warp_reduce.cuh

Go to the documentation of this file.

/******************************************************************************
 * 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.
 *
 ******************************************************************************/

#pragma once

#include "specializations/warp_reduce_shfl.cuh"
#include "specializations/warp_reduce_smem.cuh"
#include "../thread/thread_operators.cuh"
#include "../util_arch.cuh"
#include "../util_type.cuh"
#include "../util_namespace.cuh"

CUB_NS_PREFIX

namespace cub {


template <
    typename    T,
    int         LOGICAL_WARP_THREADS    = CUB_PTX_WARP_THREADS,
    int         PTX_ARCH                = CUB_PTX_ARCH>
class WarpReduce
{
private:

    /******************************************************************************
     * Constants and type definitions
     ******************************************************************************/

    enum
    {
        IS_ARCH_WARP = (LOGICAL_WARP_THREADS == CUB_WARP_THREADS(PTX_ARCH)),

        IS_POW_OF_TWO = PowerOfTwo<LOGICAL_WARP_THREADS>::VALUE,
    };

public:

    #ifndef DOXYGEN_SHOULD_SKIP_THIS    // Do not document

    typedef typename If<(PTX_ARCH >= 300) && (IS_POW_OF_TWO),
        WarpReduceShfl<T, LOGICAL_WARP_THREADS, PTX_ARCH>,
        WarpReduceSmem<T, LOGICAL_WARP_THREADS, PTX_ARCH> >::Type InternalWarpReduce;

    #endif // DOXYGEN_SHOULD_SKIP_THIS


private:

    typedef typename InternalWarpReduce::TempStorage _TempStorage;


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

    _TempStorage &temp_storage;


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

public:

    struct TempStorage : Uninitialized<_TempStorage> {};


    /******************************************************************/


    __device__ __forceinline__ WarpReduce(
        TempStorage &temp_storage)             
    :
        temp_storage(temp_storage.Alias())
    {}


    /******************************************************************/


    __device__ __forceinline__ T Sum(
        T                   input)              
    {
        return InternalWarpReduce(temp_storage).template Reduce<true>(input, LOGICAL_WARP_THREADS, cub::Sum());
    }

    __device__ __forceinline__ T Sum(
        T                   input,              
        int                 valid_items)        
    {
        // Determine if we don't need bounds checking
        return InternalWarpReduce(temp_storage).template Reduce<false>(input, valid_items, cub::Sum());
    }


    template <
        typename            FlagT>
    __device__ __forceinline__ T HeadSegmentedSum(
        T                   input,              
        FlagT                head_flag)          
    {
        return HeadSegmentedReduce(input, head_flag, cub::Sum());
    }


    template <
        typename            FlagT>
    __device__ __forceinline__ T TailSegmentedSum(
        T                   input,              
        FlagT                tail_flag)          
    {
        return TailSegmentedReduce(input, tail_flag, cub::Sum());
    }



    /******************************************************************/

    template <typename ReductionOp>
    __device__ __forceinline__ T Reduce(
        T                   input,              
        ReductionOp         reduction_op)       
    {
        return InternalWarpReduce(temp_storage).template Reduce<true>(input, LOGICAL_WARP_THREADS, reduction_op);
    }

    template <typename ReductionOp>
    __device__ __forceinline__ T Reduce(
        T                   input,              
        ReductionOp         reduction_op,       
        int                 valid_items)        
    {
        return InternalWarpReduce(temp_storage).template Reduce<false>(input, valid_items, reduction_op);
    }


    template <
        typename            ReductionOp,
        typename            FlagT>
    __device__ __forceinline__ T HeadSegmentedReduce(
        T                   input,              
        FlagT               head_flag,          
        ReductionOp         reduction_op)       
    {
        return InternalWarpReduce(temp_storage).template SegmentedReduce<true>(input, head_flag, reduction_op);
    }


    template <
        typename            ReductionOp,
        typename            FlagT>
    __device__ __forceinline__ T TailSegmentedReduce(
        T                   input,              
        FlagT               tail_flag,          
        ReductionOp         reduction_op)       
    {
        return InternalWarpReduce(temp_storage).template SegmentedReduce<false>(input, tail_flag, reduction_op);
    }



};
       // end group WarpModule

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