src/gmxlib/main.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team,
   6  * check out http://www.gromacs.org for more information.
   7  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   8  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   9  * others, as listed in the AUTHORS file in the top-level source
  10  * directory and at http://www.gromacs.org.
  11  *
  12  * GROMACS is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public License
  14  * as published by the Free Software Foundation; either version 2.1
  15  * of the License, or (at your option) any later version.
  16  *
  17  * GROMACS is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with GROMACS; if not, see
  24  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  25  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  26  *
  27  * If you want to redistribute modifications to GROMACS, please
  28  * consider that scientific software is very special. Version
  29  * control is crucial - bugs must be traceable. We will be happy to
  30  * consider code for inclusion in the official distribution, but
  31  * derived work must not be called official GROMACS. Details are found
  32  * in the README & COPYING files - if they are missing, get the
  33  * official version at http://www.gromacs.org.
  34  *
  35  * To help us fund GROMACS development, we humbly ask that you cite
  36  * the research papers on the package. Check out http://www.gromacs.org.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <stdio.h>
  43 #include <stdlib.h>
  44 #include <string.h>
  45 #include <limits.h>
  46 #include <time.h>
  47
  48 #ifdef HAVE_SYS_TIME_H
  49 #include <sys/time.h>
  50 #endif
  51
  52 #ifdef HAVE_DIRECT_H
  53 /* windows-specific include for _chdir() */
  54 #include <direct.h>
  55 #endif
  56
  57
  58 #include "smalloc.h"
  59 #include "gmx_fatal.h"
  60 #include "network.h"
  61 #include "main.h"
  62 #include "macros.h"
  63 #include "futil.h"
  64 #include "filenm.h"
  65 #include "gmxfio.h"
  66 #include "string2.h"
  67 #include "copyrite.h"
  68
  69 #ifdef GMX_THREAD_MPI
  70 #include "thread_mpi.h"
  71 #endif
  72
  73 /* The source code in this file should be thread-safe.
  74          Please keep it that way. */
  75
  76
  77 #ifdef HAVE_UNISTD_H
  78 #include <unistd.h>
  79 #endif
  80
  81 #ifdef GMX_NATIVE_WINDOWS
  82 #include <process.h>
  83 #endif
  84
  85
  86 /* Portable version of ctime_r implemented in src/gmxlib/string2.c, but we do not want it declared in public installed headers */
  87 char *
  88 gmx_ctime_r(const time_t *clock, char *buf, int n);
  89
  90
  91 #define BUFSIZE 1024
  92
  93
  94 static void par_fn(char *base, int ftp, const t_commrec *cr,
  95                    gmx_bool bAppendSimId, gmx_bool bAppendNodeId,
  96                    char buf[], int bufsize)
  97 {
  98     int n;
  99
 100     if ((size_t)bufsize < (strlen(base)+10))
 101     {
 102         gmx_mem("Character buffer too small!");
 103     }
 104
 105     /* Copy to buf, and strip extension */
 106     strcpy(buf, base);
 107     buf[strlen(base) - strlen(ftp2ext(fn2ftp(base))) - 1] = '\0';
 108
 109     if (bAppendSimId)
 110     {
 111         sprintf(buf+strlen(buf), "%d", cr->ms->sim);
 112     }
 113     if (bAppendNodeId)
 114     {
 115         strcat(buf, "_node");
 116         sprintf(buf+strlen(buf), "%d", cr->nodeid);
 117     }
 118     strcat(buf, ".");
 119
 120     /* Add extension again */
 121     strcat(buf, (ftp == efTPX) ? "tpr" : (ftp == efEDR) ? "edr" : ftp2ext(ftp));
 122     if (debug)
 123     {
 124         fprintf(debug, "node %d par_fn '%s'\n", cr->nodeid, buf);
 125         if (fn2ftp(buf) == efLOG)
 126         {
 127             fprintf(debug, "log\n");
 128         }
 129     }
 130 }
 131
 132 void check_multi_int(FILE *log, const gmx_multisim_t *ms, int val,
 133                      const char *name,
 134                      gmx_bool bQuiet)
 135 {
 136     int     *ibuf, p;
