device_histogram.cuh 53.1 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
/******************************************************************************
 * Copyright (c) 2011, Duane Merrill.  All rights reserved.
 * Copyright (c) 2011-2018, NVIDIA CORPORATION.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * 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.
 *     * Neither the name of the NVIDIA CORPORATION nor the
 *       names of its contributors may 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 NVIDIA CORPORATION 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.
 *
 ******************************************************************************/

/**
 * \file
 * cub::DeviceHistogram provides device-wide parallel operations for constructing histogram(s) from a sequence of samples data residing within device-accessible memory.
 */

#pragma once

#include <stdio.h>
#include <iterator>
#include <limits>

#include "dispatch/dispatch_histogram.cuh"
#include "../util_namespace.cuh"

/// Optional outer namespace(s)
CUB_NS_PREFIX

/// CUB namespace
namespace cub {


/**
 * \brief DeviceHistogram provides device-wide parallel operations for constructing histogram(s) from a sequence of samples data residing within device-accessible memory. ![](histogram_logo.png)
 * \ingroup SingleModule
 *
 * \par Overview
 * A <a href="http://en.wikipedia.org/wiki/Histogram"><em>histogram</em></a>
 * counts the number of observations that fall into each of the disjoint categories (known as <em>bins</em>).
 *
 * \par Usage Considerations
 * \cdp_class{DeviceHistogram}
 *
 */
struct DeviceHistogram
{
    /******************************************************************//**
     * \name Evenly-segmented bin ranges
     *********************************************************************/
    //@{

    /**
     * \brief Computes an intensity histogram from a sequence of data samples using equal-width bins.
     *
     * \par
     * - The number of histogram bins is (\p num_levels - 1)
     * - All bins comprise the same width of sample values: (\p upper_level - \p lower_level) / (\p num_levels - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of a six-bin histogram
     * from a sequence of float samples
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples and
     * // output histogram
     * int      num_samples;    // e.g., 10
     * float*   d_samples;      // e.g., [2.2, 6.0, 7.1, 2.9, 3.5, 0.3, 2.9, 2.0, 6.1, 999.5]
     * int*     d_histogram;    // e.g., [ -, -, -, -, -, -, -, -]
     * int      num_levels;     // e.g., 7       (seven level boundaries for six bins)
     * float    lower_level;    // e.g., 0.0     (lower sample value boundary of lowest bin)
     * float    upper_level;    // e.g., 12.0    (upper sample value boundary of upper bin)
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::HistogramEven(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level, num_samples);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::HistogramEven(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level, num_samples);
     *
     * // d_histogram   <-- [1, 0, 5, 0, 3, 0, 0, 0];
     *
     * \endcode
     *
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t HistogramEven(
        void*               d_temp_storage,                             ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                        ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                                  ///< [in] The pointer to the input sequence of data samples.
        CounterT*           d_histogram,                                ///< [out] The pointer to the histogram counter output array of length <tt>num_levels</tt> - 1.
        int                 num_levels,                                 ///< [in] The number of boundaries (levels) for delineating histogram samples.  Implies that the number of bins is <tt>num_levels</tt> - 1.
        LevelT              lower_level,                                ///< [in] The lower sample value bound (inclusive) for the lowest histogram bin.
        LevelT              upper_level,                                ///< [in] The upper sample value bound (exclusive) for the highest histogram bin.
        OffsetT             num_samples,                                ///< [in] The number of input samples (i.e., the length of \p d_samples)
        cudaStream_t        stream                  = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous       = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;

        CounterT*           d_histogram1[1]     = {d_histogram};
        int                 num_levels1[1]      = {num_levels};
        LevelT              lower_level1[1]     = {lower_level};
        LevelT              upper_level1[1]     = {upper_level};

        return MultiHistogramEven<1, 1>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram1,
            num_levels1,
            lower_level1,
            upper_level1,
            num_samples,
            1,
            sizeof(SampleT) * num_samples,
            stream,
            debug_synchronous);
    }


