reduce.cl 19.8 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
////////////////////////////////////////////////////////////////////////////////////////
//
//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
//  By downloading, copying, installing or using the software you agree to this license.
//  If you do not agree to this license, do not download, install,
//  copy or use the software.
//
//
//                           License Agreement
//                For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
//    Shengen Yan,yanshengen@gmail.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
//   * Redistribution's of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
//
//   * Redistribution's 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.
//
//   * The name of the copyright holders may not be used to endorse or promote products
//     derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors 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 the Intel Corporation or 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.
//

#ifdef DOUBLE_SUPPORT
#ifdef cl_amd_fp64
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#elif defined (cl_khr_fp64)
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#endif
#endif

#if defined OP_NORM_INF_MASK

#ifdef DEPTH_0
#define MIN_VAL 0
#define MAX_VAL 255
#elif defined DEPTH_1
#define MIN_VAL -128
#define MAX_VAL 127
#elif defined DEPTH_2
#define MIN_VAL 0
#define MAX_VAL 65535
#elif defined DEPTH_3
#define MIN_VAL -32768
#define MAX_VAL 32767
#elif defined DEPTH_4
#define MIN_VAL INT_MIN
#define MAX_VAL INT_MAX
#elif defined DEPTH_5
#define MIN_VAL (-FLT_MAX)
#define MAX_VAL FLT_MAX
#elif defined DEPTH_6
#define MIN_VAL (-DBL_MAX)
#define MAX_VAL DBL_MAX
#endif

#define dstT srcT
#define dstT1 srcT1

#endif // min/max stuff

#define noconvert

#ifndef kercn
#define kercn 1
#endif

#ifdef HAVE_MASK_CONT
#define MASK_INDEX int mask_index = id + mask_offset;
#else
#define MASK_INDEX int mask_index = mad24(id / cols, mask_step, mask_offset + (id % cols))
#endif

#if cn != 3
#define loadpix(addr) *(__global const srcT *)(addr)
#define storepix(val, addr)  *(__global dstT *)(addr) = val
#if kercn == 1
#define srcTSIZE (int)sizeof(srcT)
#else
#define srcTSIZE (int)sizeof(srcT1)
#endif
#define dstTSIZE (int)sizeof(dstT)
#else
#define loadpix(addr) vload3(0, (__global const srcT1 *)(addr))
#define storepix(val, addr) vstore3(val, 0, (__global dstT1 *)(addr))
#define srcTSIZE ((int)sizeof(srcT1)*3)
#define dstTSIZE ((int)sizeof(dstT1)*3)
#endif

#if ddepth <= 4
#define SUM_ABS(a) convertFromU(abs(a))
#define SUM_ABS2(a, b) convertFromU(abs_diff(a, b))
#else
#define SUM_ABS(a) fabs(a)
#define SUM_ABS2(a, b) fabs(a - b)
#endif

#ifdef HAVE_MASK
#ifdef HAVE_SRC2
#define EXTRA_PARAMS , __global const uchar * mask, int mask_step, int mask_offset, __global const uchar * src2ptr, int src2_step, int src2_offset
#else
#define EXTRA_PARAMS , __global const uchar * mask, int mask_step, int mask_offset
#endif
#else
#ifdef HAVE_SRC2
#define EXTRA_PARAMS , __global const uchar * src2ptr, int src2_step, int src2_offset
#else
#define EXTRA_PARAMS
#endif
#endif

// accumulative reduction stuff
#if defined OP_SUM || defined OP_SUM_ABS || defined OP_SUM_SQR || defined OP_DOT

#ifdef OP_DOT
#if ddepth <= 4
#define FUNC(a, b, c) a = mad24(b, c, a)
#else
#define FUNC(a, b, c) a = mad(b, c, a)
#endif

#elif defined OP_SUM
#define FUNC(a, b) a += b

#elif defined OP_SUM_ABS
#define FUNC(a, b) a += SUM_ABS(b)

#elif defined OP_SUM_SQR
#if ddepth <= 4
#define FUNC(a, b) a = mad24(b, b, a)
#else
#define FUNC(a, b) a = mad(b, b, a)
#endif
#endif

#ifdef OP_CALC2
#define DECLARE_LOCAL_MEM \
    __local dstT localmem[WGS2_ALIGNED], localmem2[WGS2_ALIGNED]
#define DEFINE_ACCUMULATOR \
    dstT accumulator = (dstT)(0), accumulator2 = (dstT)(0)
#else
#define DECLARE_LOCAL_MEM \
    __local dstT localmem[WGS2_ALIGNED]
#define DEFINE_ACCUMULATOR \
    dstT accumulator = (dstT)(0)
#endif

