX_FLOAT;\n"
"float myBestDistance2 = MAX_FLOAT;\n"
"int myBestTrainIdx1 = -1;\n"
"int myBestTrainIdx2 = -1;\n"
"for (int t = 0, endt = (train_rows + BLOCK_SIZE - 1) / BLOCK_SIZE; t < endt ; t++)\n"
"{\n"
"result_type result = 0;\n"
"int trainOffset = min(mad24(BLOCK_SIZE, t, lidy), train_rows - 1) * step;\n"
"__global TN *train_vec = (__global TN *)(train + trainOffset);\n"
"#if 0 < MAX_DESC_LEN\n"
"#pragma unroll\n"
"for (int i = 0 ; i < MAX_DESC_LEN / BLOCK_SIZE ; i++)\n"
"{\n"
"const int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"s_train[mad24(BLOCK_SIZE, lidx, lidy)] = loadx < train_cols ? train_vec[loadx] : 0;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"result += reduce_multi_block(s_query, s_train, i, lidx, lidy);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"}\n"
"#else\n"
"for (int i = 0, endq = (query_cols + BLOCK_SIZE -1) / BLOCK_SIZE; i < endq ; i++)\n"
"{\n"
"const int loadx = mad24(BLOCK_SIZE, i, lidx);\n"
"if (loadx < query_cols)\n"
"{\n"
"s_query[s_query_i] = query_vec[loadx];\n"
"s_train[s_train_i] = train_vec[loadx];\n"
"}\n"
"else\n"
"{\n"
"s_query[s_query_i] = 0;\n"
"s_train[s_train_i] = 0;\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"result += reduce_block_match(s_query, s_train, lidx, lidy);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"}\n"
"#endif\n"
"result = DIST_RES(result);\n"
"const int trainIdx = mad24(BLOCK_SIZE, t, lidx);\n"
"if (queryIdx < query_rows && trainIdx < train_rows)\n"
"{\n"
"if (result < myBestDistance1)\n"
"{\n"
"myBestDistance2 = myBestDistance1;\n"
"myBestTrainIdx2 = myBestTrainIdx1;\n"
"myBestDistance1 = result;\n"
"myBestTrainIdx1 = trainIdx;\n"
"}\n"
"else if (result < myBestDistance2)\n"
"{\n"
"myBestDistance2 = result;\n"
"myBestTrainIdx2 = trainIdx;\n"
"}\n"
"}\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"__local float *s_distance = (__local float *)sharebuffer;\n"
"__local int *s_trainIdx = (__local int *)(sharebuffer + BLOCK_SIZE_ODD * BLOCK_SIZE);\n"
"s_distance += lidy * BLOCK_SIZE_ODD;\n"
"s_trainIdx += lidy * BLOCK_SIZE_ODD;\n"
"s_distance[lidx] = myBestDistance1;\n"
"s_trainIdx[lidx] = myBestTrainIdx1;\n"
"float bestDistance1 = MAX_FLOAT;\n"
"float bestDistance2 = MAX_FLOAT;\n"
"int bestTrainIdx1 = -1;\n"
"int bestTrainIdx2 = -1;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if (lidx == 0)\n"
"{\n"
"for (int i = 0 ; i < BLOCK_SIZE ; i++)\n"
"{\n"
"float val = s_distance[i];\n"
"if (val < bestDistance1)\n"