    /**
     * \brief Computes an intensity histogram from a sequence of data samples using equal-width bins.
     *
     * \par
     * - A two-dimensional <em>region of interest</em> within \p d_samples can be specified
     *   using the \p num_row_samples, num_rows, and \p row_stride_bytes parameters.
     * - The row stride must be a whole multiple of the sample data type
     *   size, i.e., <tt>(row_stride_bytes % sizeof(SampleT)) == 0</tt>.
     * - The number of histogram bins is (\p num_levels - 1)
     * - All bins comprise the same width of sample values: (\p upper_level - \p lower_level) / (\p num_levels - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of a six-bin histogram
     * from a 2x5 region of interest within a flattened 2x7 array of float samples.
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples and
     * // output histogram
     * int      num_row_samples;    // e.g., 5
     * int      num_rows;           // e.g., 2;
     * size_t   row_stride_bytes;   // e.g., 7 * sizeof(float)
     * float*   d_samples;          // e.g., [2.2, 6.0, 7.1, 2.9, 3.5,   -, -,
     *                              //        0.3, 2.9, 2.0, 6.1, 999.5, -, -]
     * int*     d_histogram;        // e.g., [ -, -, -, -, -, -, -, -]
     * int      num_levels;         // e.g., 7       (seven level boundaries for six bins)
     * float    lower_level;        // e.g., 0.0     (lower sample value boundary of lowest bin)
     * float    upper_level;        // e.g., 12.0    (upper sample value boundary of upper bin)
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage  = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::HistogramEven(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level,
     *     num_row_samples, num_rows, row_stride_bytes);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::HistogramEven(d_temp_storage, temp_storage_bytes, d_samples, d_histogram,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level,
     *     num_row_samples, num_rows, row_stride_bytes);
     *
     * // d_histogram   <-- [1, 0, 5, 0, 3, 0, 0, 0];
     *
     * \endcode
     *
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t HistogramEven(
        void*               d_temp_storage,                             ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                        ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                                  ///< [in] The pointer to the input sequence of data samples.
        CounterT*           d_histogram,                                ///< [out] The pointer to the histogram counter output array of length <tt>num_levels</tt> - 1.
        int                 num_levels,                                 ///< [in] The number of boundaries (levels) for delineating histogram samples.  Implies that the number of bins is <tt>num_levels</tt> - 1.
        LevelT              lower_level,                                ///< [in] The lower sample value bound (inclusive) for the lowest histogram bin.
        LevelT              upper_level,                                ///< [in] The upper sample value bound (exclusive) for the highest histogram bin.
        OffsetT             num_row_samples,                            ///< [in] The number of data samples per row in the region of interest
        OffsetT             num_rows,                                   ///< [in] The number of rows in the region of interest
        size_t              row_stride_bytes,                           ///< [in] The number of bytes between starts of consecutive rows in the region of interest
        cudaStream_t        stream                  = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous       = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        CounterT*           d_histogram1[1]     = {d_histogram};
        int                 num_levels1[1]      = {num_levels};
        LevelT              lower_level1[1]     = {lower_level};
        LevelT              upper_level1[1]     = {upper_level};

        return MultiHistogramEven<1, 1>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram1,
            num_levels1,
            lower_level1,
            upper_level1,
            num_row_samples,
            num_rows,
            row_stride_bytes,
            stream,
            debug_synchronous);
    }

