copy_grid_fast.hpp

/*
 * copy_grid_fast.hpp
 *
 *  Created on: Nov 29, 2017
 *      Author: i-bird
 */
 
#ifndef OPENFPM_DATA_SRC_GRID_COPY_GRID_FAST_HPP_
#define OPENFPM_DATA_SRC_GRID_COPY_GRID_FAST_HPP_
 
#include "Grid/iterators/grid_key_dx_iterator.hpp"
 
template<unsigned int dim>
struct striding
{
    size_t striding_src[dim - 1];
    size_t striding_dst[dim - 1];
    size_t n_cpy;
    size_t tot_y;
};
 
 
template<bool lin_or_inte,typename data_type,typename S>
struct copy_fast_1d_device_memory
{
    data_type & data_src;
 
    data_type & data_dst;
 
    inline copy_fast_1d_device_memory(data_type & data_src, data_type & data_dst)
    :data_src(data_src),data_dst(data_dst)
    {};
 
    template<typename T>
    inline void operator()(T& t) const
    {
        static_cast<S *>(boost::fusion::at_c<T::value>(data_dst).mem)->copyDeviceToDevice(*static_cast<S *>(boost::fusion::at_c<T::value>(data_src).mem));
    }
};
 
template<typename data_type,typename S>
struct copy_fast_1d_device_memory<true,data_type,S>
{
    data_type & data_src;
 
    data_type & data_dst;
 
    inline copy_fast_1d_device_memory(data_type & data_src, data_type & data_dst)
    :data_src(data_src),data_dst(data_dst)
    {};
 
    template<typename T>
    inline void operator()(T& t) const
    {
        if (T::value == 0)
        {
            static_cast<S *>(data_dst.mem)->copyDeviceToDevice(*static_cast<S *>(data_src.mem));
        }
    }
};
 
 
template<bool is_complex, unsigned int N, typename grid, typename ginfo>
struct copy_grid_fast
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<N,size_t> & bx_src,
                   const Box<N,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<N> (& cnt)[1] )
    {
        grid_key_dx_iterator_sub<N,stencil_offset_compute<N,1>> sub_src(gs_src,bx_src.getKP1(),bx_src.getKP2(),cnt);
        grid_key_dx_iterator_sub<N,stencil_offset_compute<N,1>> sub_dst(gs_dst,bx_dst.getKP1(),bx_dst.getKP2(),cnt);
 
        while (sub_src.isNext())
        {
            // Option 1
            gd_dst.set(sub_dst.template getStencil<0>(),gd_src,sub_src.template getStencil<0>());
 
            ++sub_src;
            ++sub_dst;
        }
    }
};
 
template<typename grid, typename ginfo>
struct copy_grid_fast<true,3,grid,ginfo>
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<3,size_t> & bx_src,
                   const Box<3,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<3> (& cnt)[1] )
    {
        size_t lin_src = 0;
        size_t lin_dst = 0;
 
        lin_src += bx_src.getLow(2) * gs_src.size_s(1);
        lin_dst += bx_dst.getLow(2) * gs_dst.size_s(1);
        for (size_t i = bx_src.getLow(2) ; i <= bx_src.getHigh(2) ; i++)
        {
            lin_src += bx_src.getLow(1) * gs_src.size_s(0);
            lin_dst += bx_dst.getLow(1) * gs_dst.size_s(0);
            for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
            {
                lin_src += bx_src.getLow(0);
                lin_dst += bx_dst.getLow(0);
                for (size_t k = bx_src.getLow(0) ; k <= bx_src.getHigh(0) ; k++)
                {
                    gd_dst.set(lin_dst,gd_src,lin_src);
 
                    lin_src++;
                    lin_dst++;
                }
                lin_src -= bx_src.getHigh(0) + 1;
                lin_dst -= bx_dst.getHigh(0) + 1;
                lin_src += gs_src.size_s(0);
                lin_dst += gs_dst.size_s(0);
            }
            lin_src -= (bx_src.getHigh(1) + 1)*gs_src.size_s(0);
            lin_dst -= (bx_dst.getHigh(1) + 1)*gs_dst.size_s(0);
            lin_src += gs_src.size_s(1);
            lin_dst += gs_dst.size_s(1);
        }
    }
};
 
