work_queue.h

Go to the documentation of this file.

/*
 * nvbio
 * Copyright (c) 2011-2014, 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 <nvbio/basic/types.h>
#include <nvbio/basic/numbers.h>
#include <thrust/device_vector.h>

namespace nvbio {
namespace cuda {



// default tag
struct InplaceQueueTag {};

// a simple class to keep track of utilization stats
struct WorkQueueStats;
struct WorkQueueStatsView;

// return a device-side view of a stats object
inline WorkQueueStatsView view(WorkQueueStats* stats);

enum WorkQueueStatsEvent {
    STREAM_EVENT = 0,
    RUN_EVENT    = 1,
};

struct WorkQueueStatsView
{
    NVBIO_FORCEINLINE NVBIO_HOST_DEVICE
    WorkQueueStatsView() : active_lanes(NULL), issued_warps(NULL), iterations(NULL) {}

    NVBIO_FORCEINLINE NVBIO_HOST_DEVICE
    WorkQueueStatsView(uint64* _active_lanes, uint64* _issued_warps, uint64* _iterations) :
        active_lanes(_active_lanes), issued_warps(_issued_warps), iterations(_iterations) {}

    NVBIO_FORCEINLINE NVBIO_DEVICE
    void sample(const WorkQueueStatsEvent type);

    NVBIO_FORCEINLINE NVBIO_DEVICE
    void sample_iterations(const uint32 i);

    NVBIO_FORCEINLINE NVBIO_DEVICE
    bool valid() const { return active_lanes != NULL; }

    uint64* active_lanes;
    uint64* issued_warps;
    uint64* iterations;
};

struct DefaultMover
{
    template <typename WorkStreamT, typename WorkUnitT>
    NVBIO_FORCEINLINE NVBIO_HOST_DEVICE
    void move(
        const WorkStreamT& stream,
        const uint2 src_slot, WorkUnitT* src_unit,
        const uint2 dst_slot, WorkUnitT* dst_unit) const
    {
        // call WorkUnitT's copy operator and call it a day
        *dst_unit = *src_unit;
    }
};

template <
    typename PolicyTag,
    typename WorkUnitT,
    uint32   BLOCKDIM>
struct WorkQueue
{
    typedef WorkUnitT   WorkUnit;

    WorkQueue() {}

    void set_capacity(const uint32 capacity) {}

    template <typename WorkStream>
    void consume(const WorkStream stream, WorkQueueStats* stats = NULL) { consume( stream, DefaultMover(), stats ); }

    template <typename WorkStream, typename WorkMover>
    void consume(const WorkStream stream, const WorkMover mover, WorkQueueStats* stats = NULL);
};

struct WorkQueueStats
{
    typedef WorkQueueStatsView View;

    WorkQueueStats() : counters(7,0) {}

    void clear() { thrust::fill( counters.begin(), counters.end(), uint64(0u) ); }

    View view()
    {
        return View(
            thrust::raw_pointer_cast( &counters.front() ),
            thrust::raw_pointer_cast( &counters.front() ) + 2u,
            thrust::raw_pointer_cast( &counters.front() ) + 4u );
    }

    float utilization(const WorkQueueStatsEvent type) const { return counters[2 + type] ? float(counters[0 + type])/float(counters[2 + type]*cuda::Arch::WARP_SIZE) : 1.0f; }

    float avg_iterations() const { return counters[6] ? float(counters[4])/float(counters[6]) : 0.0f; }

    float max_iterations() const { return float(counters[5]); }

private:
    thrust::device_vector<uint64>   counters;
};

inline WorkQueueStatsView view(WorkQueueStats* stats) { return stats ? stats->view() : WorkQueueStatsView(); }


} // namespace cuda
} // namespace nvbio

#include <nvbio/basic/cuda/work_queue_inl.h>
#include <nvbio/basic/cuda/work_queue_persistent.h>
#include <nvbio/basic/cuda/work_queue_ordered.h>
#include <nvbio/basic/cuda/work_queue_multipass.h>