    /**
     * \brief Computes per-channel intensity histograms from a sequence of multi-channel "pixel" data samples using equal-width bins.
     *
     * \par
     * - The input is a sequence of <em>pixel</em> structures, where each pixel comprises
     *   a record of \p NUM_CHANNELS consecutive data samples (e.g., an <em>RGBA</em> pixel).
     * - Of the \p NUM_CHANNELS specified, the function will only compute histograms
     *   for the first \p NUM_ACTIVE_CHANNELS (e.g., only <em>RGB</em> histograms from <em>RGBA</em>
     *   pixel samples).
     * - The number of histogram bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
     * - For channel<sub><em>i</em></sub>, the range of values for all histogram bins
     *   have the same width: (<tt>upper_level[i]</tt> - <tt>lower_level[i]</tt>) / (<tt> num_levels[i]</tt> - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of three 256-bin <em>RGB</em> histograms
     * from a quad-channel sequence of <em>RGBA</em> pixels (8 bits per channel per pixel)
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples
     * // and output histograms
     * int              num_pixels;         // e.g., 5
     * unsigned char*   d_samples;          // e.g., [(2, 6, 7, 5), (3, 0, 2, 1), (7, 0, 6, 2),
     *                                      //        (0, 6, 7, 5), (3, 0, 2, 6)]
     * int*             d_histogram[3];     // e.g., three device pointers to three device buffers,
     *                                      //       each allocated with 256 integer counters
     * int              num_levels[3];      // e.g., {257, 257, 257};
     * unsigned int     lower_level[3];     // e.g., {0, 0, 0};
     * unsigned int     upper_level[3];     // e.g., {256, 256, 256};
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::MultiHistogramEven<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level, num_pixels);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::MultiHistogramEven<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level, num_pixels);
     *
     * // d_histogram   <-- [ [1, 0, 1, 2, 0, 0, 0, 1, 0, 0, 0, ..., 0],
     * //                     [0, 3, 0, 0, 0, 0, 2, 0, 0, 0, 0, ..., 0],
     * //                     [0, 0, 2, 0, 0, 0, 1, 2, 0, 0, 0, ..., 0] ]
     *
     * \endcode
     *
     * \tparam NUM_CHANNELS             Number of channels interleaved in the input data (may be greater than the number of channels being actively histogrammed)
     * \tparam NUM_ACTIVE_CHANNELS      <b>[inferred]</b> Number of channels actively being histogrammed
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        int                 NUM_CHANNELS,
        int                 NUM_ACTIVE_CHANNELS,
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t MultiHistogramEven(
        void*               d_temp_storage,                             ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                        ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                                  ///< [in] The pointer to the multi-channel input sequence of data samples. The samples from different channels are assumed to be interleaved (e.g., an array of 32-bit pixels where each pixel consists of four <em>RGBA</em> 8-bit samples).
        CounterT*           d_histogram[NUM_ACTIVE_CHANNELS],           ///< [out] The pointers to the histogram counter output arrays, one for each active channel.  For channel<sub><em>i</em></sub>, the allocation length of <tt>d_histogram[i]</tt> should be <tt>num_levels[i]</tt> - 1.
        int                 num_levels[NUM_ACTIVE_CHANNELS],            ///< [in] The number of boundaries (levels) for delineating histogram samples in each active channel.  Implies that the number of bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
        LevelT              lower_level[NUM_ACTIVE_CHANNELS],           ///< [in] The lower sample value bound (inclusive) for the lowest histogram bin in each active channel.
        LevelT              upper_level[NUM_ACTIVE_CHANNELS],           ///< [in] The upper sample value bound (exclusive) for the highest histogram bin in each active channel.
        OffsetT             num_pixels,                                 ///< [in] The number of multi-channel pixels (i.e., the length of \p d_samples / NUM_CHANNELS)
        cudaStream_t        stream                  = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous       = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;

        return MultiHistogramEven<NUM_CHANNELS, NUM_ACTIVE_CHANNELS>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram,
            num_levels,
            lower_level,
            upper_level,
            num_pixels,
            1,
            sizeof(SampleT) * NUM_CHANNELS * num_pixels,
            stream,
            debug_synchronous);
    }