 137     gmx_bool bCompatible;
 138
 139     if (NULL != log && !bQuiet)
 140     {
 141         fprintf(log, "Multi-checking %s ... ", name);
 142     }
 143
 144     if (ms == NULL)
 145     {
 146         gmx_fatal(FARGS,
 147                   "check_multi_int called with a NULL communication pointer");
 148     }
 149
 150     snew(ibuf, ms->nsim);
 151     ibuf[ms->sim] = val;
 152     gmx_sumi_sim(ms->nsim, ibuf, ms);
 153
 154     bCompatible = TRUE;
 155     for (p = 1; p < ms->nsim; p++)
 156     {
 157         bCompatible = bCompatible && (ibuf[p-1] == ibuf[p]);
 158     }
 159
 160     if (bCompatible)
 161     {
 162         if (NULL != log && !bQuiet)
 163         {
 164             fprintf(log, "OK\n");
 165         }
 166     }
 167     else
 168     {
 169         if (NULL != log)
 170         {
 171             fprintf(log, "\n%s is not equal for all subsystems\n", name);
 172             for (p = 0; p < ms->nsim; p++)
 173             {
 174                 fprintf(log, "  subsystem %d: %d\n", p, ibuf[p]);
 175             }
 176         }
 177         gmx_fatal(FARGS, "The %d subsystems are not compatible\n", ms->nsim);
 178     }
 179
 180     sfree(ibuf);
 181 }
 182
 183 void check_multi_large_int(FILE *log, const gmx_multisim_t *ms,
 184                            gmx_large_int_t val, const char *name,
 185                            gmx_bool bQuiet)
 186 {
 187     gmx_large_int_t  *ibuf;
 188     int               p;
 189     gmx_bool          bCompatible;
 190
 191     if (NULL != log && !bQuiet)
 192     {
 193         fprintf(log, "Multi-checking %s ... ", name);
 194     }
 195
 196     if (ms == NULL)
 197     {
 198         gmx_fatal(FARGS,
 199                   "check_multi_int called with a NULL communication pointer");
 200     }
 201
 202     snew(ibuf, ms->nsim);
 203     ibuf[ms->sim] = val;
 204     gmx_sumli_sim(ms->nsim, ibuf, ms);
 205
 206     bCompatible = TRUE;
 207     for (p = 1; p < ms->nsim; p++)
 208     {
 209         bCompatible = bCompatible && (ibuf[p-1] == ibuf[p]);
 210     }
 211
 212     if (bCompatible)
 213     {
 214         if (NULL != log && !bQuiet)
 215         {
 216             fprintf(log, "OK\n");
 217         }
 218     }
 219     else
 220     {
 221         if (NULL != log)
 222         {
 223             fprintf(log, "\n%s is not equal for all subsystems\n", name);
 224             for (p = 0; p < ms->nsim; p++)
 225             {
 226                 char strbuf[255];
 227                 /* first make the format string */
 228                 snprintf(strbuf, 255, "  subsystem %%d: %s\n",
 229                          gmx_large_int_pfmt);
 230                 fprintf(log, strbuf, p, ibuf[p]);
 231             }
 232         }
 233         gmx_fatal(FARGS, "The %d subsystems are not compatible\n", ms->nsim);
 234     }
 235
 236     sfree(ibuf);
 237 }
 238
 239
 240 int gmx_gethostname(char *name, size_t len)
 241 {
 242     if (len < 8)
 243     {
 244         gmx_incons("gmx_gethostname called with len<8");
 245     }
 246 #if defined(HAVE_UNISTD_H) && !defined(__native_client__)
 247     if (gethostname(name, len-1) != 0)
 248     {
 249         strncpy(name, "unknown", 8);
 250         return -1;
 251     }
 252     return 0;
 253 #else
 254     strncpy(name, "unknown", 8);
 255     return -1;
 256 #endif
 257 }
 258
 259
 260 void gmx_log_open(const char *lognm, const t_commrec *cr, gmx_bool bMasterOnly,
 261                   gmx_bool bAppendFiles, FILE** fplog)
 262 {
 263     int    len, testlen, pid;
 264     char   buf[256], host[256];
 265     time_t t;
 266     char   timebuf[STRLEN];
 267     FILE  *fp = *fplog;
 268     char  *tmpnm;
 269
 270     debug_gmx();
 271
 272     /* Communicate the filename for logfile */
 273     if (cr->nnodes > 1 && !bMasterOnly
 274 #ifdef GMX_THREAD_MPI
 275         /* With thread MPI the non-master log files are opened later
 276          * when the files names are already known on all nodes.