 
template<typename grid, typename ginfo>
struct copy_grid_fast<true,2,grid,ginfo>
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<2,size_t> & bx_src,
                   const Box<2,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<2> (& cnt)[1] )
    {
        size_t lin_src = 0;
        size_t lin_dst = 0;
 
 
        lin_src += bx_src.getLow(1) * gs_src.size_s(0);
        lin_dst += bx_dst.getLow(1) * gs_dst.size_s(0);
        for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
        {
            lin_src += bx_src.getLow(0);
            lin_dst += bx_dst.getLow(0);
            for (size_t k = bx_src.getLow(0) ; k <= bx_src.getHigh(0) ; k++)
            {
                gd_dst.set(lin_dst,gd_src,lin_src);
 
                lin_src++;
                lin_dst++;
            }
            lin_src -= bx_src.getHigh(0) + 1;
            lin_dst -= bx_dst.getHigh(0) + 1;
            lin_src += gs_src.size_s(0);
            lin_dst += gs_dst.size_s(0);
        }
 
    }
};
 
 
 
template<typename grid, typename ginfo>
struct copy_grid_fast<true,1,grid,ginfo>
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<1,size_t> & bx_src,
                   const Box<1,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<1> (& cnt)[1] )
    {
        size_t lin_src = 0;
        size_t lin_dst = 0;
 
        lin_src += bx_src.getLow(0);
        lin_dst += bx_dst.getLow(0);
        for (size_t k = bx_src.getLow(0) ; k <= bx_src.getHigh(0) ; k++)
        {
            gd_dst.set(lin_dst,gd_src,lin_src);
 
            lin_src++;
            lin_dst++;
        }
 
 
    }
};
 
 
template<unsigned int object_size>
void copy_grid_fast_longx_3(const Box<3,size_t> & bx_src,
                            unsigned char * ptr_dst,
                            unsigned char * ptr_src,
                            striding<3> & sr)
{
    for (size_t i = bx_src.getLow(2) ; i <= bx_src.getHigh(2) ; i++)
    {
        for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
        {
            memcpy(ptr_dst,ptr_src,sr.n_cpy*object_size);
 
            ptr_dst += sr.striding_dst[0];
            ptr_src += sr.striding_src[0];
        }
        ptr_dst += sr.striding_dst[1] - sr.tot_y*sr.striding_dst[0];
        ptr_src += sr.striding_src[1] - sr.tot_y*sr.striding_src[0];
    }
}
 
template<unsigned int object_size, unsigned int n_cpy>
void copy_grid_fast_shortx_3(const Box<3,size_t> & bx_src,
        unsigned char * ptr_dst,
        unsigned char * ptr_src,
        striding<3> & sr)
{
    for (size_t i = bx_src.getLow(2) ; i <= bx_src.getHigh(2) ; i++)
    {
        for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
        {
            __builtin_memcpy(ptr_dst,ptr_src,n_cpy*object_size);
 
            ptr_dst += sr.striding_dst[0];
            ptr_src += sr.striding_src[0];
        }
        ptr_dst += sr.striding_dst[1] - sr.tot_y*sr.striding_dst[0];
        ptr_src += sr.striding_src[1] - sr.tot_y*sr.striding_src[0];
    }
}
 
 
 
template<unsigned int object_size>
void copy_grid_fast_longx_2(const Box<2,size_t> & bx_src,
                            unsigned char * ptr_dst,
                            unsigned char * ptr_src,
                            striding<2> & sr)
{
    for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
    {
        memcpy(ptr_dst,ptr_src,sr.n_cpy*object_size);
 
        ptr_dst += sr.striding_dst[0];
        ptr_src += sr.striding_src[0];
    }
}
 
template<unsigned int object_size, unsigned int n_cpy>
void copy_grid_fast_shortx_2(const Box<2,size_t> & bx_src,
        unsigned char * ptr_dst,
        unsigned char * ptr_src,
        striding<2> & sr)
{
    for (size_t j = bx_src.getLow(1) ; j <= bx_src.getHigh(1) ; j++)
    {
        __builtin_memcpy(ptr_dst,ptr_src,n_cpy*object_size);
 
