matrix_fold.hpp

Go to the documentation of this file.

/*!
 * \brief       kernels for folding matrices with openCL
 * 
 * \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_CLBLAS_MATRIX_FOLD_HPP
#define REMORA_KERNELS_CLBLAS_MATRIX_FOLD_HPP

#include "../../expression_types.hpp"
#include "../../detail/traits.hpp"
#include <boost/compute/kernel.hpp>
#include <boost/compute/detail/meta_kernel.hpp>
#include <boost/compute/container/array.hpp>
#include <boost/compute/algorithm/copy_n.hpp>
namespace remora{namespace bindings{

template<class F, class MatA, class Orientation>
void matrix_fold(matrix_expression<MatA, gpu_tag> const& A, typename F::result_type& value, Orientation, dense_tag) {
    auto& queue = A().queue();
    typedef typename F::result_type value_type;
    boost::compute::detail::meta_kernel k("blas_matrix_fold");
    std::size_t size1_index = k.add_arg<std::size_t>("size1");
    std::size_t size2_index = k.add_arg<std::size_t>("size2");
    boost::compute::array<value_type,1> device_result;
    boost::compute::copy_n(&value, 1, device_result.begin(), queue);
    device_result.front() = value;
    F f;
    
    //read all tiles in the assigned rows and apply f
    k << "__local " <<k.decl<value_type>("subfold")<< "[TILE_DIM][TILE_DIM+1];";
    k << "subfold[get_local_id(0)][get_local_id(1)]  = "<<device_result.begin()[0]<<';';
    k << "for(uint i = get_local_id(0) ; i < size1; i += TILE_DIM){";
    k << "    for(uint j = get_local_id(1) ; j < size2; j += TILE_DIM){";
    auto exprSubFold = k.expr<value_type>("subfold[get_local_id(0)][get_local_id(1)]");
    k<< exprSubFold << '=' << f(exprSubFold,A()(k.expr<cl_uint>("i"),k.expr<cl_uint>("j")))<<";";
    k<<"}}";
    k << "barrier(CLK_LOCAL_MEM_FENCE);";//wait until all threads are done with copying
    //sum up the rows
    k << "if(get_local_id(0) == 0){";
    k << "    for(uint i = 1 ; i < TILE_DIM; ++i){";
    k << "        subfold[0][get_local_id(1)] =" 
        << f(
            k.expr<value_type>("subfold[0][get_local_id(1)]"),
            k.expr<value_type>("subfold[i][get_local_id(1)]")
        )<<';';
    k << "    }";
    k <<"    if(get_local_id(1) == 0){";
    k << "       for(uint i = 1 ; i < TILE_DIM; ++i){";
    k <<"            subfold[0][0] =" << f(k.expr<value_type>("subfold[0][0]"),k.expr<value_type>("subfold[0][i]"))<<';';
    k <<"        }";
    k <<device_result.begin()[0]<< "= subfold[0][0];";
    k<< "}}";
    
    //compile kernel
    std::size_t TILE_DIM = 1;
    char const* options ="-DTILE_DIM=1";
    boost::compute::kernel kernel = k.compile(queue.get_context(), options);
    //enqueue kernel
    kernel.set_arg(size1_index, A().size1());
    kernel.set_arg(size2_index, A().size2());
    
    std::size_t global_work_size[2] = {TILE_DIM,TILE_DIM};
    std::size_t local_work_size[2] = {TILE_DIM, TILE_DIM};
    queue.enqueue_nd_range_kernel(kernel, 2,nullptr, global_work_size, local_work_size);
    boost::compute::copy_n(device_result.begin(), 1, &value, queue);
}
}}
#endif