 277          */
 278         && FALSE
 279 #endif
 280         )
 281     {
 282         if (MASTER(cr))
 283         {
 284             len = strlen(lognm) + 1;
 285         }
 286         gmx_bcast(sizeof(len), &len, cr);
 287         if (!MASTER(cr))
 288         {
 289             snew(tmpnm, len+8);
 290         }
 291         else
 292         {
 293             tmpnm = gmx_strdup(lognm);
 294         }
 295         gmx_bcast(len*sizeof(*tmpnm), tmpnm, cr);
 296     }
 297     else
 298     {
 299         tmpnm = gmx_strdup(lognm);
 300     }
 301
 302     debug_gmx();
 303
 304     if (!bMasterOnly && !MASTER(cr))
 305     {
 306         /* Since log always ends with '.log' let's use this info */
 307         par_fn(tmpnm, efLOG, cr, FALSE, !bMasterOnly, buf, 255);
 308         fp = gmx_fio_fopen(buf, bAppendFiles ? "a+" : "w+" );
 309     }
 310     else if (!bAppendFiles)
 311     {
 312         fp = gmx_fio_fopen(tmpnm, bAppendFiles ? "a+" : "w+" );
 313     }
 314
 315     sfree(tmpnm);
 316
 317     gmx_fatal_set_log_file(fp);
 318
 319     /* Get some machine parameters */
 320     gmx_gethostname(host, 256);
 321
 322     time(&t);
 323
 324 #ifndef NO_GETPID
 325 #   ifdef GMX_NATIVE_WINDOWS
 326     pid = _getpid();
 327 #   else
 328     pid = getpid();
 329 #   endif
 330 #else
 331     pid = 0;
 332 #endif
 333
 334     if (bAppendFiles)
 335     {
 336         fprintf(fp,
 337                 "\n"
 338                 "\n"
 339                 "-----------------------------------------------------------\n"
 340                 "Restarting from checkpoint, appending to previous log file.\n"
 341                 "\n"
 342                 );
 343     }
 344
 345     gmx_ctime_r(&t, timebuf, STRLEN);
 346
 347     fprintf(fp,
 348             "Log file opened on %s"
 349             "Host: %s  pid: %d  nodeid: %d  nnodes:  %d\n",
 350             timebuf, host, pid, cr->nodeid, cr->nnodes);
 351     gmx_print_version_info(fp);
 352     fprintf(fp, "\n\n");
 353
 354     fflush(fp);
 355     debug_gmx();
 356
 357     *fplog = fp;
 358 }
 359
 360 void gmx_log_close(FILE *fp)
 361 {
 362     if (fp)
 363     {
 364         gmx_fatal_set_log_file(NULL);
 365         gmx_fio_fclose(fp);
 366     }
 367 }
 368
 369 static void comm_args(const t_commrec *cr, int *argc, char ***argv)
 370 {
 371     int i, len;
 372
 373     if (PAR(cr))
 374     {
 375         gmx_bcast(sizeof(*argc), argc, cr);
 376     }
 377
 378     if (!MASTER(cr))
 379     {
 380         snew(*argv, *argc+1);
 381     }
 382     if (debug)
 383     {
 384         fprintf(debug, "NODEID=%d argc=%d\n", cr->nodeid, *argc);
 385     }
 386     for (i = 0; (i < *argc); i++)
 387     {
 388         if (MASTER(cr))
 389         {
 390             len = strlen((*argv)[i])+1;
 391         }
 392         gmx_bcast(sizeof(len), &len, cr);
 393         if (!MASTER(cr))
 394         {
 395             snew((*argv)[i], len);
 396         }
 397         /*gmx_bcast(len*sizeof((*argv)[i][0]),(*argv)[i],cr);*/
 398         gmx_bcast(len*sizeof(char), (*argv)[i], cr);
 399     }
 400     debug_gmx();
 401 }
 402
 403 void init_multisystem(t_commrec *cr, int nsim, char **multidirs,
 404                       int nfile, const t_filenm fnm[], gmx_bool bParFn)
 405 {
 406     gmx_multisim_t *ms;