        ptr_dst += sr.striding_dst[0];
        ptr_src += sr.striding_src[0];
    }
}
 
template<unsigned int dim>
struct copy_ndim_fast_selector
{
    template<unsigned int object_size>
    static void call(unsigned char * ptr_src,
             unsigned char * ptr_dst,
             striding<dim> & sr,
  const Box<dim,size_t> & bx_src)
    {
        std::cout << __FILE__ << ":" << __LINE__ << " error, wrong specialization. this case is not handled" << std::endl;
    }
};
 
template<>
struct copy_ndim_fast_selector<3>
{
    template<unsigned int object_size>
    static void call(unsigned char * ptr_src,
                     unsigned char * ptr_dst,
                     striding<3> & sr,
           const Box<3,size_t> & bx_src)
    {
        switch (sr.n_cpy)
        {
        case 1:
                copy_grid_fast_shortx_3<object_size,1>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 2:
                copy_grid_fast_shortx_3<object_size,2>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 3:
                copy_grid_fast_shortx_3<object_size,3>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 4:
                copy_grid_fast_shortx_3<object_size,4>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 5:
                copy_grid_fast_shortx_3<object_size,5>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 6:
                copy_grid_fast_shortx_3<object_size,6>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 7:
                copy_grid_fast_shortx_3<object_size,7>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 8:
                copy_grid_fast_shortx_3<object_size,8>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        default:
                copy_grid_fast_longx_3<object_size>(bx_src,ptr_dst,ptr_src,sr);
        }
    }
};
 
template<>
struct copy_ndim_fast_selector<2>
{
    template<unsigned int object_size>
    static void call(unsigned char * ptr_src,
             unsigned char * ptr_dst,
             striding<2> & sr,
  const Box<2,size_t> & bx_src)
    {
        switch (sr.n_cpy)
        {
        case 1:
                copy_grid_fast_shortx_2<object_size,1>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 2:
                copy_grid_fast_shortx_2<object_size,2>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 3:
                copy_grid_fast_shortx_2<object_size,3>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 4:
                copy_grid_fast_shortx_2<object_size,4>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 5:
                copy_grid_fast_shortx_2<object_size,5>(bx_src,ptr_dst,ptr_src,sr);
                break;
        case 6:
                copy_grid_fast_shortx_2<object_size,6>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 7:
                copy_grid_fast_shortx_2<object_size,7>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        case 8:
                copy_grid_fast_shortx_2<object_size,8>(bx_src,ptr_dst,ptr_src,sr);
                break;
 
        default:
                copy_grid_fast_longx_2<object_size>(bx_src,ptr_dst,ptr_src,sr);
        }
    }
};
 
 
template<unsigned int dim_prp, unsigned int prp,typename grid_type>
struct get_pointer
{
    static unsigned char * get(const grid_type & gd, grid_key_dx<grid_type::dims> & key, unsigned int (& id)[dim_prp])
    {
        return (unsigned char *)(&gd.template get_unsafe<prp>(key));
    }
};
 
template<unsigned int prp,typename grid_type>
struct get_pointer<1,prp,grid_type>
{
    static unsigned char * get(const grid_type & gd, grid_key_dx<grid_type::dims> & key, unsigned int (& id)[1])
    {
        return (unsigned char *)(&gd.template get_unsafe<prp>(key)[id[0]]);
    }
};
 
template<unsigned int prp,typename grid_type>
struct get_pointer<2,prp,grid_type>
{
    static unsigned char * get(const grid_type & gd, grid_key_dx<grid_type::dims> & key, unsigned int (& id)[2])
    {
        return (unsigned char *)(&gd.template get_unsafe<prp>(key)[id[0]][id[1]]);
    }
};
 
template<unsigned int dim,unsigned int dim_prp,unsigned int prp, typename grid_type>
struct get_striding
{
    static striding<dim> get(const grid_type & gd_src,grid_type &gd_dst,
                               const Box<dim,size_t> & bx_src, unsigned int (& id)[dim_prp])
    {
        striding<dim> sr;
 
        grid_key_dx<dim> zero;
        zero.zero();
        grid_key_dx<dim> one = zero;
        one.set_d(1,1);
 
        unsigned char * ptr_final_src = get_pointer<dim_prp,prp,grid_type>::get(gd_src,one,id);
        unsigned char * ptr_start_src = get_pointer<dim_prp,prp,grid_type>::get(gd_src,zero,id);
 
