GPUMatrix.cpp

Go to the documentation of this file.

/*
 * Copyright (c) 2014, Shogun Toolbox Foundation
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:

 * 1. Redistributions of source code must retain the above copyright notice,
 * this list of conditions and the following disclaimer.
 *
 * 2. 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.
 *
 * 3. Neither the name of the copyright holder 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 THE COPYRIGHT HOLDER OR CONTRIBUTORS 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.
 *
 * Written (W) 2014 Khaled Nasr
 */

#include <shogun/lib/config.h>

#ifdef HAVE_VIENNACL
#ifdef HAVE_CXX11

#include <shogun/lib/GPUMatrix.h>
#include <viennacl/matrix.hpp>

#include <shogun/mathematics/eigen3.h>
#include <shogun/lib/SGMatrix.h>

#include <type_traits>

namespace shogun
{

template <class T>
CGPUMatrix<T>::CGPUMatrix()
{
    init();
}

template <class T>
CGPUMatrix<T>::CGPUMatrix(index_t nrows, index_t ncols) : matrix(new VCLMemoryArray())
{
    init();

    num_rows = nrows;
    num_cols = ncols;

    viennacl::backend::memory_create(*matrix, sizeof(T)*num_rows*num_cols,
        viennacl::context());
}

template <class T>
CGPUMatrix<T>::CGPUMatrix(std::shared_ptr<VCLMemoryArray> mem, index_t nrows, index_t ncols,
    index_t mem_offset)
{
    init();

    matrix = mem;
    num_rows = nrows;
    num_cols = ncols;
    offset = mem_offset;
}

template <class T>
CGPUMatrix<T>::CGPUMatrix(const SGMatrix< T >& cpu_mat) : matrix(new VCLMemoryArray())
{
    init();

    num_rows = cpu_mat.num_rows;
    num_cols = cpu_mat.num_cols;

    viennacl::backend::memory_create(*matrix, sizeof(T)*num_rows*num_cols,
        viennacl::context());

    viennacl::backend::memory_write(*matrix, 0, num_rows*num_cols*sizeof(T),
        cpu_mat.matrix);
}

template <class T>
CGPUMatrix<T>::CGPUMatrix(const EigenMatrixXt& cpu_mat)
: matrix(new VCLMemoryArray())
{
    init();

    num_rows = cpu_mat.rows();
    num_cols = cpu_mat.cols();

    viennacl::backend::memory_create(*matrix, sizeof(T)*num_rows*num_cols,
        viennacl::context());

    viennacl::backend::memory_write(*matrix, 0, num_rows*num_cols*sizeof(T),
        cpu_mat.data());
}

template <class T>
CGPUMatrix<T>::operator EigenMatrixXt() const
{
    EigenMatrixXt cpu_mat(num_rows, num_cols);

    viennacl::backend::memory_read(*matrix, offset*sizeof(T), num_rows*num_cols*sizeof(T),
        cpu_mat.data());

    return cpu_mat;
}

template <class T>
CGPUMatrix<T>::operator SGMatrix<T>() const
{
    SGMatrix<T> cpu_mat(num_rows, num_cols);

    viennacl::backend::memory_read(*matrix, offset*sizeof(T), num_rows*num_cols*sizeof(T),
        cpu_mat.matrix);

    return cpu_mat;
}

template <class T>
typename CGPUMatrix<T>::VCLMatrixBase CGPUMatrix<T>::vcl_matrix()
{
#if VIENNACL_VERSION >= 10600
    return VCLMatrixBase(*matrix, num_rows, offset, 1, num_rows, num_cols, 0, 1, num_cols, false);
#else
    return VCLMatrixBase(*matrix, num_rows, offset, 1, num_rows, num_cols, 0, 1, num_cols);
#endif
}

template <class T>
void CGPUMatrix<T>::display_matrix(const char* name) const
{
    ((SGMatrix<T>)*this).display_matrix(name);
}

template <class T>
void CGPUMatrix<T>::zero()
{
    vcl_matrix().clear();
}

template <class T>
void CGPUMatrix<T>::set_const(T value)
{
    VCLMatrixBase m = vcl_matrix();
    viennacl::linalg::matrix_assign(m, value);
}

template <class T>
viennacl::const_entry_proxy<T> CGPUMatrix<T>::operator()(index_t i, index_t j) const
{
    return viennacl::const_entry_proxy<T>(offset+i+j*num_rows, *matrix);
}

template <class T>
viennacl::entry_proxy< T > CGPUMatrix<T>::operator()(index_t i, index_t j)
{
    return viennacl::entry_proxy<T>(offset+i+j*num_rows, *matrix);
}

template <class T>
viennacl::const_entry_proxy< T > CGPUMatrix<T>::operator[](index_t index) const
{
    return viennacl::const_entry_proxy<T>(offset+index, *matrix);
}

template <class T>
viennacl::entry_proxy< T > CGPUMatrix<T>::operator[](index_t index)
{
    return viennacl::entry_proxy<T>(offset+index, *matrix);
}

template <class T>
void CGPUMatrix<T>::init()
{
    num_rows = 0;
    num_cols = 0;
    offset = 0;
}

template<typename T> struct dummy {};
template<typename T> class CGPUMatrix<dummy<T> > {};

template class CGPUMatrix<char>;
template class CGPUMatrix<uint8_t>;
template class CGPUMatrix<int16_t>;
template class CGPUMatrix<uint16_t>;
template class CGPUMatrix<int32_t>;
template class CGPUMatrix<uint32_t>;
template class CGPUMatrix<std::conditional<viennacl::is_primitive_type<int64_t>::value, int64_t, dummy<int64_t> >::type>;
template class CGPUMatrix<std::conditional<viennacl::is_primitive_type<uint64_t>::value, uint64_t, dummy<uint64_t> >::type>;
template class CGPUMatrix<float32_t>;
template class CGPUMatrix<float64_t>;
}

#endif // HAVE_CXX11
#endif // HAVE_VIENNACL