 407     int             nnodes, nnodpersim, sim, i, ftp;
 408     char            buf[256];
 409 #ifdef GMX_MPI
 410     MPI_Group       mpi_group_world;
 411 #endif
 412     int            *rank;
 413
 414 #ifndef GMX_MPI
 415     if (nsim > 1)
 416     {
 417         gmx_fatal(FARGS, "This binary is compiled without MPI support, can not do multiple simulations.");
 418     }
 419 #endif
 420
 421     nnodes  = cr->nnodes;
 422     if (nnodes % nsim != 0)
 423     {
 424         gmx_fatal(FARGS, "The number of nodes (%d) is not a multiple of the number of simulations (%d)", nnodes, nsim);
 425     }
 426
 427     nnodpersim = nnodes/nsim;
 428     sim        = cr->nodeid/nnodpersim;
 429
 430     if (debug)
 431     {
 432         fprintf(debug, "We have %d simulations, %d nodes per simulation, local simulation is %d\n", nsim, nnodpersim, sim);
 433     }
 434
 435     snew(ms, 1);
 436     cr->ms   = ms;
 437     ms->nsim = nsim;
 438     ms->sim  = sim;
 439 #ifdef GMX_MPI
 440     /* Create a communicator for the master nodes */
 441     snew(rank, ms->nsim);
 442     for (i = 0; i < ms->nsim; i++)
 443     {
 444         rank[i] = i*nnodpersim;
 445     }
 446     MPI_Comm_group(MPI_COMM_WORLD, &mpi_group_world);
 447     MPI_Group_incl(mpi_group_world, nsim, rank, &ms->mpi_group_masters);
 448     sfree(rank);
 449     MPI_Comm_create(MPI_COMM_WORLD, ms->mpi_group_masters,
 450                     &ms->mpi_comm_masters);
 451
 452 #if !defined(GMX_THREAD_MPI) && !defined(MPI_IN_PLACE_EXISTS)
 453     /* initialize the MPI_IN_PLACE replacement buffers */
 454     snew(ms->mpb, 1);
 455     ms->mpb->ibuf        = NULL;
 456     ms->mpb->libuf       = NULL;
 457     ms->mpb->fbuf        = NULL;
 458     ms->mpb->dbuf        = NULL;
 459     ms->mpb->ibuf_alloc  = 0;
 460     ms->mpb->libuf_alloc = 0;
 461     ms->mpb->fbuf_alloc  = 0;
 462     ms->mpb->dbuf_alloc  = 0;
 463 #endif
 464
 465 #endif
 466
 467     /* Reduce the intra-simulation communication */
 468     cr->sim_nodeid = cr->nodeid % nnodpersim;
 469     cr->nnodes     = nnodpersim;
 470 #ifdef GMX_MPI
 471     MPI_Comm_split(MPI_COMM_WORLD, sim, cr->sim_nodeid, &cr->mpi_comm_mysim);
 472     cr->mpi_comm_mygroup = cr->mpi_comm_mysim;
 473     cr->nodeid           = cr->sim_nodeid;
 474 #endif
 475
 476     if (debug)
 477     {
 478         fprintf(debug, "This is simulation %d", cr->ms->sim);
 479         if (PAR(cr))
 480         {
 481             fprintf(debug, ", local number of nodes %d, local nodeid %d",
 482                     cr->nnodes, cr->sim_nodeid);
 483         }
 484         fprintf(debug, "\n\n");
 485     }
 486
 487     if (multidirs)
 488     {
 489         int ret;
 490         if (debug)
 491         {
 492             fprintf(debug, "Changing to directory %s\n", multidirs[cr->ms->sim]);
 493         }
 494         if (chdir(multidirs[cr->ms->sim]) != 0)
 495         {
 496             gmx_fatal(FARGS, "Couldn't change directory to %s: %s",
 497                       multidirs[cr->ms->sim],
 498                       strerror(errno));
 499         }
 500     }
 501     else if (bParFn)
 502     {
 503         /* Patch output and tpx, cpt and rerun input file names */
 504         for (i = 0; (i < nfile); i++)
 505         {