        unsigned char * ptr_final_dst = get_pointer<dim_prp,prp,grid_type>::get(gd_dst,one,id);
        unsigned char * ptr_start_dst = get_pointer<dim_prp,prp,grid_type>::get(gd_dst,zero,id);
 
        sr.n_cpy = bx_src.getHigh(0) - bx_src.getLow(0) + 1;
        sr.tot_y = bx_src.getHigh(1) - bx_src.getLow(1) + 1;
 
        sr.striding_src[0] = ptr_final_src - ptr_start_src;
        sr.striding_dst[0] = ptr_final_dst - ptr_start_dst;
 
        if (dim == 3)
        {
            grid_key_dx<dim> one2 = zero;
            one2.set_d(2,1);
 
            unsigned char * ptr_final_src = get_pointer<dim_prp,prp,grid_type>::get(gd_src,one2,id);
            unsigned char * ptr_start_src = get_pointer<dim_prp,prp,grid_type>::get(gd_src,zero,id);
 
            unsigned char * ptr_final_dst = get_pointer<dim_prp,prp,grid_type>::get(gd_dst,one2,id);
            unsigned char * ptr_start_dst = get_pointer<dim_prp,prp,grid_type>::get(gd_dst,zero,id);
 
            sr.striding_src[1] = ptr_final_src - ptr_start_src;
            sr.striding_dst[1] = ptr_final_dst - ptr_start_dst;
        }
 
        return sr;
    };
};
 
template<unsigned int dim, typename T>
struct mp_funct_impl
{
    template<unsigned int prp, typename grid>
    static void process(const grid & gd_src,grid & gd_dst,
                        const Box<dim,size_t> & bx_src, const Box<grid::dims,size_t> & bx_dst)
    {
        unsigned int id[0];
        striding<dim> sr = get_striding<dim,0,prp,grid>::get(gd_src,gd_dst,bx_src,id);
 
        unsigned char * ptr_src = (unsigned char *)(&gd_src.template get<prp>(bx_src.getKP1()));
        unsigned char * ptr_dst = (unsigned char *)(&gd_dst.template get<prp>(bx_dst.getKP1()));
 
        copy_ndim_fast_selector<dim>::template call<sizeof(T)>(ptr_src,ptr_dst,sr,bx_src);
    }
};
 
template<unsigned int dim, typename T, unsigned int N1>
struct mp_funct_impl<dim,T[N1]>
{
    template<unsigned int prp, typename grid>
    static void process(const grid & gd_src,grid & gd_dst,
                        const Box<dim,size_t> & bx_src, const Box<grid::dims,size_t> & bx_dst)
    {
        unsigned int id[1];
 
        for (id[0] = 0 ; id[0] < N1 ; id[0]++)
        {
            striding<dim> sr = get_striding<dim,1,prp,grid>::get(gd_src,gd_dst,bx_src,id);
 
            unsigned char * ptr_src = (unsigned char *)(&gd_src.template get<prp>(bx_src.getKP1())[id[0]]);
            unsigned char * ptr_dst = (unsigned char *)(&gd_dst.template get<prp>(bx_dst.getKP1())[id[0]]);
 
            copy_ndim_fast_selector<dim>::template call<sizeof(T)>(ptr_src,ptr_dst,sr,bx_src);
        }
    }
};
 
 
template<unsigned int dim, typename T, unsigned int N1, unsigned int N2>
struct mp_funct_impl<dim,T[N1][N2]>
{
    template<unsigned int prp, typename grid>
    static void process(const grid & gd_src,grid & gd_dst,
                        const Box<dim,size_t> & bx_src, const Box<grid::dims,size_t> & bx_dst)
    {
        unsigned int id[2];
 
        for (id[1] = 0 ; id[1] < N1 ; id[1]++)
        {
            for (id[0] = 0 ; id[0] < N2 ; id[0]++)
            {
                striding<dim> sr = get_striding<dim,2,prp,grid>::get(gd_src,gd_dst,bx_src,id);
 