"{\n"
"bestDistance2 = bestDistance1;\n"
"bestTrainIdx2 = bestTrainIdx1;\n"
"bestDistance1 = val;\n"
"bestTrainIdx1 = s_trainIdx[i];\n"
"}\n"
"else if (val < bestDistance2)\n"
"{\n"
"bestDistance2 = val;\n"
"bestTrainIdx2 = s_trainIdx[i];\n"
"}\n"
"}\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"s_distance[lidx] = myBestDistance2;\n"
"s_trainIdx[lidx] = myBestTrainIdx2;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if (lidx == 0)\n"
"{\n"
"for (int i = 0 ; i < BLOCK_SIZE ; i++)\n"
"{\n"
"float val = s_distance[i];\n"
"if (val < bestDistance2)\n"
"{\n"
"bestDistance2 = val;\n"
"bestTrainIdx2 = s_trainIdx[i];\n"
"}\n"
"}\n"
"}\n"
"myBestDistance1 = bestDistance1;\n"
"myBestDistance2 = bestDistance2;\n"
"myBestTrainIdx1 = bestTrainIdx1;\n"
"myBestTrainIdx2 = bestTrainIdx2;\n"
"if (queryIdx < query_rows && lidx == 0)\n"
"{\n"
"bestTrainIdx[queryIdx] = (int2)(myBestTrainIdx1, myBestTrainIdx2);\n"
"bestDistance[queryIdx] = (float2)(myBestDistance1, myBestDistance2);\n"
"}\n"
"}\n"
, "35c3a1e231d446e4088561e3604fb94f", NULL};
struct cv::ocl::internal::ProgramEntry fast_oclsrc={moduleName, "fast",
"inline int cornerScore(__global const uchar* img, int step)\n"
"{\n"
"int k, tofs, v = img[0], a0 = 0, b0;\n"
"int d[16];\n"
"#define LOAD2(idx, ofs) \\\n"
"tofs = ofs; d[idx] = (short)(v - img[tofs]); d[idx+8] = (short)(v - img[-tofs])\n"
"LOAD2(0, 3);\n"
"LOAD2(1, -step+3);\n"
"LOAD2(2, -step*2+2);\n"
"LOAD2(3, -step*3+1);\n"
"LOAD2(4, -step*3);\n"
"LOAD2(5, -step*3-1);\n"
"LOAD2(6, -step*2-2);\n"
"LOAD2(7, -step-3);\n"
"#pragma unroll\n"
"for( k = 0; k < 16; k += 2 )\n"
"{\n"
"int a = min((int)d[(k+1)&15], (int)d[(k+2)&15]);\n"
"a = min(a, (int)d[(k+3)&15]);\n"
"a = min(a, (int)d[(k+4)&15]);\n"
"a = min(a, (int)d[(k+5)&15]);\n"
"a = min(a, (int)d[(k+6)&15]);\n"
"a = min(a, (int)d[(k+7)&15]);\n"
"a = min(a, (int)d[(k+8)&15]);\n"
"a0 = max(a0, min(a, (int)d[k&15]));\n"
"a0 = max(a0, min(a, (int)d[(k+9)&15]));\n"
"}\n"
"b0 = -a0;\n"
"#pragma unroll\n"
"for( k = 0; k < 16; k += 2 )\n"
"{\n"
"int b = max((int)d[(k+1)&15], (int)d[(k+2)&15]);\n"
"b = max(b, (int)d[(k+3)&15]);\n"
"b = max(b, (int)d[(k+4)&15]);\n"
"b = max(b, (int)d[(k+5)&15]);\n"
"b = max(b, (int)d[(k+6)&15]);\n"
"b = max(b, (int)d[(k+7)&15]);\n"
"b = max(b, (int)d[(k+8)&15]);\n"
"b0 = min(b0, max(b, (int)d[k]));\n"
"b0 = min(b0, max(b, (int)d[(k+9)&15]));\n"
"}\n"
"return -b0-1;\n"
"}\n"
"__kernel\n"
"void FAST_findKeypoints(\n"
