include/gmx_parallel_3dfft.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2005
   5  * David van der Spoel, Erik Lindahl, University of Groningen.
   6  * Copyright (c) 2012, by the GROMACS development team, led by
   7  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   8  * others, as listed in the AUTHORS file in the top-level source
   9  * directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37
  38 #ifndef _gmx_parallel_3dfft_h_
  39 #define _gmx_parallel_3dfft_h_
  40
  41
  42 #include "types/simple.h"
  43 #include "types/commrec.h"
  44 #include "gmxcomplex.h"
  45 #include "gmx_fft.h"
  46
  47 typedef struct gmx_parallel_3dfft *
  48 gmx_parallel_3dfft_t;
  49
  50
  51
  52 /*! \brief Initialize parallel MPI-based 3D-FFT.
  53  *
  54  *  This routine performs real-to-complex and complex-to-real parallel 3D FFTs,
  55  *  but not complex-to-complex.
  56  *
  57  *  The routine is optimized for small-to-medium size FFTs used for PME and
  58  *  PPPM algorithms, and do allocate extra workspace whenever it might improve
  59  *  performance.
  60  *
  61  *  \param pfft_setup     Pointer to parallel 3dfft setup structure, previously
  62  *                        allocated or with automatic storage.
  63  *  \param ngridx         Global number of grid cells in the x direction. Must be
  64  *                        divisible by the number of nodes.
  65  *  \param ngridy         Global number of grid cells in the y direction. Must be
  66  *                        divisible by the number of nodes.
  67  *  \param ngridz         Global number of grid cells in the z direction.
  68  *  \param node2slab      Node id to slab index array, can be NULL.
  69  *  \param slab2grid_x    Slab index to grid_x array (nnodes+1), can be NULL.
  70  *  \param comm           MPI communicator, must have been initialized.
  71  *  \param bReproducible  Try to avoid FFT timing optimizations and other stuff
  72  *                        that could make results differ for two runs with
  73  *                        identical input (reproducibility for debugging).
  74  *  \param nthreads       Run in parallel using n threads
  75  *
  76  *  \return 0 or a standard error code.
  77  */
  78 int
  79 gmx_parallel_3dfft_init   (gmx_parallel_3dfft_t *    pfft_setup,
  80                            ivec                      ndata,
  81                            real **                   real_data,
  82                            t_complex **              complex_data,
  83                            MPI_Comm                  comm[2],
  84                            int *                     slab2index_major,
  85                            int *                     slab2index_minor,
  86                            gmx_bool                  bReproducible,
  87                            int                       nthreads);
  88
  89
  90
  91
  92
  93 /*! \brief Get direct space grid index limits
  94  */
  95 int
  96 gmx_parallel_3dfft_real_limits(gmx_parallel_3dfft_t      pfft_setup,
  97                                ivec                      local_ndata,
  98                                ivec                      local_offset,
  99                                ivec                      local_size);
 100
 101
 102 /*! \brief Get reciprocal space grid index limits
 103  */
 104 int
 105 gmx_parallel_3dfft_complex_limits(gmx_parallel_3dfft_t      pfft_setup,
 106                                   ivec                      complex_order,
 107                                   ivec                      local_ndata,
 108                                   ivec                      local_offset,
 109                                   ivec                      local_size);
 110
 111
 112 int
 113 gmx_parallel_3dfft_execute(gmx_parallel_3dfft_t    pfft_setup,
 114                            enum gmx_fft_direction  dir,
 115                            void *                  in_data,
 116                            void *                  out_data,
 117                            int                     thread,
 118                            gmx_wallcycle_t         wcycle);
 119
 120
 121 /*! \brief Release all data in parallel fft setup
 122  *
 123  *  All temporary storage and FFT plans are released. The structure itself
 124  *  is not released, but the contents is invalid after this call.
 125  *
 126  *  \param pfft_setup Parallel 3dfft setup.
 127  *  \param in_data    Input data.
 128  *  \param out_data   Output data.
 129  *  \param thread     Thread index of the calling thread, i.e. index to the part
 130  *                    of the data operated on last by the calling thread. This
 131  *                    is needed to start the FFT without an OpenMP barrier.
 132  *  \param wcycle     Wall cycle counters.
 133  *
 134  *  \return 0 or a standard error code.
 135  */
 136 int
 137 gmx_parallel_3dfft_destroy(gmx_parallel_3dfft_t    pfft_setup);
 138
 139
 140
 141
 142 #endif /* _gmx_parallel_3dfft_h_ */
 143