gemv.hpp

Go to the documentation of this file.

//===========================================================================
/*!
 * 
 *
 * \brief       -
 *
 * \author      O. Krause
 * \date        2016
 *
 *
 * \par Copyright 1995-2015 Shark Development Team
 * 
 * <BR><HR>
 * This file is part of Shark.
 * <http://image.diku.dk/shark/>
 * 
 * Shark is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published 
 * by the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 * 
 * Shark is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Lesser General Public License for more details.
 * 
 * You should have received a copy of the GNU Lesser General Public License
 * along with Shark.  If not, see <http://www.gnu.org/licenses/>.
 *
 */
//===========================================================================
#ifndef REMORA_KERNELS_GPU_GEMV_HPP
#define REMORA_KERNELS_GPU_GEMV_HPP

#include "../../expression_types.hpp"
#include "../../detail/traits.hpp"
#include <boost/compute/kernel.hpp>
#include <boost/compute/detail/meta_kernel.hpp>
#include <boost/compute/functional/operator.hpp> //for multiplies

namespace remora{ namespace kernels{

// v <- v + alpha * A * x
template <typename MatA, typename VecX, typename VecV>
void gemv(
    matrix_expression<MatA, gpu_tag> const& A,
    vector_expression<VecX, gpu_tag> const& x,
    vector_expression<VecV, gpu_tag>& v, 
    typename VecV::value_type const& alpha
) {
    REMORA_SIZE_CHECK(A().size1() == v().size());
    REMORA_SIZE_CHECK(A().size2() == x().size());
    
    
    typedef typename VecV::value_type value_type;
    boost::compute::multiplies<value_type> prod;
    boost::compute::detail::meta_kernel k("blas_gemv");
    std::size_t alpha_index = k.add_arg<value_type>("alpha");
    std::size_t size1_index = k.add_arg<std::size_t>("size1");
    std::size_t size2_index = k.add_arg<std::size_t>("size2");
    //read all tiles in the assigned rows and compute the inner product
    k << "__local " <<k.decl<value_type>("results")<< "[TILE_DIM][TILE_DIM+2];";
    k << "uint rowid = get_global_id(0);";
    k << "results[get_local_id(0)][get_local_id(1)]  = 0.0;";
    k << "for(uint i = get_local_id(1) ; i < size2 && rowid < size1; i += TILE_DIM){";
    auto exprRow = k.expr<cl_uint>("rowid");
    auto exprCol = k.expr<cl_uint>("i");
    k<< "    results[get_local_id(0)][get_local_id(1)] += "<< prod(A()(exprRow,exprCol),x()(exprCol))<<";";
    k<<'}';
    k << "barrier(CLK_LOCAL_MEM_FENCE);";//wait until all threads are done with computing
    //sum up the rows
    k << "if(get_local_id(1) == 0 && rowid < size1){";
    k << "    for(uint i = 1 ; i < TILE_DIM; ++i){";
    k << "        results[get_local_id(0)][0] +=results[get_local_id(0)][i];";
    k << "    }";
    k << v()(exprRow) << "+= alpha * results[get_local_id(0)][0];";
    k<< "}";
    //create source

    std::size_t TILE_DIM = 16;
    char const* options ="-DTILE_DIM=16";
    boost::compute::kernel kernel = k.compile(v().queue().get_context(), options);
    //enqueue kernel
    kernel.set_arg(alpha_index, alpha);
    kernel.set_arg(size1_index, A().size1());
    kernel.set_arg(size2_index, A().size2());
    
    std::size_t global_work_size[2] = {
        ((A().size1()+TILE_DIM-1)/TILE_DIM) * TILE_DIM,
        TILE_DIM
    };
    std::size_t local_work_size[2] = {TILE_DIM,TILE_DIM};
    v().queue().enqueue_nd_range_kernel(kernel, 2,nullptr, global_work_size, local_work_size);
}

}}

#endif