#ifdef HAVE_SRC2
#ifdef OP_CALC2
#define PROCESS_ELEMS \
    dstT temp = convertToDT(loadpix(srcptr + src_index)); \
    dstT temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator2, temp2); \
    FUNC(accumulator, temp)
#else
#define PROCESS_ELEMS \
    dstT temp = convertToDT(loadpix(srcptr + src_index)); \
    dstT temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp)
#endif
#else
#define PROCESS_ELEMS \
    dstT temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp)
#endif

#ifdef HAVE_MASK
#define REDUCE_GLOBAL \
    MASK_INDEX; \
    if (mask[mask_index]) \
    { \
        PROCESS_ELEMS; \
    }
#elif defined OP_DOT

#ifdef HAVE_SRC2_CONT
#define SRC2_INDEX int src2_index = mad24(id, srcTSIZE, src2_offset);
#else
#define SRC2_INDEX int src2_index = mad24(id / cols, src2_step, mad24(id % cols, srcTSIZE, src2_offset))
#endif

#if kercn == 1
#define REDUCE_GLOBAL \
    SRC2_INDEX; \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)), temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    FUNC(accumulator, temp, temp2)
#elif kercn == 2
#define REDUCE_GLOBAL \
    SRC2_INDEX; \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)), temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    FUNC(accumulator, temp.s0, temp2.s0); \
    FUNC(accumulator, temp.s1, temp2.s1)
#elif kercn == 4
#define REDUCE_GLOBAL \
    SRC2_INDEX; \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)), temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    FUNC(accumulator, temp.s0, temp2.s0); \
    FUNC(accumulator, temp.s1, temp2.s1); \
    FUNC(accumulator, temp.s2, temp2.s2); \
    FUNC(accumulator, temp.s3, temp2.s3)
#elif kercn == 8
#define REDUCE_GLOBAL \
    SRC2_INDEX; \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)), temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    FUNC(accumulator, temp.s0, temp2.s0); \
    FUNC(accumulator, temp.s1, temp2.s1); \
    FUNC(accumulator, temp.s2, temp2.s2); \
    FUNC(accumulator, temp.s3, temp2.s3); \
    FUNC(accumulator, temp.s4, temp2.s4); \
    FUNC(accumulator, temp.s5, temp2.s5); \
    FUNC(accumulator, temp.s6, temp2.s6); \
    FUNC(accumulator, temp.s7, temp2.s7)
#elif kercn == 16
#define REDUCE_GLOBAL \
    SRC2_INDEX; \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)), temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    FUNC(accumulator, temp.s0, temp2.s0); \
    FUNC(accumulator, temp.s1, temp2.s1); \
    FUNC(accumulator, temp.s2, temp2.s2); \
    FUNC(accumulator, temp.s3, temp2.s3); \
    FUNC(accumulator, temp.s4, temp2.s4); \
    FUNC(accumulator, temp.s5, temp2.s5); \
    FUNC(accumulator, temp.s6, temp2.s6); \
    FUNC(accumulator, temp.s7, temp2.s7); \
    FUNC(accumulator, temp.s8, temp2.s8); \
    FUNC(accumulator, temp.s9, temp2.s9); \
    FUNC(accumulator, temp.sA, temp2.sA); \
    FUNC(accumulator, temp.sB, temp2.sB); \
    FUNC(accumulator, temp.sC, temp2.sC); \
    FUNC(accumulator, temp.sD, temp2.sD); \
    FUNC(accumulator, temp.sE, temp2.sE); \
    FUNC(accumulator, temp.sF, temp2.sF)
#endif

