view release on metacpan or  search on metacpan

xgboost/cub/cub/warp/specializations/warp_scan_smem.cuh  view on Meta::CPAN

 * Copyright (c) 2011-2016, 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::WarpScanSmem provides smem-based variants of parallel prefix scan of items partitioned across a CUDA thread warp.
 */

#pragma once

#include "../../thread/thread_operators.cuh"
#include "../../thread/thread_load.cuh"
#include "../../thread/thread_store.cuh"
#include "../../util_type.cuh"
#include "../../util_namespace.cuh"

/// Optional outer namespace(s)
CUB_NS_PREFIX

/// CUB namespace
namespace cub {

/**
 * \brief WarpScanSmem provides smem-based variants of parallel prefix scan of items partitioned across a CUDA thread warp.
 */
template <
    typename    T,                      ///< Data type being scanned
    int         LOGICAL_WARP_THREADS,   ///< Number of threads per logical warp
    int         PTX_ARCH>               ///< The PTX compute capability for which to to specialize this collective
struct WarpScanSmem
{
    /******************************************************************************
     * Constants and type definitions
     ******************************************************************************/

    enum
    {
        /// Whether the logical warp size and the PTX warp size coincide
        IS_ARCH_WARP = (LOGICAL_WARP_THREADS == CUB_WARP_THREADS(PTX_ARCH)),

        /// Whether the logical warp size is a power-of-two
        IS_POW_OF_TWO = PowerOfTwo<LOGICAL_WARP_THREADS>::VALUE,

        /// The number of warp scan steps
        STEPS = Log2<LOGICAL_WARP_THREADS>::VALUE,

        /// The number of threads in half a warp
        HALF_WARP_THREADS = 1 << (STEPS - 1),

        /// The number of shared memory elements per warp
        WARP_SMEM_ELEMENTS =  LOGICAL_WARP_THREADS + HALF_WARP_THREADS,
    };

    /// Storage cell type (workaround for SM1x compiler bugs with custom-ops like Max() on signed chars)
    typedef typename If<((Equals<T, char>::VALUE || Equals<T, signed char>::VALUE) && (PTX_ARCH < 200)), int, T>::Type CellT;

    /// Shared memory storage layout type (1.5 warps-worth of elements for each warp)
    typedef CellT _TempStorage[WARP_SMEM_ELEMENTS];

    // Alias wrapper allowing storage to be unioned
    struct TempStorage : Uninitialized<_TempStorage> {};


    /******************************************************************************
     * Thread fields
     ******************************************************************************/

    _TempStorage    &temp_storage;
    unsigned int    lane_id;
    unsigned int    member_mask;


    /******************************************************************************
     * Construction
     ******************************************************************************/

    /// Constructor
    __device__ __forceinline__ WarpScanSmem(
        TempStorage     &temp_storage)
    :
        temp_storage(temp_storage.Alias()),
        lane_id(IS_ARCH_WARP ?
            LaneId() :
            LaneId() % LOGICAL_WARP_THREADS),
        member_mask(!IS_POW_OF_TWO ?
            (0xffffffff >> (32 - LOGICAL_WARP_THREADS)) :                                       // non-power-of-two subwarps cannot be tiled
            (0xffffffff >> (32 - LOGICAL_WARP_THREADS)) << (LaneId() / LOGICAL_WARP_THREADS))
    {}


    /******************************************************************************
     * Utility methods
     ******************************************************************************/

    /// Basic inclusive scan iteration (template unrolled, inductive-case specialization)
    template <
        bool        HAS_IDENTITY,
        int         STEP,