    /**
     * \brief Computes per-channel intensity histograms from a sequence of multi-channel "pixel" data samples using equal-width bins.
     *
     * \par
     * - The input is a sequence of <em>pixel</em> structures, where each pixel comprises
     *   a record of \p NUM_CHANNELS consecutive data samples (e.g., an <em>RGBA</em> pixel).
     * - Of the \p NUM_CHANNELS specified, the function will only compute histograms
     *   for the first \p NUM_ACTIVE_CHANNELS (e.g., only <em>RGB</em> histograms from <em>RGBA</em>
     *   pixel samples).
     * - A two-dimensional <em>region of interest</em> within \p d_samples can be specified
     *   using the \p num_row_samples, num_rows, and \p row_stride_bytes parameters.
     * - The row stride must be a whole multiple of the sample data type
     *   size, i.e., <tt>(row_stride_bytes % sizeof(SampleT)) == 0</tt>.
     * - The number of histogram bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
     * - For channel<sub><em>i</em></sub>, the range of values for all histogram bins
     *   have the same width: (<tt>upper_level[i]</tt> - <tt>lower_level[i]</tt>) / (<tt> num_levels[i]</tt> - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of three 256-bin <em>RGB</em> histograms from a 2x3 region of
     * interest of within a flattened 2x4 array of quad-channel <em>RGBA</em> pixels (8 bits per channel per pixel).
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples
     * // and output histograms
     * int              num_row_pixels;     // e.g., 3
     * int              num_rows;           // e.g., 2
     * size_t           row_stride_bytes;   // e.g., 4 * sizeof(unsigned char) * NUM_CHANNELS
     * unsigned char*   d_samples;          // e.g., [(2, 6, 7, 5), (3, 0, 2, 1), (7, 0, 6, 2), (-, -, -, -),
     *                                      //        (0, 6, 7, 5), (3, 0, 2, 6), (1, 1, 1, 1), (-, -, -, -)]
     * int*             d_histogram[3];     // e.g., three device pointers to three device buffers,
     *                                      //       each allocated with 256 integer counters
     * int              num_levels[3];      // e.g., {257, 257, 257};
     * unsigned int     lower_level[3];     // e.g., {0, 0, 0};
     * unsigned int     upper_level[3];     // e.g., {256, 256, 256};
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::MultiHistogramEven<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level,
     *     num_row_pixels, num_rows, row_stride_bytes);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::MultiHistogramEven<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, lower_level, upper_level,
     *     num_row_pixels, num_rows, row_stride_bytes);
     *
     * // d_histogram   <-- [ [1, 1, 1, 2, 0, 0, 0, 1, 0, 0, 0, ..., 0],
     * //                     [0, 4, 0, 0, 0, 0, 2, 0, 0, 0, 0, ..., 0],
     * //                     [0, 1, 2, 0, 0, 0, 1, 2, 0, 0, 0, ..., 0] ]
     *
     * \endcode
     *
     * \tparam NUM_CHANNELS             Number of channels interleaved in the input data (may be greater than the number of channels being actively histogrammed)
     * \tparam NUM_ACTIVE_CHANNELS      <b>[inferred]</b> Number of channels actively being histogrammed
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        int                 NUM_CHANNELS,
        int                 NUM_ACTIVE_CHANNELS,
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t MultiHistogramEven(
        void*               d_temp_storage,                             ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                        ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                                  ///< [in] The pointer to the multi-channel input sequence of data samples. The samples from different channels are assumed to be interleaved (e.g., an array of 32-bit pixels where each pixel consists of four <em>RGBA</em> 8-bit samples).
        CounterT*           d_histogram[NUM_ACTIVE_CHANNELS],           ///< [out] The pointers to the histogram counter output arrays, one for each active channel.  For channel<sub><em>i</em></sub>, the allocation length of <tt>d_histogram[i]</tt> should be <tt>num_levels[i]</tt> - 1.
        int                 num_levels[NUM_ACTIVE_CHANNELS],            ///< [in] The number of boundaries (levels) for delineating histogram samples in each active channel.  Implies that the number of bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
        LevelT              lower_level[NUM_ACTIVE_CHANNELS],           ///< [in] The lower sample value bound (inclusive) for the lowest histogram bin in each active channel.
        LevelT              upper_level[NUM_ACTIVE_CHANNELS],           ///< [in] The upper sample value bound (exclusive) for the highest histogram bin in each active channel.
        OffsetT             num_row_pixels,                             ///< [in] The number of multi-channel pixels per row in the region of interest
        OffsetT             num_rows,                                   ///< [in] The number of rows in the region of interest
        size_t              row_stride_bytes,                           ///< [in] The number of bytes between starts of consecutive rows in the region of interest
        cudaStream_t        stream                  = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous       = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;
        Int2Type<sizeof(SampleT) == 1> is_byte_sample;

        if ((sizeof(OffsetT) > sizeof(int)) &&
            ((unsigned long long) (num_rows * row_stride_bytes) < (unsigned long long) std::numeric_limits<int>::max()))
        {
            // Down-convert OffsetT data type


            return DipatchHistogram<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, SampleIteratorT, CounterT, LevelT, int>::DispatchEven(
                d_temp_storage, temp_storage_bytes, d_samples, d_histogram, num_levels, lower_level, upper_level,
                (int) num_row_pixels, (int) num_rows, (int) (row_stride_bytes / sizeof(SampleT)),
                stream, debug_synchronous, is_byte_sample);
        }

        return DipatchHistogram<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, SampleIteratorT, CounterT, LevelT, OffsetT>::DispatchEven(
            d_temp_storage, temp_storage_bytes, d_samples, d_histogram, num_levels, lower_level, upper_level,
            num_row_pixels, num_rows, (OffsetT) (row_stride_bytes / sizeof(SampleT)),
            stream, debug_synchronous, is_byte_sample);
    }


    //@}  end member group
    /******************************************************************//**
     * \name Custom bin ranges
     *********************************************************************/
    //@{