#else // sum or norm with 2 args
#ifdef HAVE_SRC2
#ifdef OP_CALC2 // norm relative
#if kercn == 1
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator, temp); \
    FUNC(accumulator2, temp2)
#elif kercn == 2
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator2, temp2.s0); \
    FUNC(accumulator2, temp2.s1)
#elif kercn == 4
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator2, temp2.s0); \
    FUNC(accumulator2, temp2.s1); \
    FUNC(accumulator2, temp2.s2); \
    FUNC(accumulator2, temp2.s3)
#elif kercn == 8
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7); \
    FUNC(accumulator2, temp2.s0); \
    FUNC(accumulator2, temp2.s1); \
    FUNC(accumulator2, temp2.s2); \
    FUNC(accumulator2, temp2.s3); \
    FUNC(accumulator2, temp2.s4); \
    FUNC(accumulator2, temp2.s5); \
    FUNC(accumulator2, temp2.s6); \
    FUNC(accumulator2, temp2.s7)
#elif kercn == 16
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    temp2 = SUM_ABS(temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7); \
    FUNC(accumulator, temp.s8); \
    FUNC(accumulator, temp.s9); \
    FUNC(accumulator, temp.sA); \
    FUNC(accumulator, temp.sB); \
    FUNC(accumulator, temp.sC); \
    FUNC(accumulator, temp.sD); \
    FUNC(accumulator, temp.sE); \
    FUNC(accumulator, temp.sF); \
    FUNC(accumulator2, temp2.s0); \
    FUNC(accumulator2, temp2.s1); \
    FUNC(accumulator2, temp2.s2); \
    FUNC(accumulator2, temp2.s3); \
    FUNC(accumulator2, temp2.s4); \
    FUNC(accumulator2, temp2.s5); \
    FUNC(accumulator2, temp2.s6); \
    FUNC(accumulator2, temp2.s7); \
    FUNC(accumulator2, temp2.s8); \
    FUNC(accumulator2, temp2.s9); \
    FUNC(accumulator2, temp2.sA); \
    FUNC(accumulator2, temp2.sB); \
    FUNC(accumulator2, temp2.sC); \
    FUNC(accumulator2, temp2.sD); \
    FUNC(accumulator2, temp2.sE); \
    FUNC(accumulator2, temp2.sF)
#endif
#else // norm with 2 args
#if kercn == 1
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp)
#elif kercn == 2
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1)
#elif kercn == 4
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3)
#elif kercn == 8
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7)
#elif kercn == 16
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    dstTK temp2 = convertToDT(loadpix(src2ptr + src2_index)); \
    temp = SUM_ABS2(temp, temp2); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7); \
    FUNC(accumulator, temp.s8); \
    FUNC(accumulator, temp.s9); \
    FUNC(accumulator, temp.sA); \
    FUNC(accumulator, temp.sB); \
    FUNC(accumulator, temp.sC); \
    FUNC(accumulator, temp.sD); \
    FUNC(accumulator, temp.sE); \
    FUNC(accumulator, temp.sF)
#endif
#endif

#else // sum
#if kercn == 1
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp)
#elif kercn == 2
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1)
#elif kercn == 4
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3)
#elif kercn == 8
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7)
#elif kercn == 16
#define REDUCE_GLOBAL \
    dstTK temp = convertToDT(loadpix(srcptr + src_index)); \
    FUNC(accumulator, temp.s0); \
    FUNC(accumulator, temp.s1); \
    FUNC(accumulator, temp.s2); \
    FUNC(accumulator, temp.s3); \
    FUNC(accumulator, temp.s4); \
    FUNC(accumulator, temp.s5); \
    FUNC(accumulator, temp.s6); \
    FUNC(accumulator, temp.s7); \
    FUNC(accumulator, temp.s8); \
    FUNC(accumulator, temp.s9); \
    FUNC(accumulator, temp.sA); \
    FUNC(accumulator, temp.sB); \
    FUNC(accumulator, temp.sC); \
    FUNC(accumulator, temp.sD); \
    FUNC(accumulator, temp.sE); \
    FUNC(accumulator, temp.sF)
#endif
#endif
#endif

#ifdef OP_CALC2
#define SET_LOCAL_1 \
    localmem[lid] = accumulator; \
    localmem2[lid] = accumulator2
#define REDUCE_LOCAL_1 \
    localmem[lid - WGS2_ALIGNED] += accumulator; \
    localmem2[lid - WGS2_ALIGNED] += accumulator2
#define REDUCE_LOCAL_2 \
    localmem[lid] += localmem[lid2]; \
    localmem2[lid] += localmem2[lid2]
#define CALC_RESULT \
    storepix(localmem[0], dstptr + dstTSIZE * gid); \
    storepix(localmem2[0], dstptr + mad24(groupnum, dstTSIZE, dstTSIZE * gid))
#else
#define SET_LOCAL_1 \
    localmem[lid] = accumulator
#define REDUCE_LOCAL_1 \
    localmem[lid - WGS2_ALIGNED] += accumulator
#define REDUCE_LOCAL_2 \
    localmem[lid] += localmem[lid2]
#define CALC_RESULT \
    storepix(localmem[0], dstptr + dstTSIZE * gid)
#endif

