sparse_matrix_operations.hpp

Go to the documentation of this file.

#ifndef VIENNACL_LINALG_SPARSE_MATRIX_OPERATIONS_HPP_
#define VIENNACL_LINALG_SPARSE_MATRIX_OPERATIONS_HPP_

/* =========================================================================
   Copyright (c) 2010-2016, Institute for Microelectronics,
                            Institute for Analysis and Scientific Computing,
                            TU Wien.
   Portions of this software are copyright by UChicago Argonne, LLC.

                            -----------------
                  ViennaCL - The Vienna Computing Library
                            -----------------

   Project Head:    Karl Rupp                   rupp@iue.tuwien.ac.at

   (A list of authors and contributors can be found in the manual)

   License:         MIT (X11), see file LICENSE in the base directory
============================================================================= */

#include "viennacl/forwards.h"
#include "viennacl/scalar.hpp"
#include "viennacl/vector.hpp"
#include "viennacl/matrix.hpp"
#include "viennacl/tools/tools.hpp"
#include "viennacl/linalg/host_based/sparse_matrix_operations.hpp"

#ifdef VIENNACL_WITH_OPENCL
  #include "viennacl/linalg/opencl/sparse_matrix_operations.hpp"
#endif

#ifdef VIENNACL_WITH_CUDA
  #include "viennacl/linalg/cuda/sparse_matrix_operations.hpp"
#endif

namespace viennacl
{
  namespace linalg
  {

    namespace detail
    {

      template<typename SparseMatrixType, typename SCALARTYPE, unsigned int VEC_ALIGNMENT>
      typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value >::type
      row_info(SparseMatrixType const & mat,
               vector<SCALARTYPE, VEC_ALIGNMENT> & vec,
               row_info_types info_selector)
      {
        switch (viennacl::traits::handle(mat).get_active_handle_id())
        {
          case viennacl::MAIN_MEMORY:
            viennacl::linalg::host_based::detail::row_info(mat, vec, info_selector);
            break;
#ifdef VIENNACL_WITH_OPENCL
          case viennacl::OPENCL_MEMORY:
            viennacl::linalg::opencl::detail::row_info(mat, vec, info_selector);
            break;
#endif
#ifdef VIENNACL_WITH_CUDA
          case viennacl::CUDA_MEMORY:
            viennacl::linalg::cuda::detail::row_info(mat, vec, info_selector);
            break;
#endif
          case viennacl::MEMORY_NOT_INITIALIZED:
            throw memory_exception("not initialised!");
          default:
            throw memory_exception("not implemented");
        }
      }

    }



    // A * x

    template<typename SparseMatrixType, class ScalarType>
    typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
    prod_impl(const SparseMatrixType & mat,
              const viennacl::vector_base<ScalarType> & vec,
              ScalarType alpha,
                    viennacl::vector_base<ScalarType> & result,
              ScalarType beta)
    {
      assert( (mat.size1() == result.size()) && bool("Size check failed for compressed matrix-vector product: size1(mat) != size(result)"));
      assert( (mat.size2() == vec.size())    && bool("Size check failed for compressed matrix-vector product: size2(mat) != size(x)"));

      switch (viennacl::traits::handle(mat).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
        viennacl::linalg::host_based::prod_impl(mat, vec, alpha, result, beta);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::prod_impl(mat, vec, alpha, result, beta);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::prod_impl(mat, vec, alpha, result, beta);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }


    // A * B
    template<typename SparseMatrixType, class ScalarType>
    typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
    prod_impl(const SparseMatrixType & sp_mat,
              const viennacl::matrix_base<ScalarType> & d_mat,
                    viennacl::matrix_base<ScalarType> & result)
    {
      assert( (sp_mat.size1() == result.size1()) && bool("Size check failed for compressed matrix - dense matrix product: size1(sp_mat) != size1(result)"));
      assert( (sp_mat.size2() == d_mat.size1()) && bool("Size check failed for compressed matrix - dense matrix product: size2(sp_mat) != size1(d_mat)"));

      switch (viennacl::traits::handle(sp_mat).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
          viennacl::linalg::host_based::prod_impl(sp_mat, d_mat, result);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::prod_impl(sp_mat, d_mat, result);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::prod_impl(sp_mat, d_mat, result);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }

    // A * transpose(B)
    template<typename SparseMatrixType, class ScalarType>
    typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
    prod_impl(const SparseMatrixType & sp_mat,
              const viennacl::matrix_expression<const viennacl::matrix_base<ScalarType>,
                                                const viennacl::matrix_base<ScalarType>,
                                                viennacl::op_trans>& d_mat,
                    viennacl::matrix_base<ScalarType> & result)
    {
      assert( (sp_mat.size1() == result.size1()) && bool("Size check failed for compressed matrix - dense matrix product: size1(sp_mat) != size1(result)"));
      assert( (sp_mat.size2() == d_mat.size1()) && bool("Size check failed for compressed matrix - dense matrix product: size2(sp_mat) != size1(d_mat)"));

      switch (viennacl::traits::handle(sp_mat).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
          viennacl::linalg::host_based::prod_impl(sp_mat, d_mat, result);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::prod_impl(sp_mat, d_mat, result);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::prod_impl(sp_mat, d_mat, result);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }

    // A * B with both A and B sparse

    template<typename NumericT>
    void
    prod_impl(const viennacl::compressed_matrix<NumericT> & A,
              const viennacl::compressed_matrix<NumericT> & B,
                    viennacl::compressed_matrix<NumericT> & C)
    {
      assert( (A.size2() == B.size1())                    && bool("Size check failed for sparse matrix-matrix product: size2(A) != size1(B)"));
      assert( (C.size1() == 0 || C.size1() == A.size1())  && bool("Size check failed for sparse matrix-matrix product: size1(A) != size1(C)"));
      assert( (C.size2() == 0 || C.size2() == B.size2())  && bool("Size check failed for sparse matrix-matrix product: size2(B) != size2(B)"));

      switch (viennacl::traits::handle(A).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
          viennacl::linalg::host_based::prod_impl(A, B, C);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::prod_impl(A, B, C);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::prod_impl(A, B, C);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }


    template<typename SparseMatrixType, class ScalarType, typename SOLVERTAG>
    typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
    inplace_solve(const SparseMatrixType & mat,
                  viennacl::vector_base<ScalarType> & vec,
                  SOLVERTAG tag)
    {
      assert( (mat.size1() == mat.size2()) && bool("Size check failed for triangular solve on compressed matrix: size1(mat) != size2(mat)"));
      assert( (mat.size2() == vec.size())    && bool("Size check failed for compressed matrix-vector product: size2(mat) != size(x)"));

      switch (viennacl::traits::handle(mat).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
          viennacl::linalg::host_based::inplace_solve(mat, vec, tag);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::inplace_solve(mat, vec, tag);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::inplace_solve(mat, vec, tag);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }


    template<typename SparseMatrixType, class ScalarType, typename SOLVERTAG>
    typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
    inplace_solve(const matrix_expression<const SparseMatrixType, const SparseMatrixType, op_trans> & mat,
                  viennacl::vector_base<ScalarType> & vec,
                  SOLVERTAG tag)
    {
      assert( (mat.size1() == mat.size2()) && bool("Size check failed for triangular solve on transposed compressed matrix: size1(mat) != size2(mat)"));
      assert( (mat.size1() == vec.size())    && bool("Size check failed for transposed compressed matrix triangular solve: size1(mat) != size(x)"));

      switch (viennacl::traits::handle(mat.lhs()).get_active_handle_id())
      {
        case viennacl::MAIN_MEMORY:
          viennacl::linalg::host_based::inplace_solve(mat, vec, tag);
          break;
#ifdef VIENNACL_WITH_OPENCL
        case viennacl::OPENCL_MEMORY:
          viennacl::linalg::opencl::inplace_solve(mat, vec, tag);
          break;
#endif
#ifdef VIENNACL_WITH_CUDA
        case viennacl::CUDA_MEMORY:
          viennacl::linalg::cuda::inplace_solve(mat, vec, tag);
          break;
#endif
        case viennacl::MEMORY_NOT_INITIALIZED:
          throw memory_exception("not initialised!");
        default:
          throw memory_exception("not implemented");
      }
    }