    /**
     * \brief Computes an intensity histogram from a sequence of data samples using the specified bin boundary levels.
     *
     * \par
     * - The number of histogram bins is (\p num_levels - 1)
     * - The value range for bin<sub><em>i</em></sub> is [<tt>level[i]</tt>, <tt>level[i+1]</tt>)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of an six-bin histogram
     * from a sequence of float samples
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples and
     * // output histogram
     * int      num_samples;    // e.g., 10
     * float*   d_samples;      // e.g., [2.2, 6.0, 7.1, 2.9, 3.5, 0.3, 2.9, 2.0, 6.1, 999.5]
     * int*     d_histogram;    // e.g., [ -, -, -, -, -, -, -, -]
     * int      num_levels      // e.g., 7 (seven level boundaries for six bins)
     * float*   d_levels;       // e.g., [0.0, 2.0, 4.0, 6.0, 8.0, 12.0, 16.0]
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::HistogramRange(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_samples);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::HistogramRange(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_samples);
     *
     * // d_histogram   <-- [1, 0, 5, 0, 3, 0, 0, 0];
     *
     * \endcode
     *
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t HistogramRange(
        void*               d_temp_storage,                         ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                    ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                              ///< [in] The pointer to the input sequence of data samples.
        CounterT*           d_histogram,                            ///< [out] The pointer to the histogram counter output array of length <tt>num_levels</tt> - 1.
        int                 num_levels,                             ///< [in] The number of boundaries (levels) for delineating histogram samples.  Implies that the number of bins is <tt>num_levels</tt> - 1.
        LevelT*             d_levels,                               ///< [in] The pointer to the array of boundaries (levels).  Bin ranges are defined by consecutive boundary pairings: lower sample value boundaries are inclusive and upper sample value boundaries are exclusive.
        OffsetT             num_samples,                            ///< [in] The number of data samples per row in the region of interest
        cudaStream_t        stream              = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous   = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;

        CounterT*           d_histogram1[1] = {d_histogram};
        int                 num_levels1[1]  = {num_levels};
        LevelT*             d_levels1[1]    = {d_levels};

        return MultiHistogramRange<1, 1>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram1,
            num_levels1,
            d_levels1,
            num_samples,
            1,
            sizeof(SampleT) * num_samples,
            stream,
            debug_synchronous);
    }


    /**
     * \brief Computes an intensity histogram from a sequence of data samples using the specified bin boundary levels.
     *
     * \par
     * - A two-dimensional <em>region of interest</em> within \p d_samples can be specified
     *   using the \p num_row_samples, num_rows, and \p row_stride_bytes parameters.
     * - The row stride must be a whole multiple of the sample data type
     *   size, i.e., <tt>(row_stride_bytes % sizeof(SampleT)) == 0</tt>.
     * - The number of histogram bins is (\p num_levels - 1)
     * - The value range for bin<sub><em>i</em></sub> is [<tt>level[i]</tt>, <tt>level[i+1]</tt>)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of a six-bin histogram
     * from a 2x5 region of interest within a flattened 2x7 array of float samples.
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples and
     * // output histogram
     * int      num_row_samples;    // e.g., 5
     * int      num_rows;           // e.g., 2;
     * int      row_stride_bytes;   // e.g., 7 * sizeof(float)
     * float*   d_samples;          // e.g., [2.2, 6.0, 7.1, 2.9, 3.5,   -, -,
     *                              //        0.3, 2.9, 2.0, 6.1, 999.5, -, -]
     * int*     d_histogram;        // e.g., [ , , , , , , , ]
     * int      num_levels          // e.g., 7 (seven level boundaries for six bins)
     * float    *d_levels;          // e.g., [0.0, 2.0, 4.0, 6.0, 8.0, 12.0, 16.0]
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::HistogramRange(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels,
     *     num_row_samples, num_rows, row_stride_bytes);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::HistogramRange(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels,
     *     num_row_samples, num_rows, row_stride_bytes);
     *
     * // d_histogram   <-- [1, 0, 5, 0, 3, 0, 0, 0];
     *
     * \endcode
     *
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t HistogramRange(
        void*               d_temp_storage,                         ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                    ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                              ///< [in] The pointer to the input sequence of data samples.
        CounterT*           d_histogram,                            ///< [out] The pointer to the histogram counter output array of length <tt>num_levels</tt> - 1.
        int                 num_levels,                             ///< [in] The number of boundaries (levels) for delineating histogram samples.  Implies that the number of bins is <tt>num_levels</tt> - 1.
        LevelT*             d_levels,                               ///< [in] The pointer to the array of boundaries (levels).  Bin ranges are defined by consecutive boundary pairings: lower sample value boundaries are inclusive and upper sample value boundaries are exclusive.
        OffsetT             num_row_samples,                        ///< [in] The number of data samples per row in the region of interest
        OffsetT             num_rows,                               ///< [in] The number of rows in the region of interest
        size_t              row_stride_bytes,                       ///< [in] The number of bytes between starts of consecutive rows in the region of interest
        cudaStream_t        stream              = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous   = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        CounterT*           d_histogram1[1]     = {d_histogram};
        int                 num_levels1[1]      = {num_levels};
        LevelT*             d_levels1[1]        = {d_levels};