// countNonZero stuff
#elif defined OP_COUNT_NON_ZERO
#define dstT int
#define DECLARE_LOCAL_MEM \
    __local dstT localmem[WGS2_ALIGNED]
#define DEFINE_ACCUMULATOR \
    dstT accumulator = (dstT)(0); \
    srcT1 zero = (srcT1)(0), one = (srcT1)(1)
#if kercn == 1
#define REDUCE_GLOBAL \
    accumulator += loadpix(srcptr + src_index) == zero ? zero : one
#elif kercn == 2
#define REDUCE_GLOBAL \
    srcT value = loadpix(srcptr + src_index); \
    accumulator += value.s0 == zero ? zero : one; \
    accumulator += value.s1 == zero ? zero : one
#elif kercn == 4
#define REDUCE_GLOBAL \
    srcT value = loadpix(srcptr + src_index); \
    accumulator += value.s0 == zero ? zero : one; \
    accumulator += value.s1 == zero ? zero : one; \
    accumulator += value.s2 == zero ? zero : one; \
    accumulator += value.s3 == zero ? zero : one
#elif kercn == 8
#define REDUCE_GLOBAL \
    srcT value = loadpix(srcptr + src_index); \
    accumulator += value.s0 == zero ? zero : one; \
    accumulator += value.s1 == zero ? zero : one; \
    accumulator += value.s2 == zero ? zero : one; \
    accumulator += value.s3 == zero ? zero : one; \
    accumulator += value.s4 == zero ? zero : one; \
    accumulator += value.s5 == zero ? zero : one; \
    accumulator += value.s6 == zero ? zero : one; \
    accumulator += value.s7 == zero ? zero : one
#elif kercn == 16
#define REDUCE_GLOBAL \
    srcT value = loadpix(srcptr + src_index); \
    accumulator += value.s0 == zero ? zero : one; \
    accumulator += value.s1 == zero ? zero : one; \
    accumulator += value.s2 == zero ? zero : one; \
    accumulator += value.s3 == zero ? zero : one; \
    accumulator += value.s4 == zero ? zero : one; \
    accumulator += value.s5 == zero ? zero : one; \
    accumulator += value.s6 == zero ? zero : one; \
    accumulator += value.s7 == zero ? zero : one; \
    accumulator += value.s8 == zero ? zero : one; \
    accumulator += value.s9 == zero ? zero : one; \
    accumulator += value.sA == zero ? zero : one; \
    accumulator += value.sB == zero ? zero : one; \
    accumulator += value.sC == zero ? zero : one; \
    accumulator += value.sD == zero ? zero : one; \
    accumulator += value.sE == zero ? zero : one; \
    accumulator += value.sF == zero ? zero : one
#endif

#define SET_LOCAL_1 \
    localmem[lid] = accumulator
#define REDUCE_LOCAL_1 \
    localmem[lid - WGS2_ALIGNED] += accumulator
#define REDUCE_LOCAL_2 \
    localmem[lid] += localmem[lid2]
#define CALC_RESULT \
    storepix(localmem[0], dstptr + dstTSIZE * gid)

#else
#error "No operation"
#endif

#ifdef OP_DOT
#undef EXTRA_PARAMS
#define EXTRA_PARAMS , __global uchar * src2ptr, int src2_step, int src2_offset
#endif

__kernel void reduce(__global const uchar * srcptr, int src_step, int src_offset, int cols,
                     int total, int groupnum, __global uchar * dstptr EXTRA_PARAMS)
{
    int lid = get_local_id(0);
    int gid = get_group_id(0);
    int  id = get_global_id(0) * kercn;

    srcptr += src_offset;
#ifdef HAVE_SRC2
    src2ptr += src2_offset;
#endif

    DECLARE_LOCAL_MEM;
    DEFINE_ACCUMULATOR;

    for (int grain = groupnum * WGS * kercn; id < total; id += grain)
    {
#ifdef HAVE_SRC_CONT
        int src_index = id * srcTSIZE;
#else
        int src_index = mad24(id / cols, src_step, mul24(id % cols, srcTSIZE));
#endif
#ifdef HAVE_SRC2
#ifdef HAVE_SRC2_CONT
        int src2_index = id * srcTSIZE;
#else
        int src2_index = mad24(id / cols, src2_step, mul24(id % cols, srcTSIZE));
#endif
#endif
        REDUCE_GLOBAL;
    }

    if (lid < WGS2_ALIGNED)
    {
        SET_LOCAL_1;
    }
    barrier(CLK_LOCAL_MEM_FENCE);

    if (lid >= WGS2_ALIGNED && total >= WGS2_ALIGNED)
    {
        REDUCE_LOCAL_1;
    }
    barrier(CLK_LOCAL_MEM_FENCE);

    for (int lsize = WGS2_ALIGNED >> 1; lsize > 0; lsize >>= 1)
    {
        if (lid < lsize)
        {
           int lid2 = lsize + lid;
           REDUCE_LOCAL_2;
        }
        barrier(CLK_LOCAL_MEM_FENCE);
    }

    if (lid == 0)
    {
        CALC_RESULT;
    }
}