"__global const uchar * _img, int step, int img_offset,\n"
"int img_rows, int img_cols,\n"
"volatile __global int* kp_loc,\n"
"int max_keypoints, int threshold )\n"
"{\n"
"int j = get_global_id(0) + 3;\n"
"int i = get_global_id(1) + 3;\n"
"if (i < img_rows - 3 && j < img_cols - 3)\n"
"{\n"
"__global const uchar* img = _img + mad24(i, step, j + img_offset);\n"
"int v = img[0], t0 = v - threshold, t1 = v + threshold;\n"
"int k, tofs, v0, v1;\n"
"int m0 = 0, m1 = 0;\n"
"#define UPDATE_MASK(idx, ofs) \\\n"
"tofs = ofs; v0 = img[tofs]; v1 = img[-tofs]; \\\n"
"m0 |= ((v0 < t0) << idx) | ((v1 < t0) << (8 + idx)); \\\n"
"m1 |= ((v0 > t1) << idx) | ((v1 > t1) << (8 + idx))\n"
"UPDATE_MASK(0, 3);\n"
"if( (m0 | m1) == 0 )\n"
"return;\n"
"UPDATE_MASK(2, -step*2+2);\n"
"UPDATE_MASK(4, -step*3);\n"
"UPDATE_MASK(6, -step*2-2);\n"
"#define EVEN_MASK (1+4+16+64)\n"
"if( ((m0 | (m0 >> 8)) & EVEN_MASK) != EVEN_MASK &&\n"
"((m1 | (m1 >> 8)) & EVEN_MASK) != EVEN_MASK )\n"
"return;\n"
"UPDATE_MASK(1, -step+3);\n"
"UPDATE_MASK(3, -step*3+1);\n"
"UPDATE_MASK(5, -step*3-1);\n"
"UPDATE_MASK(7, -step-3);\n"
"if( ((m0 | (m0 >> 8)) & 255) != 255 &&\n"
"((m1 | (m1 >> 8)) & 255) != 255 )\n"
"return;\n"
"m0 |= m0 << 16;\n"
"m1 |= m1 << 16;\n"
"#define CHECK0(i) ((m0 & (511 << i)) == (511 << i))\n"
"#define CHECK1(i) ((m1 & (511 << i)) == (511 << i))\n"
"if( CHECK0(0) + CHECK0(1) + CHECK0(2) + CHECK0(3) +\n"
"CHECK0(4) + CHECK0(5) + CHECK0(6) + CHECK0(7) +\n"
"CHECK0(8) + CHECK0(9) + CHECK0(10) + CHECK0(11) +\n"
"CHECK0(12) + CHECK0(13) + CHECK0(14) + CHECK0(15) +\n"
"CHECK1(0) + CHECK1(1) + CHECK1(2) + CHECK1(3) +\n"
"CHECK1(4) + CHECK1(5) + CHECK1(6) + CHECK1(7) +\n"
"CHECK1(8) + CHECK1(9) + CHECK1(10) + CHECK1(11) +\n"
"CHECK1(12) + CHECK1(13) + CHECK1(14) + CHECK1(15) == 0 )\n"
"return;\n"
"{\n"
"int idx = atomic_inc(kp_loc);\n"
"if( idx < max_keypoints )\n"
"{\n"
"kp_loc[1 + 2*idx] = j;\n"
"kp_loc[2 + 2*idx] = i;\n"
"}\n"
"}\n"
"}\n"
"}\n"
"__kernel\n"
"void FAST_nonmaxSupression(\n"
"__global const int* kp_in, volatile __global int* kp_out,\n"
"__global const uchar * _img, int step, int img_offset,\n"
"int rows, int cols, int counter, int max_keypoints)\n"
"{\n"
"const int idx = get_global_id(0);\n"
"if (idx < counter)\n"
"{\n"
"int x = kp_in[1 + 2*idx];\n"
"int y = kp_in[2 + 2*idx];\n"
"__global const uchar* img = _img + mad24(y, step, x + img_offset);\n"
"int s = cornerScore(img, step);\n"
"if( (x < 4 || s > cornerScore(img-1, step)) +\n"
"(y < 4 || s > cornerScore(img-step, step)) != 2 )\n"