        return MultiHistogramRange<1, 1>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram1,
            num_levels1,
            d_levels1,
            num_row_samples,
            num_rows,
            row_stride_bytes,
            stream,
            debug_synchronous);
    }

    /**
     * \brief Computes per-channel intensity histograms from a sequence of multi-channel "pixel" data samples using the specified bin boundary levels.
     *
     * \par
     * - The input is a sequence of <em>pixel</em> structures, where each pixel comprises
     *   a record of \p NUM_CHANNELS consecutive data samples (e.g., an <em>RGBA</em> pixel).
     * - Of the \p NUM_CHANNELS specified, the function will only compute histograms
     *   for the first \p NUM_ACTIVE_CHANNELS (e.g., <em>RGB</em> histograms from <em>RGBA</em>
     *   pixel samples).
     * - The number of histogram bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
     * - For channel<sub><em>i</em></sub>, the range of values for all histogram bins
     *   have the same width: (<tt>upper_level[i]</tt> - <tt>lower_level[i]</tt>) / (<tt> num_levels[i]</tt> - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of three 4-bin <em>RGB</em> histograms
     * from a quad-channel sequence of <em>RGBA</em> pixels (8 bits per channel per pixel)
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples
     * // and output histograms
     * int            num_pixels;       // e.g., 5
     * unsigned char  *d_samples;       // e.g., [(2, 6, 7, 5),(3, 0, 2, 1),(7, 0, 6, 2),
     *                                  //        (0, 6, 7, 5),(3, 0, 2, 6)]
     * unsigned int   *d_histogram[3];  // e.g., [[ -, -, -, -],[ -, -, -, -],[ -, -, -, -]];
     * int            num_levels[3];    // e.g., {5, 5, 5};
     * unsigned int   *d_levels[3];     // e.g., [ [0, 2, 4, 6, 8],
     *                                  //         [0, 2, 4, 6, 8],
     *                                  //         [0, 2, 4, 6, 8] ];
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::MultiHistogramRange<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_pixels);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::MultiHistogramRange<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_pixels);
     *
     * // d_histogram   <-- [ [1, 3, 0, 1],
     * //                     [3, 0, 0, 2],
     * //                     [0, 2, 0, 3] ]
     *
     * \endcode
     *
     * \tparam NUM_CHANNELS             Number of channels interleaved in the input data (may be greater than the number of channels being actively histogrammed)
     * \tparam NUM_ACTIVE_CHANNELS      <b>[inferred]</b> Number of channels actively being histogrammed
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        int                 NUM_CHANNELS,
        int                 NUM_ACTIVE_CHANNELS,
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t MultiHistogramRange(
        void*               d_temp_storage,                         ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                    ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                              ///< [in] The pointer to the multi-channel input sequence of data samples. The samples from different channels are assumed to be interleaved (e.g., an array of 32-bit pixels where each pixel consists of four <em>RGBA</em> 8-bit samples).
        CounterT*           d_histogram[NUM_ACTIVE_CHANNELS],       ///< [out] The pointers to the histogram counter output arrays, one for each active channel.  For channel<sub><em>i</em></sub>, the allocation length of <tt>d_histogram[i]</tt> should be <tt>num_levels[i]</tt> - 1.
        int                 num_levels[NUM_ACTIVE_CHANNELS],        ///< [in] The number of boundaries (levels) for delineating histogram samples in each active channel.  Implies that the number of bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
        LevelT*             d_levels[NUM_ACTIVE_CHANNELS],          ///< [in] The pointers to the arrays of boundaries (levels), one for each active channel.  Bin ranges are defined by consecutive boundary pairings: lower sample value boundaries are inclusive and upper sample value boundaries are exclusive.
        OffsetT             num_pixels,                             ///< [in] The number of multi-channel pixels (i.e., the length of \p d_samples / NUM_CHANNELS)
        cudaStream_t        stream              = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous   = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;

