block_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/block_reduce_raking.cuh"
#include "specializations/block_reduce_raking_commutative_only.cuh"
#include "specializations/block_reduce_warp_reductions.cuh"
#include "../util_ptx.cuh"
#include "../util_type.cuh"
#include "../thread/thread_operators.cuh"
#include "../util_namespace.cuh"

CUB_NS_PREFIX

namespace cub {



/******************************************************************************
 * Algorithmic variants
 ******************************************************************************/

enum BlockReduceAlgorithm
{

    BLOCK_REDUCE_RAKING_COMMUTATIVE_ONLY,


    BLOCK_REDUCE_RAKING,


    BLOCK_REDUCE_WARP_REDUCTIONS,
};


/******************************************************************************
 * Block reduce
 ******************************************************************************/

template <
    typename                T,
    int                     BLOCK_DIM_X,
    BlockReduceAlgorithm    ALGORITHM       = BLOCK_REDUCE_WARP_REDUCTIONS,
    int                     BLOCK_DIM_Y     = 1,
    int                     BLOCK_DIM_Z     = 1,
    int                     PTX_ARCH        = CUB_PTX_ARCH>
class BlockReduce
{
private:

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

    enum
    {
        BLOCK_THREADS = BLOCK_DIM_X * BLOCK_DIM_Y * BLOCK_DIM_Z,
    };

    typedef BlockReduceWarpReductions<T, BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z, PTX_ARCH>           WarpReductions;
    typedef BlockReduceRakingCommutativeOnly<T, BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z, PTX_ARCH>    RakingCommutativeOnly;
    typedef BlockReduceRaking<T, BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z, PTX_ARCH>                   Raking;

    typedef typename If<(ALGORITHM == BLOCK_REDUCE_WARP_REDUCTIONS),
        WarpReductions,
        typename If<(ALGORITHM == BLOCK_REDUCE_RAKING_COMMUTATIVE_ONLY),
            RakingCommutativeOnly,
            Raking>::Type>::Type InternalBlockReduce;     // BlockReduceRaking

    typedef typename InternalBlockReduce::TempStorage _TempStorage;


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

    __device__ __forceinline__ _TempStorage& PrivateStorage()
    {
        __shared__ _TempStorage private_storage;
        return private_storage;
    }


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

    _TempStorage &temp_storage;

    unsigned int linear_tid;


public:

    struct TempStorage : Uninitialized<_TempStorage> {};


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

    __device__ __forceinline__ BlockReduce()
    :
        temp_storage(PrivateStorage()),
        linear_tid(RowMajorTid(BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z))
    {}


    __device__ __forceinline__ BlockReduce(
        TempStorage &temp_storage)             
    :
        temp_storage(temp_storage.Alias()),
        linear_tid(RowMajorTid(BLOCK_DIM_X, BLOCK_DIM_Y, BLOCK_DIM_Z))
    {}


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


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


    template <
        int ITEMS_PER_THREAD,
        typename ReductionOp>
    __device__ __forceinline__ T Reduce(
        T               (&inputs)[ITEMS_PER_THREAD],    
        ReductionOp     reduction_op)                   
    {
        // Reduce partials
        T partial = internal::ThreadReduce(inputs, reduction_op);
        return Reduce(partial, reduction_op);
    }


    template <typename ReductionOp>
    __device__ __forceinline__ T Reduce(
        T                   input,                  
        ReductionOp         reduction_op,           
        int                 num_valid)              
    {
        // Determine if we scan skip bounds checking
        if (num_valid >= BLOCK_THREADS)
        {
            return InternalBlockReduce(temp_storage).template Reduce<true>(input, num_valid, reduction_op);
        }
        else
        {
            return InternalBlockReduce(temp_storage).template Reduce<false>(input, num_valid, reduction_op);
        }
    }


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


    __device__ __forceinline__ T Sum(
        T   input)                      
    {
        return InternalBlockReduce(temp_storage).template Sum<true>(input, BLOCK_THREADS);
    }

    template <int ITEMS_PER_THREAD>
    __device__ __forceinline__ T Sum(
        T   (&inputs)[ITEMS_PER_THREAD])    
    {
        // Reduce partials
        T partial = internal::ThreadReduce(inputs, cub::Sum());
        return Sum(partial);
    }


    __device__ __forceinline__ T Sum(
        T   input,                  
        int num_valid)              
    {
        // Determine if we scan skip bounds checking
        if (num_valid >= BLOCK_THREADS)
        {
            return InternalBlockReduce(temp_storage).template Sum<true>(input, num_valid);
        }
        else
        {
            return InternalBlockReduce(temp_storage).template Sum<false>(input, num_valid);
        }
    }


};

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