 506             /* Because of possible multiple extensions per type we must look
 507              * at the actual file name
 508              */
 509             if (is_output(&fnm[i]) ||
 510                 fnm[i].ftp == efTPX || fnm[i].ftp == efCPT ||
 511                 strcmp(fnm[i].opt, "-rerun") == 0)
 512             {
 513                 ftp = fn2ftp(fnm[i].fns[0]);
 514                 par_fn(fnm[i].fns[0], ftp, cr, TRUE, FALSE, buf, 255);
 515                 sfree(fnm[i].fns[0]);
 516                 fnm[i].fns[0] = gmx_strdup(buf);
 517             }
 518         }
 519     }
 520 }
 521
 522 t_commrec *init_par(int *argc, char ***argv_ptr)
 523 {
 524     t_commrec    *cr;
 525     char        **argv;
 526     int           i;
 527     gmx_bool      pe = FALSE;
 528
 529     snew(cr, 1);
 530
 531     argv = argv_ptr ? *argv_ptr : NULL;
 532
 533 #if defined GMX_MPI && !defined GMX_THREAD_MPI
 534     cr->sim_nodeid = gmx_setup(argc, argv, &cr->nnodes);
 535
 536     if (!PAR(cr) && (cr->sim_nodeid != 0))
 537     {
 538         gmx_comm("(!PAR(cr) && (cr->sim_nodeid != 0))");
 539     }
 540
 541     cr->mpi_comm_mysim   = MPI_COMM_WORLD;
 542     cr->mpi_comm_mygroup = cr->mpi_comm_mysim;
 543 #else
 544     /* These should never be accessed */
 545     cr->mpi_comm_mysim   = NULL;
 546     cr->mpi_comm_mygroup = NULL;
 547     cr->nnodes           = 1;
 548     cr->sim_nodeid       = 0;
 549 #endif
 550
 551     cr->nodeid = cr->sim_nodeid;
 552
 553     cr->duty = (DUTY_PP | DUTY_PME);
 554
 555     /* Communicate arguments if parallel */
 556 #ifndef GMX_THREAD_MPI
 557     if (PAR(cr))
 558     {
 559         comm_args(cr, argc, argv_ptr);
 560     }
 561 #endif /* GMX_THREAD_MPI */
 562
 563 #ifdef GMX_MPI
 564 #if !defined(GMX_THREAD_MPI) && !defined(MPI_IN_PLACE_EXISTS)
 565     /* initialize the MPI_IN_PLACE replacement buffers */
 566     snew(cr->mpb, 1);
 567     cr->mpb->ibuf        = NULL;
 568     cr->mpb->libuf       = NULL;
 569     cr->mpb->fbuf        = NULL;
 570     cr->mpb->dbuf        = NULL;
 571     cr->mpb->ibuf_alloc  = 0;
 572     cr->mpb->libuf_alloc = 0;
 573     cr->mpb->fbuf_alloc  = 0;
 574     cr->mpb->dbuf_alloc  = 0;
 575 #endif
 576 #endif
 577
 578     return cr;
 579 }
 580
 581 t_commrec *init_par_threads(const t_commrec *cro)
 582 {
 583 #ifdef GMX_THREAD_MPI
 584     int        initialized;
 585     t_commrec *cr;
 586
 587     /* make a thread-specific commrec */
 588     snew(cr, 1);
 589     /* now copy the whole thing, so settings like the number of PME nodes
 590        get propagated. */
 591     *cr = *cro;
 592
 593     /* and we start setting our own thread-specific values for things */
 594     MPI_Initialized(&initialized);
 595     if (!initialized)
 596     {
 597         gmx_comm("Initializing threads without comm");
 598     }
 599     /* once threads will be used together with MPI, we'll
 600        fill the cr structure with distinct data here. This might even work: */
 601     cr->sim_nodeid = gmx_setup(0, NULL, &cr->nnodes);
 602
 603     cr->mpi_comm_mysim   = MPI_COMM_WORLD;
 604     cr->mpi_comm_mygroup = cr->mpi_comm_mysim;
 605     cr->nodeid           = cr->sim_nodeid;
 606     cr->duty             = (DUTY_PP | DUTY_PME);
 607
 608     return cr;
 609 #else
 610     return NULL;
 611 #endif
 612 }