        return MultiHistogramRange<NUM_CHANNELS, NUM_ACTIVE_CHANNELS>(
            d_temp_storage,
            temp_storage_bytes,
            d_samples,
            d_histogram,
            num_levels,
            d_levels,
            num_pixels,
            1,
            sizeof(SampleT) * NUM_CHANNELS * num_pixels,
            stream,
            debug_synchronous);
    }


    /**
     * \brief Computes per-channel intensity histograms from a sequence of multi-channel "pixel" data samples using the specified bin boundary levels.
     *
     * \par
     * - The input is a sequence of <em>pixel</em> structures, where each pixel comprises
     *   a record of \p NUM_CHANNELS consecutive data samples (e.g., an <em>RGBA</em> pixel).
     * - Of the \p NUM_CHANNELS specified, the function will only compute histograms
     *   for the first \p NUM_ACTIVE_CHANNELS (e.g., <em>RGB</em> histograms from <em>RGBA</em>
     *   pixel samples).
     * - A two-dimensional <em>region of interest</em> within \p d_samples can be specified
     *   using the \p num_row_samples, num_rows, and \p row_stride_bytes parameters.
     * - The row stride must be a whole multiple of the sample data type
     *   size, i.e., <tt>(row_stride_bytes % sizeof(SampleT)) == 0</tt>.
     * - The number of histogram bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
     * - For channel<sub><em>i</em></sub>, the range of values for all histogram bins
     *   have the same width: (<tt>upper_level[i]</tt> - <tt>lower_level[i]</tt>) / (<tt> num_levels[i]</tt> - 1)
     * - \devicestorage
     *
     * \par Snippet
     * The code snippet below illustrates the computation of three 4-bin <em>RGB</em> histograms from a 2x3 region of
     * interest of within a flattened 2x4 array of quad-channel <em>RGBA</em> pixels (8 bits per channel per pixel).
     *
     * \par
     * \code
     * #include <cub/cub.cuh>   // or equivalently <cub/device/device_histogram.cuh>
     *
     * // Declare, allocate, and initialize device-accessible pointers for input samples
     * // and output histograms
     * int              num_row_pixels;     // e.g., 3
     * int              num_rows;           // e.g., 2
     * size_t           row_stride_bytes;   // e.g., 4 * sizeof(unsigned char) * NUM_CHANNELS
     * unsigned char*   d_samples;          // e.g., [(2, 6, 7, 5),(3, 0, 2, 1),(1, 1, 1, 1),(-, -, -, -),
     *                                      //        (7, 0, 6, 2),(0, 6, 7, 5),(3, 0, 2, 6),(-, -, -, -)]
     * int*             d_histogram[3];     // e.g., [[ -, -, -, -],[ -, -, -, -],[ -, -, -, -]];
     * int              num_levels[3];      // e.g., {5, 5, 5};
     * unsigned int*    d_levels[3];        // e.g., [ [0, 2, 4, 6, 8],
     *                                      //         [0, 2, 4, 6, 8],
     *                                      //         [0, 2, 4, 6, 8] ];
     * ...
     *
     * // Determine temporary device storage requirements
     * void*    d_temp_storage = NULL;
     * size_t   temp_storage_bytes = 0;
     * cub::DeviceHistogram::MultiHistogramRange<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_row_pixels, num_rows, row_stride_bytes);
     *
     * // Allocate temporary storage
     * cudaMalloc(&d_temp_storage, temp_storage_bytes);
     *
     * // Compute histograms
     * cub::DeviceHistogram::MultiHistogramRange<4, 3>(d_temp_storage, temp_storage_bytes,
     *     d_samples, d_histogram, num_levels, d_levels, num_row_pixels, num_rows, row_stride_bytes);
     *
     * // d_histogram   <-- [ [2, 3, 0, 1],
     * //                     [3, 0, 0, 2],
     * //                     [1, 2, 0, 3] ]
     *
     * \endcode
     *
     * \tparam NUM_CHANNELS             Number of channels interleaved in the input data (may be greater than the number of channels being actively histogrammed)
     * \tparam NUM_ACTIVE_CHANNELS      <b>[inferred]</b> Number of channels actively being histogrammed
     * \tparam SampleIteratorT          <b>[inferred]</b> Random-access input iterator type for reading input samples. \iterator
     * \tparam CounterT                 <b>[inferred]</b> Integer type for histogram bin counters
     * \tparam LevelT                   <b>[inferred]</b> Type for specifying boundaries (levels)