                unsigned char * ptr_src = (unsigned char *)(&gd_src.template get<prp>(bx_src.getKP1())[id[1]][id[0]]);
                unsigned char * ptr_dst = (unsigned char *)(&gd_dst.template get<prp>(bx_dst.getKP1())[id[1]][id[0]]);
 
                copy_ndim_fast_selector<dim>::template call<sizeof(T)>(ptr_src,ptr_dst,sr,bx_src);
            }
        }
    }
};
 
template<typename grid, typename ginfo>
class mp_funct
{
    const grid & gd_src;
    grid & gd_dst;
    ginfo & gs_src;
    ginfo & gs_dst;
 
    const Box<grid::dims,size_t> & bx_src;
    const Box<grid::dims,size_t> & bx_dst;
 
    grid_key_dx<grid::dims> (& cnt)[1];
 
public:
 
    mp_funct(const grid & gd_src,grid & gd_dst,
             ginfo & gs_src,ginfo & gs_dst,
             const Box<grid::dims,size_t> & bx_src , const Box<grid::dims,size_t> & bx_dst,
             grid_key_dx<grid::dims> (& cnt)[1])
    :gd_src(gd_src),gd_dst(gd_dst),gs_src(gs_src),gs_dst(gs_dst),bx_src(bx_src),bx_dst(bx_dst),cnt(cnt)
    {}
 
    template<typename T>
    inline void operator()(T& t)
    {
        typedef typename boost::mpl::at<typename grid::value_type::type,boost::mpl::int_<T::value>>::type prp_type;
 
        mp_funct_impl<grid::dims,prp_type>::template process<T::value>(gd_src,gd_dst,bx_src,bx_dst);
    }
};
 
template<bool is_inte,unsigned int dim,typename grid, typename ginfo>
struct copy_grid_fast_layout_switch
{
    static void copy(const ginfo & gs_src,
               const ginfo & gs_dst,
               const Box<dim,size_t> & bx_src,
               const Box<dim,size_t> & bx_dst,
               const grid & gd_src,
               grid & gd_dst,
               grid_key_dx<dim> (& cnt)[1])
    {
        mp_funct_impl<dim,typename grid::value_type>::template process<0>(gd_src,gd_dst,bx_src,bx_dst);
 
//      copy_ndim_fast_selector<grid::dims>::template call<0>(gs_src,gs_dst,bx_src,bx_dst,
//                                                            gd_src,gd_dst,cnt);
    }
};
 
template<unsigned int dim, typename grid, typename ginfo>
struct copy_grid_fast_layout_switch<true,dim,grid,ginfo>
{
    static void copy(const ginfo & gs_src,
               const ginfo & gs_dst,
               const Box<dim,size_t> & bx_src,
               const Box<dim,size_t> & bx_dst,
               const grid & gd_src,
               grid & gd_dst,
               grid_key_dx<dim> (& cnt)[1])
    {
        mp_funct<grid,ginfo> mpf(gd_src,gd_dst,gs_src,gs_dst,bx_src,bx_dst,cnt);
 
        boost::mpl::for_each_ref<boost::mpl::range_c<int,0,grid::value_type::max_prop>>(mpf);
    }
};
 
template<typename grid, typename ginfo>
struct copy_grid_fast<false,3,grid,ginfo>
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<3,size_t> & bx_src,
                   const Box<3,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<3> (& cnt)[1] )
    {
        copy_grid_fast_layout_switch<is_layout_inte<typename grid::layout_base_>::value,3,grid,ginfo>::copy(gs_src,gs_dst,bx_src,bx_dst,gd_src,gd_dst,cnt);
    }
};
 
 
 
 
 
template<typename grid, typename ginfo>
struct copy_grid_fast<false,2,grid,ginfo>
{
    static void copy(ginfo & gs_src,
                   ginfo & gs_dst,
                   const Box<2,size_t> & bx_src,
                   const Box<2,size_t> & bx_dst,
                   const grid & gd_src,
                   grid & gd_dst,
                   grid_key_dx<2> (& cnt)[1] )
    {
        copy_grid_fast_layout_switch<is_layout_inte<typename grid::layout_base_>::value,2,grid,ginfo>::copy(gs_src,gs_dst,bx_src,bx_dst,gd_src,gd_dst,cnt);
    }
};
 