"return;\n"
"if( (x >= cols - 4 || s > cornerScore(img+1, step)) +\n"
"(y >= rows - 4 || s > cornerScore(img+step, step)) +\n"
"(x < 4 || y < 4 || s > cornerScore(img-step-1, step)) +\n"
"(x >= cols - 4 || y < 4 || s > cornerScore(img-step+1, step)) +\n"
"(x < 4 || y >= rows - 4 || s > cornerScore(img+step-1, step)) +\n"
"(x >= cols - 4 || y >= rows - 4 || s > cornerScore(img+step+1, step)) == 6)\n"
"{\n"
"int new_idx = atomic_inc(kp_out);\n"
"if( new_idx < max_keypoints )\n"
"{\n"
"kp_out[1 + 3*new_idx] = x;\n"
"kp_out[2 + 3*new_idx] = y;\n"
"kp_out[3 + 3*new_idx] = s;\n"
"}\n"
"}\n"
"}\n"
"}\n"
, "f5e6f463f21a7ed77bd4d2c753478305", NULL};
struct cv::ocl::internal::ProgramEntry orb_oclsrc={moduleName, "orb",
"#define LAYERINFO_SIZE 1\n"
"#define LAYERINFO_OFS 0\n"
"#define KEYPOINT_SIZE 3\n"
"#define ORIENTED_KEYPOINT_SIZE 4\n"
"#define KEYPOINT_X 0\n"
"#define KEYPOINT_Y 1\n"
"#define KEYPOINT_Z 2\n"
"#define KEYPOINT_ANGLE 3\n"
"#ifdef ORB_RESPONSES\n"
"__kernel void\n"
"ORB_HarrisResponses(__global const uchar* imgbuf, int imgstep, int imgoffset0,\n"
"__global const int* layerinfo, __global const int* keypoints,\n"
"__global float* responses, int nkeypoints )\n"
"{\n"
"int idx = get_global_id(0);\n"
"if( idx < nkeypoints )\n"
"{\n"
"__global const int* kpt = keypoints + idx*KEYPOINT_SIZE;\n"
"__global const int* layer = layerinfo + kpt[KEYPOINT_Z]*LAYERINFO_SIZE;\n"
"__global const uchar* img = imgbuf + imgoffset0 + layer[LAYERINFO_OFS] +\n"
"(kpt[KEYPOINT_Y] - blockSize/2)*imgstep + (kpt[KEYPOINT_X] - blockSize/2);\n"
"int i, j;\n"
"int a = 0, b = 0, c = 0;\n"
"for( i = 0; i < blockSize; i++, img += imgstep-blockSize )\n"
"{\n"
"for( j = 0; j < blockSize; j++, img++ )\n"
"{\n"
"int Ix = (img[1] - img[-1])*2 + img[-imgstep+1] - img[-imgstep-1] + img[imgstep+1] - img[imgstep-1];\n"
"int Iy = (img[imgstep] - img[-imgstep])*2 + img[imgstep-1] - img[-imgstep-1] + img[imgstep+1] - img[-imgstep+1];\n"
"a += Ix*Ix;\n"
"b += Iy*Iy;\n"
"c += Ix*Iy;\n"
"}\n"
"}\n"
"responses[idx] = ((float)a * b - (float)c * c - HARRIS_K * (float)(a + b) * (a + b))*scale_sq_sq;\n"
"}\n"
"}\n"
"#endif\n"
"#ifdef ORB_ANGLES\n"
"#define _DBL_EPSILON 2.2204460492503131e-16f\n"
"#define atan2_p1 (0.9997878412794807f*57.29577951308232f)\n"
"#define atan2_p3 (-0.3258083974640975f*57.29577951308232f)\n"
"#define atan2_p5 (0.1555786518463281f*57.29577951308232f)\n"
"#define atan2_p7 (-0.04432655554792128f*57.29577951308232f)\n"