     * \tparam OffsetT                  <b>[inferred]</b> Signed integer type for sequence offsets, list lengths, pointer differences, etc.  \offset_size1
     */
    template <
        int                 NUM_CHANNELS,
        int                 NUM_ACTIVE_CHANNELS,
        typename            SampleIteratorT,
        typename            CounterT,
        typename            LevelT,
        typename            OffsetT>
    CUB_RUNTIME_FUNCTION
    static cudaError_t MultiHistogramRange(
        void*               d_temp_storage,                         ///< [in] %Device-accessible allocation of temporary storage.  When NULL, the required allocation size is written to \p temp_storage_bytes and no work is done.
        size_t&             temp_storage_bytes,                    ///< [in,out] Reference to size in bytes of \p d_temp_storage allocation
        SampleIteratorT     d_samples,                              ///< [in] The pointer to the multi-channel input sequence of data samples. The samples from different channels are assumed to be interleaved (e.g., an array of 32-bit pixels where each pixel consists of four <em>RGBA</em> 8-bit samples).
        CounterT*           d_histogram[NUM_ACTIVE_CHANNELS],       ///< [out] The pointers to the histogram counter output arrays, one for each active channel.  For channel<sub><em>i</em></sub>, the allocation length of <tt>d_histogram[i]</tt> should be <tt>num_levels[i]</tt> - 1.
        int                 num_levels[NUM_ACTIVE_CHANNELS],        ///< [in] The number of boundaries (levels) for delineating histogram samples in each active channel.  Implies that the number of bins for channel<sub><em>i</em></sub> is <tt>num_levels[i]</tt> - 1.
        LevelT*             d_levels[NUM_ACTIVE_CHANNELS],          ///< [in] The pointers to the arrays of boundaries (levels), one for each active channel.  Bin ranges are defined by consecutive boundary pairings: lower sample value boundaries are inclusive and upper sample value boundaries are exclusive.
        OffsetT             num_row_pixels,                         ///< [in] The number of multi-channel pixels per row in the region of interest
        OffsetT             num_rows,                               ///< [in] The number of rows in the region of interest
        size_t              row_stride_bytes,                       ///< [in] The number of bytes between starts of consecutive rows in the region of interest
        cudaStream_t        stream              = 0,                ///< [in] <b>[optional]</b> CUDA stream to launch kernels within.  Default is stream<sub>0</sub>.
        bool                debug_synchronous   = false)            ///< [in] <b>[optional]</b> Whether or not to synchronize the stream after every kernel launch to check for errors.  May cause significant slowdown.  Default is \p false.
    {
        /// The sample value type of the input iterator
        typedef typename std::iterator_traits<SampleIteratorT>::value_type SampleT;
        Int2Type<sizeof(SampleT) == 1> is_byte_sample;

        if ((sizeof(OffsetT) > sizeof(int)) &&
            ((unsigned long long) (num_rows * row_stride_bytes) < (unsigned long long) std::numeric_limits<int>::max()))
        {
            // Down-convert OffsetT data type
            return DipatchHistogram<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, SampleIteratorT, CounterT, LevelT, int>::DispatchRange(
                d_temp_storage, temp_storage_bytes, d_samples, d_histogram, num_levels, d_levels,
                (int) num_row_pixels, (int) num_rows, (int) (row_stride_bytes / sizeof(SampleT)),
                stream, debug_synchronous, is_byte_sample);
        }

        return DipatchHistogram<NUM_CHANNELS, NUM_ACTIVE_CHANNELS, SampleIteratorT, CounterT, LevelT, OffsetT>::DispatchRange(
            d_temp_storage, temp_storage_bytes, d_samples, d_histogram, num_levels, d_levels,
            num_row_pixels, num_rows, (OffsetT) (row_stride_bytes / sizeof(SampleT)),
            stream, debug_synchronous, is_byte_sample);
    }



    //@}  end member group
};

}               // CUB namespace
CUB_NS_POSTFIX  // Optional outer namespace(s)