 
 
template <bool is_complex,
          unsigned int dim,
          typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename dtype,
          int ... prp>
struct pack_with_iterator
{
    static void pack(grid & gr, it & sub_it, dtype & dest)
    {
        size_t id = 0;
 
        // Packing the information
        while (sub_it.isNext())
        {
            // Copy only the selected properties
            object_si_d<encap_src,encap_dst,OBJ_ENCAP,prp...>(gr.get_o(sub_it.get()),dest.get(id));
 
            ++id;
            ++sub_it;
        }
    }
};
 
 
 
 
template <typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename dtype,
          int ... prp>
struct pack_with_iterator<true,3,grid,encap_src,encap_dst,boost_vct,it,dtype,prp...>
{
    static void pack(grid & gr, it & sub_it, dtype & dest)
    {
        size_t id = 0;
 
        size_t lin_src = 0;
 
        auto & gs_src = gr.getGrid();
        grid_key_dx<3> start = sub_it.getStart();
        grid_key_dx<3> stop = sub_it.getStop();
 
        lin_src += start.get(2) * gs_src.size_s(1);
        for (long int i = start.get(2) ; i <= stop.get(2) ; i++)
        {
            lin_src += start.get(1) * gs_src.size_s(0);
            for (long int j = start.get(1) ; j <= stop.get(1) ; j++)
            {
                lin_src += start.get(0);
                for (long int k = start.get(0) ; k <= stop.get(0) ; k++)
                {
                    // Copy only the selected properties
                    object_si_d<encap_src,encap_dst,OBJ_ENCAP,prp...>(gr.get_o(lin_src),dest.get(id));
 
                    ++id;
                    ++lin_src;
                }
                lin_src -= stop.get(0) + 1;
                lin_src += gs_src.size_s(0);
            }
            lin_src -= (stop.get(1) + 1)*gs_src.size_s(0);
            lin_src += gs_src.size_s(1);
        }
    }
};
 
template <typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename dtype,
          int ... prp>
struct pack_with_iterator<true,2,grid,encap_src,encap_dst,boost_vct,it,dtype,prp...>
{
    static void pack(grid & gr, it & sub_it, dtype & dest)
    {
        size_t id = 0;
 
        size_t lin_src = 0;
 
        auto & gs_src = gr.getGrid();
        grid_key_dx<2> start = sub_it.getStart();
        grid_key_dx<2> stop = sub_it.getStop();
 
        lin_src += start.get(1) * gs_src.size_s(0);
        for (long int j = start.get(1) ; j <= stop.get(1) ; j++)
        {
            lin_src += start.get(0);
            for (long int k = start.get(0) ; k <= stop.get(0) ; k++)
            {
                // Copy only the selected properties
                object_si_d<encap_src,encap_dst,OBJ_ENCAP,prp...>(gr.get_o(lin_src),dest.get(id));
 
                ++id;
                ++lin_src;
            }
            lin_src -= stop.get(0) + 1;
            lin_src += gs_src.size_s(0);
        }
 
    }
};
 
 
template<unsigned int dim, int obj_byte, typename git, typename grid>
struct pack_with_iterator_longx
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x,
              size_t n_cpy)
    {
        std::cout << __FILE__ << ":" << __LINE__ << " critical error, we shoukd never be here" << std::endl;
    }
};
 
template<int obj_byte, typename git, typename grid>
struct pack_with_iterator_longx<2,obj_byte,git,grid>
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x,
              size_t n_cpy)
    {
        size_t tot_y = sub_it.getStop().get(1) - sub_it.getStart().get(1) + 1;
 
        for (size_t i = 0 ; i < tot_y ; i++)
        {
            memcpy(ptr_dest,ptr,n_cpy * obj_byte);
 
            ptr += stride_x;
            ptr_dest += n_cpy * obj_byte;
        }
    }
};
 
template<int obj_byte, typename git, typename grid>
struct pack_with_iterator_longx<3,obj_byte,git,grid>
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x,
              size_t n_cpy)
    {
        size_t tot_y = sub_it.getStop().get(1) - sub_it.getStart().get(1) + 1;
        size_t tot_z = sub_it.getStop().get(2) - sub_it.getStart().get(2) + 1;
 
        grid_key_dx<3> zero;
        zero.zero();
        grid_key_dx<3> one = zero;
        one.set_d(2,1);
 