    namespace detail
    {

      template<typename SparseMatrixType, class ScalarType, typename SOLVERTAG>
      typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value>::type
      block_inplace_solve(const matrix_expression<const SparseMatrixType, const SparseMatrixType, op_trans> & mat,
                          viennacl::backend::mem_handle const & block_index_array, vcl_size_t num_blocks,
                          viennacl::vector_base<ScalarType> const & mat_diagonal,
                          viennacl::vector_base<ScalarType> & vec,
                          SOLVERTAG tag)
      {
        assert( (mat.size1() == mat.size2()) && bool("Size check failed for triangular solve on transposed compressed matrix: size1(mat) != size2(mat)"));
        assert( (mat.size1() == vec.size())  && bool("Size check failed for transposed compressed matrix triangular solve: size1(mat) != size(x)"));

        switch (viennacl::traits::handle(mat.lhs()).get_active_handle_id())
        {
          case viennacl::MAIN_MEMORY:
            viennacl::linalg::host_based::detail::block_inplace_solve(mat, block_index_array, num_blocks, mat_diagonal, vec, tag);
            break;
  #ifdef VIENNACL_WITH_OPENCL
          case viennacl::OPENCL_MEMORY:
            viennacl::linalg::opencl::detail::block_inplace_solve(mat, block_index_array, num_blocks, mat_diagonal, vec, tag);
            break;
  #endif
  #ifdef VIENNACL_WITH_CUDA
          case viennacl::CUDA_MEMORY:
            viennacl::linalg::cuda::detail::block_inplace_solve(mat, block_index_array, num_blocks, mat_diagonal, vec, tag);
            break;
  #endif
          case viennacl::MEMORY_NOT_INITIALIZED:
            throw memory_exception("not initialised!");
          default:
            throw memory_exception("not implemented");
        }
      }


    }



  } //namespace linalg


  template<typename M1>
  typename viennacl::enable_if<viennacl::is_any_sparse_matrix<M1>::value,
                                matrix_expression< const M1, const M1, op_trans>
                              >::type
  trans(const M1 & mat)
  {
    return matrix_expression< const M1, const M1, op_trans>(mat, mat);
  }

  //free functions:
  template<typename SCALARTYPE, typename SparseMatrixType>
  typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value,
                                viennacl::vector<SCALARTYPE> >::type
  operator+(viennacl::vector_base<SCALARTYPE> & result,
            const viennacl::vector_expression< const SparseMatrixType, const viennacl::vector_base<SCALARTYPE>, viennacl::op_prod> & proxy)
  {
    assert(proxy.lhs().size1() == result.size() && bool("Dimensions for addition of sparse matrix-vector product to vector don't match!"));
    vector<SCALARTYPE> temp(proxy.lhs().size1());
    viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), temp);
    result += temp;
    return result;
  }

  template<typename SCALARTYPE, typename SparseMatrixType>
  typename viennacl::enable_if< viennacl::is_any_sparse_matrix<SparseMatrixType>::value,
                                viennacl::vector<SCALARTYPE> >::type
  operator-(viennacl::vector_base<SCALARTYPE> & result,
            const viennacl::vector_expression< const SparseMatrixType, const viennacl::vector_base<SCALARTYPE>, viennacl::op_prod> & proxy)
  {
    assert(proxy.lhs().size1() == result.size() && bool("Dimensions for addition of sparse matrix-vector product to vector don't match!"));
    vector<SCALARTYPE> temp(proxy.lhs().size1());
    viennacl::linalg::prod_impl(proxy.lhs(), proxy.rhs(), temp);
    result += temp;
    return result;
  }

} //namespace viennacl


#endif