src/external/thread_mpi/src/gather.c

   1 /*
   2    This source code file is part of thread_mpi.
   3    Written by Sander Pronk, Erik Lindahl, and possibly others.
   4
   5    Copyright (c) 2009,2016, Sander Pronk, Erik Lindahl.
   6    All rights reserved.
   7
   8    Redistribution and use in source and binary forms, with or without
   9    modification, are permitted provided that the following conditions are met:
  10    1) Redistributions of source code must retain the above copyright
  11    notice, this list of conditions and the following disclaimer.
  12    2) Redistributions in binary form must reproduce the above copyright
  13    notice, this list of conditions and the following disclaimer in the
  14    documentation and/or other materials provided with the distribution.
  15    3) Neither the name of the copyright holders nor the
  16    names of its contributors may be used to endorse or promote products
  17    derived from this software without specific prior written permission.
  18
  19    THIS SOFTWARE IS PROVIDED BY US ''AS IS'' AND ANY
  20    EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  21    WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  22    DISCLAIMED. IN NO EVENT SHALL WE BE LIABLE FOR ANY
  23    DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  24    (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  25    LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
  26    ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  27    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  28    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  29
  30    If you want to redistribute modifications, please consider that
  31    scientific software is very special. Version control is crucial -
  32    bugs must be traceable. We will be happy to consider code for
  33    inclusion in the official distribution, but derived work should not
  34    be called official thread_mpi. Details are found in the README & COPYING
  35    files.
  36  */
  37
  38 #ifdef HAVE_TMPI_CONFIG_H
  39 #include "tmpi_config.h"
  40 #endif
  41
  42 #ifdef HAVE_CONFIG_H
  43 #include "config.h"
  44 #endif
  45
  46
  47 #ifdef HAVE_UNISTD_H
  48 #include <unistd.h>
  49 #endif
  50
  51 #include <errno.h>
  52 #include <stdlib.h>
  53 #include <stdio.h>
  54 #include <stdarg.h>
  55 #include <string.h>
  56
  57 #include "impl.h"
  58 #include "collective.h"
  59
  60
  61 int tMPI_Gather(const void* sendbuf, int sendcount, tMPI_Datatype sendtype,
  62                 void* recvbuf, int recvcount, tMPI_Datatype recvtype,
  63                 int root, tMPI_Comm comm)
  64 {
  65     int                 synct;
  66     struct coll_env    *cev;
  67     int                 myrank;
  68     int                 ret = TMPI_SUCCESS;
  69     struct tmpi_thread *cur = tMPI_Get_current();
  70
  71 #ifdef TMPI_PROFILE
  72     tMPI_Profile_count_start(cur);
  73 #endif
  74 #ifdef TMPI_TRACE
  75     tMPI_Trace_print("tMPI_Gather(%p, %d, %p, %p, %d, %p, %d, %p)",
  76                      sendbuf, sendcount, sendtype,
  77                      recvbuf, recvcount, recvtype, root, comm);
  78 #endif
  79
  80     if (!comm)
  81     {
  82         return tMPI_Error(TMPI_COMM_WORLD, TMPI_ERR_COMM);
  83     }
  84     myrank = tMPI_Comm_seek_rank(comm, cur);
  85
  86     /* we increase our counter, and determine which coll_env we get */
  87     cev = tMPI_Get_cev(comm, myrank, &synct);
  88
  89     if (myrank == root)
  90     {
  91         int i;
  92         int n_remaining = comm->grp.N-1;
  93         /* do root transfer */
  94         if (sendbuf != TMPI_IN_PLACE)
  95         {
  96             tMPI_Coll_root_xfer(comm, sendtype, recvtype,
  97                                 sendtype->size*sendcount,
  98                                 recvtype->size*recvcount,
  99                                 (void*)sendbuf,
 100                                 (char*)recvbuf+myrank*recvcount*recvtype->size,
 101                                 &ret);
 102         }
 103         for (i = 0; i < comm->grp.N; i++)
 104         {
 105             cev->met[myrank].read_data[i] = FALSE;
 106         }
 107         cev->met[myrank].read_data[myrank] = TRUE;
 108
 109         /* wait for data availability as long as there are xfers to be done */
 110         while (n_remaining > 0)
 111         {
 112 #if defined(TMPI_PROFILE) && defined(TMPI_CYCLE_COUNT)
 113             tMPI_Profile_wait_start(cur);
 114 #endif
 115             tMPI_Event_wait( &(cev->met[myrank]).recv_ev );
 116 #if defined(TMPI_PROFILE) && defined(TMPI_CYCLE_COUNT)
 117             tMPI_Profile_wait_stop(cur, TMPIWAIT_Coll_recv);
 118 #endif
 119             /* now check all of them */
 120             for (i = 0; i < comm->grp.N; i++)
 121             {
 122                 if (!cev->met[myrank].read_data[i] &&
 123                     (tMPI_Atomic_get(&(cev->met[i].current_sync)) == synct))
 124                 {
 125                     tMPI_Mult_recv(comm, cev, i, 0, TMPI_GATHER_TAG, recvtype,
 126                                    recvcount*recvtype->size,
 127                                    (char*)recvbuf+i*recvcount*recvtype->size,