        unsigned char * ptr_final = (unsigned char *)&(gr.template get<0>(one));
        unsigned char * ptr_start = (unsigned char *)&(gr.template get<0>(zero));
 
        size_t stride_y = ptr_final - ptr_start;
 
        for (size_t i = 0 ; i < tot_z ; i++)
        {
            for (size_t i = 0 ; i < tot_y ; i++)
            {
                memcpy(ptr_dest,ptr,n_cpy * obj_byte);
 
                ptr += stride_x;
                ptr_dest += n_cpy * obj_byte;
            }
            ptr += stride_y - tot_y*stride_x;
        }
    }
};
 
 
 
template<unsigned int dim, unsigned int n_cpy ,int obj_byte, typename git, typename grid>
struct pack_with_iterator_shortx
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x)
    {
        std::cout << __FILE__ << ":" << __LINE__ << " critical error, we shoukd never be here" << std::endl;
    }
};
 
template<unsigned int n_cpy, int obj_byte, typename git, typename grid>
struct pack_with_iterator_shortx<2,n_cpy,obj_byte,git,grid>
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x)
    {
        size_t tot_y = sub_it.getStop().get(1) - sub_it.getStart().get(1) + 1;
 
        for (size_t i = 0 ; i < tot_y ; i++)
        {
            __builtin_memcpy(ptr_dest,ptr,n_cpy * obj_byte);
 
            ptr += stride_x;
            ptr_dest += n_cpy * obj_byte;
        }
    }
};
 
template<unsigned int n_cpy, int obj_byte, typename git, typename grid>
struct pack_with_iterator_shortx<3,n_cpy,obj_byte,git,grid>
{
    static void pack(grid & gr,
              git & sub_it,
              unsigned char * ptr_dest,
              unsigned char * ptr,
              size_t stride_x)
    {
        size_t tot_y = sub_it.getStop().get(1) - sub_it.getStart().get(1) + 1;
        size_t tot_z = sub_it.getStop().get(2) - sub_it.getStart().get(2) + 1;
 
        grid_key_dx<3> zero;
        zero.zero();
        grid_key_dx<3> one = zero;
        one.set_d(2,1);
 
        unsigned char * ptr_final = (unsigned char *)(&gr.template get_unsafe<0>(one));
        unsigned char * ptr_start = (unsigned char *)(&gr.template get<0>(zero));
 
        size_t stride_y = ptr_final - ptr_start;
 
        for (size_t i = 0 ; i < tot_z ; i++)
        {
            for (size_t i = 0 ; i < tot_y ; i++)
            {
                __builtin_memcpy(ptr_dest,ptr,n_cpy * obj_byte);
 
                ptr += stride_x;
                ptr_dest += n_cpy * obj_byte;
            }
            ptr += stride_y - tot_y*stride_x;
        }
    }
};
 
 
template <unsigned int dim,
          typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename dtype,
          int ... prp>
struct pack_with_iterator<false,dim,grid,encap_src,encap_dst,boost_vct,it,dtype,prp...>
{
    static void pack(grid & gr, it & sub_it, dtype & dest)
    {
        // We do not have an optimized version for dimension different from 3 and 2
        if (dim == 1 || dim > 3)
        {
            pack_with_iterator<true,dim,grid,encap_src,encap_dst,boost_vct,it,dtype,prp...>::pack(gr,sub_it,dest);
            return;
        }
 
        // Sending property object
        typedef object<typename object_creator<
                                        boost_vct,
                                        prp...>::type
                      > prp_object;
 
 
        // fast strided copy
 
        grid_key_dx<dim> zero;
        zero.zero();
        grid_key_dx<dim> one = zero;
        one.set_d(1,1);
 
        unsigned char * ptr_final = (unsigned char *)(&gr.template get_unsafe<0>(one));
        unsigned char * ptr_start = (unsigned char *)(&gr.template get<0>(zero));
 
        size_t stride = ptr_final - ptr_start;
 
        size_t n_tot = 0;
        for (size_t i = 1 ; i < dim ; i++)
        {n_tot += sub_it.getStop().get(i) - sub_it.getStart().get(i) + 1;}
 