"inline float fastAtan2( float y, float x )\n"
"{\n"
"float ax = fabs(x), ay = fabs(y);\n"
"float a, c, c2;\n"
"if( ax >= ay )\n"
"{\n"
"c = ay/(ax + _DBL_EPSILON);\n"
"c2 = c*c;\n"
"a = (((atan2_p7*c2 + atan2_p5)*c2 + atan2_p3)*c2 + atan2_p1)*c;\n"
"}\n"
"else\n"
"{\n"
"c = ax/(ay + _DBL_EPSILON);\n"
"c2 = c*c;\n"
"a = 90.f - (((atan2_p7*c2 + atan2_p5)*c2 + atan2_p3)*c2 + atan2_p1)*c;\n"
"}\n"
"if( x < 0 )\n"
"a = 180.f - a;\n"
"if( y < 0 )\n"
"a = 360.f - a;\n"
"return a;\n"
"}\n"
"__kernel void\n"
"ORB_ICAngle(__global const uchar* imgbuf, int imgstep, int imgoffset0,\n"
"__global const int* layerinfo, __global const int* keypoints,\n"
"__global float* responses, const __global int* u_max,\n"
"int nkeypoints, int half_k )\n"
"{\n"
"int idx = get_global_id(0);\n"
"if( idx < nkeypoints )\n"
"{\n"
"__global const int* kpt = keypoints + idx*KEYPOINT_SIZE;\n"
"__global const int* layer = layerinfo + kpt[KEYPOINT_Z]*LAYERINFO_SIZE;\n"
"__global const uchar* center = imgbuf + imgoffset0 + layer[LAYERINFO_OFS] +\n"
"kpt[KEYPOINT_Y]*imgstep + kpt[KEYPOINT_X];\n"
"int u, v, m_01 = 0, m_10 = 0;\n"
"for( u = -half_k; u <= half_k; u++ )\n"
"m_10 += u * center[u];\n"
"for( v = 1; v <= half_k; v++ )\n"
"{\n"
"int v_sum = 0;\n"
"int d = u_max[v];\n"
"for( u = -d; u <= d; u++ )\n"
"{\n"
"int val_plus = center[u + v*imgstep], val_minus = center[u - v*imgstep];\n"
"v_sum += (val_plus - val_minus);\n"
"m_10 += u * (val_plus + val_minus);\n"
"}\n"
"m_01 += v * v_sum;\n"
"}\n"
"responses[idx] = fastAtan2((float)m_01, (float)m_10);\n"
"}\n"
"}\n"
"#endif\n"
"#ifdef ORB_DESCRIPTORS\n"
"__kernel void\n"
"ORB_computeDescriptor(__global const uchar* imgbuf, int imgstep, int imgoffset0,\n"
"__global const int* layerinfo, __global const int* keypoints,\n"
"__global uchar* _desc, const __global int* pattern,\n"
"int nkeypoints, int dsize )\n"
"{\n"
"int idx = get_global_id(0);\n"
"if( idx < nkeypoints )\n"
"{\n"
"int i;\n"
"__global const int* kpt = keypoints + idx*ORIENTED_KEYPOINT_SIZE;\n"
"__global const int* layer = layerinfo + kpt[KEYPOINT_Z]*LAYERINFO_SIZE;\n"
"__global const uchar* center = imgbuf + imgoffset0 + layer[LAYERINFO_OFS] +\n"
"kpt[KEYPOINT_Y]*imgstep + kpt[KEYPOINT_X];\n"
"float angle = as_float(kpt[KEYPOINT_ANGLE]);\n"
"angle *= 0.01745329251994329547f;\n"
"float cosa;\n"
"float sina = sincos(angle, &cosa);\n"
"__global uchar* desc = _desc + idx*dsize;\n"
"#define GET_VALUE(idx) \\\n"