 128                                    &ret);
 129                     tMPI_Event_process( &(cev->met[myrank]).recv_ev, 1);
 130                     if (ret != TMPI_SUCCESS)
 131                     {
 132                         return ret;
 133                     }
 134                     cev->met[myrank].read_data[i] = TRUE;
 135                     n_remaining--;
 136                 }
 137             }
 138         }
 139     }
 140     else
 141     {
 142         if (!sendbuf && sendcount > 0) /* don't do pointer arithmetic on a NULL ptr */
 143         {
 144             return tMPI_Error(comm, TMPI_ERR_BUF);
 145         }
 146
 147         /* first set up the data just to root. */
 148         ret = tMPI_Post_multi(cev, myrank, 0, TMPI_GATHER_TAG, sendtype,
 149                               sendcount*sendtype->size, (void*)sendbuf, 1, synct, root);
 150         if (ret != TMPI_SUCCESS)
 151         {
 152             return ret;
 153         }
 154         /* and wait until root is done copying */
 155         tMPI_Wait_for_others(cev, myrank);
 156     }
 157 #ifdef TMPI_PROFILE
 158     tMPI_Profile_count_stop(cur, TMPIFN_Gather);
 159 #endif
 160     return ret;
 161 }
 162
 163
 164
 165
 166
 167
 168 int tMPI_Gatherv(const void* sendbuf, int sendcount, tMPI_Datatype sendtype,
 169                  void* recvbuf, int *recvcounts, int *displs,
 170                  tMPI_Datatype recvtype, int root, tMPI_Comm comm)
 171 {
 172     int                 synct;
 173     struct coll_env    *cev;
 174     int                 myrank;
 175     int                 ret = TMPI_SUCCESS;
 176     struct tmpi_thread *cur = tMPI_Get_current();
 177
 178 #ifdef TMPI_PROFILE
 179     tMPI_Profile_count_start(cur);
 180 #endif
 181 #ifdef TMPI_TRACE
 182     tMPI_Trace_print("tMPI_Gatherv(%p, %d, %p, %p, %p, %p, %p, %d, %p)",
 183                      sendbuf, sendcount, sendtype, recvbuf,
 184                      recvcounts, displs, recvtype, root, comm);
 185 #endif
 186
 187     if (!comm)
 188     {
 189         return tMPI_Error(TMPI_COMM_WORLD, TMPI_ERR_COMM);
 190     }
 191     myrank = tMPI_Comm_seek_rank(comm, cur);
 192
 193     /* we increase our counter, and determine which coll_env we get */
 194     cev = tMPI_Get_cev(comm, myrank, &synct);
 195
 196     if (myrank == root)
 197     {
 198         int i;
 199         int n_remaining = comm->grp.N-1;
 200         /* do root transfer */
 201         if (sendbuf != TMPI_IN_PLACE)
 202         {
 203             tMPI_Coll_root_xfer(comm, sendtype, recvtype,
 204                                 sendtype->size*sendcount,
 205                                 recvtype->size*recvcounts[myrank],
 206                                 (void*)sendbuf,
 207                                 (char*)recvbuf+displs[myrank]*recvtype->size,
 208                                 &ret);
 209         }
 210         for (i = 0; i < comm->grp.N; i++)
 211         {
 212             cev->met[myrank].read_data[i] = FALSE;
 213         }
 214         cev->met[myrank].read_data[myrank] = TRUE;
 215
 216         /* wait for data availability as long as there are xfers to be done */
 217         while (n_remaining > 0)
 218         {
 219 #if defined(TMPI_PROFILE) && defined(TMPI_CYCLE_COUNT)
 220             tMPI_Profile_wait_start(cur);
 221 #endif
 222             tMPI_Event_wait( &(cev->met[myrank]).recv_ev );
 223 #if defined(TMPI_PROFILE) && defined(TMPI_CYCLE_COUNT)
 224             tMPI_Profile_wait_stop(cur, TMPIWAIT_Coll_recv);
 225 #endif
 226             for (i = 0; i < comm->grp.N; i++)
 227             {
 228                 if (!cev->met[myrank].read_data[i] &&
 229                     (tMPI_Atomic_get(&(cev->met[i].current_sync)) == synct) )
 230                 {
 231                     tMPI_Event_process( &(cev->met[myrank]).recv_ev, 1);
 232                     tMPI_Mult_recv(comm, cev, i, 0, TMPI_GATHERV_TAG, recvtype,
 233                                    recvcounts[i]*recvtype->size,
 234                                    (char*)recvbuf+displs[i]*recvtype->size,
 235                                    &ret);
 236                     if (ret != TMPI_SUCCESS)
 237                     {
 238                         return ret;
 239                     }
 240                     cev->met[myrank].read_data[i] = TRUE;
 241                     n_remaining--;
 242                 }
 243             }
 244         }
 245     }
 246     else
 247     {
 248         if (!sendbuf && sendcount > 0) /* don't do pointer arithmetic on a NULL ptr */
 249         {
 250             return tMPI_Error(comm, TMPI_ERR_BUF);
 251         }
 252
 253         /* first set up the data just to root. */
 254         ret = tMPI_Post_multi(cev, myrank, 0, TMPI_GATHERV_TAG, sendtype,
 255                               sendcount*sendtype->size, (void*)sendbuf, 1, synct, root);
 256         if (ret != TMPI_SUCCESS)
 257         {
 258             return ret;
 259         }
 260         /* and wait until root is done copying */
 261         tMPI_Wait_for_others(cev, myrank);
 262     }
 263
 264 #ifdef TMPI_PROFILE
 265     tMPI_Profile_count_stop(cur, TMPIFN_Gatherv);
 266 #endif
 267     return ret;
 268 }