        size_t n_cpy = sub_it.getStop().get(0) - sub_it.getStart().get(0) + 1;
        unsigned char * ptr = (unsigned char *)&(gr.template get<first_variadic<prp...>::type::value>(sub_it.getStart()));
        unsigned char * ptr_dest = (unsigned char *)dest.getPointer();
 
        switch (n_cpy)
        {
        case 1:
            pack_with_iterator_shortx<dim,1,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 2:
            pack_with_iterator_shortx<dim,2,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 3:
            pack_with_iterator_shortx<dim,3,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 4:
            pack_with_iterator_shortx<dim,4,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 5:
            pack_with_iterator_shortx<dim,5,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 6:
            pack_with_iterator_shortx<dim,6,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 7:
            pack_with_iterator_shortx<dim,7,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        case 8:
            pack_with_iterator_shortx<dim,8,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride);
            break;
        default:
            pack_with_iterator_longx<dim,sizeof(prp_object),it,grid>::pack(gr,
                                                            sub_it,
                                                            ptr_dest,
                                                            ptr,
                                                            stride,
                                                            n_cpy);
 
        }
 
    }
};
 
template <typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename dtype,
          int ... prp>
struct pack_with_iterator<true,1,grid,encap_src,encap_dst,boost_vct,it,dtype,prp...>
{
    static void pack(grid & gr, it & sub_it, dtype & dest)
    {
        size_t id = 0;
 
        size_t lin_src = 0;
 
        auto & gs_src = gr.getGrid();
        grid_key_dx<1> start = sub_it.getStart();
        grid_key_dx<1> stop = sub_it.getStop();
 
 
        lin_src += start.get(0);
        for (long int k = start.get(0) ; k <= stop.get(0) ; k++)
        {
            // Copy only the selected properties
            object_si_d<encap_src,encap_dst,OBJ_ENCAP,prp...>(gr.get_o(lin_src),dest.get(id));
 
            ++id;
            ++lin_src;
        }
        lin_src -= stop.get(0) + 1;
        lin_src += gs_src.size_s(0);
 
    }
};
 
 
template <unsigned int dim,
          typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename stype,
          int ... prp>
struct unpack_with_iterator
{
    static void unpack(grid & gr, it & sub_it, stype & src)
    {
        size_t id = 0;
 
        // unpacking the information
        while (sub_it.isNext())
        {
 
            // Copy only the selected properties
            object_s_di<encap_src,encap_dst,OBJ_ENCAP,prp...>(src.get(id),gr.get_o(sub_it.get()));
 
            ++id;
            ++sub_it;
        }
    }
};
 
 
template <typename grid,
          typename encap_src,
          typename encap_dst,
          typename boost_vct,
          typename it,
          typename stype,
          int ... prp>
struct unpack_with_iterator<3,grid,
                            encap_src,
                            encap_dst,
                            boost_vct,
                            it,
                            stype,
                            prp ...>
{
    static void unpack(grid & gr, it & sub_it, stype & src)
    {
        size_t id = 0;
 
        size_t lin_dst = 0;
 
        auto & gs_dst = gr.getGrid();
        grid_key_dx<3> start = sub_it.getStart();
        grid_key_dx<3> stop = sub_it.getStop();
 
        // unpacking the information
 
        lin_dst += start.get(2) * gs_dst.size_s(1);
        for (long int i = start.get(2) ; i <= stop.get(2) ; i++)
        {
            lin_dst += start.get(1) * gs_dst.size_s(0);
            for (long int j = start.get(1) ; j <= stop.get(1) ; j++)
            {
                lin_dst += start.get(0);
                for (long int k = start.get(0) ; k <= stop.get(0) ; k++)
                {
                    // Copy only the selected properties
                    object_s_di<encap_src,encap_dst,OBJ_ENCAP,prp...>(src.get(id),gr.get_o(lin_dst));
 
                    ++id;
                    ++lin_dst;
                }
                lin_dst -= stop.get(0) + 1;
                lin_dst += gs_dst.size_s(0);
            }
            lin_dst -= (stop.get(1) + 1)*gs_dst.size_s(0);
            lin_dst += gs_dst.size_s(1);
        }
    }
};
 
#endif /* OPENFPM_DATA_SRC_GRID_COPY_GRID_FAST_HPP_ */