"center[mad24(convert_int_rte(pattern[(idx)*2] * sina + pattern[(idx)*2+1] * cosa), imgstep, \\\n"
"convert_int_rte(pattern[(idx)*2] * cosa - pattern[(idx)*2+1] * sina))]\n"
"for( i = 0; i < dsize; i++ )\n"
"{\n"
"int val;\n"
"#if WTA_K == 2\n"
"int t0, t1;\n"
"t0 = GET_VALUE(0); t1 = GET_VALUE(1);\n"
"val = t0 < t1;\n"
"t0 = GET_VALUE(2); t1 = GET_VALUE(3);\n"
"val |= (t0 < t1) << 1;\n"
"t0 = GET_VALUE(4); t1 = GET_VALUE(5);\n"
"val |= (t0 < t1) << 2;\n"
"t0 = GET_VALUE(6); t1 = GET_VALUE(7);\n"
"val |= (t0 < t1) << 3;\n"
"t0 = GET_VALUE(8); t1 = GET_VALUE(9);\n"
"val |= (t0 < t1) << 4;\n"
"t0 = GET_VALUE(10); t1 = GET_VALUE(11);\n"
"val |= (t0 < t1) << 5;\n"
"t0 = GET_VALUE(12); t1 = GET_VALUE(13);\n"
"val |= (t0 < t1) << 6;\n"
"t0 = GET_VALUE(14); t1 = GET_VALUE(15);\n"
"val |= (t0 < t1) << 7;\n"
"pattern += 16*2;\n"
"#elif WTA_K == 3\n"
"int t0, t1, t2;\n"
"t0 = GET_VALUE(0); t1 = GET_VALUE(1); t2 = GET_VALUE(2);\n"
"val = t2 > t1 ? (t2 > t0 ? 2 : 0) : (t1 > t0);\n"
"t0 = GET_VALUE(3); t1 = GET_VALUE(4); t2 = GET_VALUE(5);\n"
"val |= (t2 > t1 ? (t2 > t0 ? 2 : 0) : (t1 > t0)) << 2;\n"
"t0 = GET_VALUE(6); t1 = GET_VALUE(7); t2 = GET_VALUE(8);\n"
"val |= (t2 > t1 ? (t2 > t0 ? 2 : 0) : (t1 > t0)) << 4;\n"
"t0 = GET_VALUE(9); t1 = GET_VALUE(10); t2 = GET_VALUE(11);\n"
"val |= (t2 > t1 ? (t2 > t0 ? 2 : 0) : (t1 > t0)) << 6;\n"
"pattern += 12*2;\n"
"#elif WTA_K == 4\n"
"int t0, t1, t2, t3, k;\n"
"int a, b;\n"
"t0 = GET_VALUE(0); t1 = GET_VALUE(1);\n"
"t2 = GET_VALUE(2); t3 = GET_VALUE(3);\n"
"a = 0, b = 2;\n"
"if( t1 > t0 ) t0 = t1, a = 1;\n"
"if( t3 > t2 ) t2 = t3, b = 3;\n"
"k = t0 > t2 ? a : b;\n"
"val = k;\n"
"t0 = GET_VALUE(4); t1 = GET_VALUE(5);\n"
"t2 = GET_VALUE(6); t3 = GET_VALUE(7);\n"
"a = 0, b = 2;\n"
"if( t1 > t0 ) t0 = t1, a = 1;\n"
"if( t3 > t2 ) t2 = t3, b = 3;\n"
"k = t0 > t2 ? a : b;\n"
"val |= k << 2;\n"
"t0 = GET_VALUE(8); t1 = GET_VALUE(9);\n"
"t2 = GET_VALUE(10); t3 = GET_VALUE(11);\n"
"a = 0, b = 2;\n"
"if( t1 > t0 ) t0 = t1, a = 1;\n"
"if( t3 > t2 ) t2 = t3, b = 3;\n"
"k = t0 > t2 ? a : b;\n"
"val |= k << 4;\n"
"t0 = GET_VALUE(12); t1 = GET_VALUE(13);\n"
"t2 = GET_VALUE(14); t3 = GET_VALUE(15);\n"
"a = 0, b = 2;\n"
"if( t1 > t0 ) t0 = t1, a = 1;\n"
"if( t3 > t2 ) t2 = t3, b = 3;\n"
"k = t0 > t2 ? a : b;\n"
"val |= k << 6;\n"
"pattern += 16*2;\n"
"#else\n"
"#error \"unknown/undefined WTA_K value; should be 2, 3 or 4\"\n"
"#endif\n"
"desc[i] = (uchar)val;\n"
"}\n"
"}\n"
"}\n"
"#endif\n"
, "a7c2cfaeda19907b637211b1cc91d253", NULL};

}}}
#endif