src/kernel/runner.c

   1 /* -*- mode: c; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; c-file-style: "stroustrup"; -*-
   2  *
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  *          GROningen MAchine for Chemical Simulations
   9  *
  10  *                        VERSION 3.2.0
  11  * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
  12  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
  13  * Copyright (c) 2001-2004, The GROMACS development team,
  14  * check out http://www.gromacs.org for more information.
  15
  16  * This program is free software; you can redistribute it and/or
  17  * modify it under the terms of the GNU General Public License
  18  * as published by the Free Software Foundation; either version 2
  19  * of the License, or (at your option) any later version.
  20  *
  21  * If you want to redistribute modifications, please consider that
  22  * scientific software is very special. Version control is crucial -
  23  * bugs must be traceable. We will be happy to consider code for
  24  * inclusion in the official distribution, but derived work must not
  25  * be called official GROMACS. Details are found in the README & COPYING
  26  * files - if they are missing, get the official version at www.gromacs.org.
  27  *
  28  * To help us fund GROMACS development, we humbly ask that you cite
  29  * the papers on the package - you can find them in the top README file.
  30  *
  31  * For more info, check our website at http://www.gromacs.org
  32  *
  33  * And Hey:
  34  * Gallium Rubidium Oxygen Manganese Argon Carbon Silicon
  35  */
  36 #ifdef HAVE_CONFIG_H
  37 #include <config.h>
  38 #endif
  39
  40 #include <signal.h>
  41 #include <stdlib.h>
  42
  43 #if ((defined WIN32 || defined _WIN32 || defined WIN64 || defined _WIN64) && !defined __CYGWIN__ && !defined __CYGWIN32__)
  44 /* _isnan() */
  45 #include <float.h>
  46 #endif
  47
  48 #include "typedefs.h"
  49 #include "smalloc.h"
  50 #include "sysstuff.h"
  51 #include "statutil.h"
  52 #include "mdrun.h"
  53 #include "network.h"
  54 #include "pull.h"
  55 #include "names.h"
  56 #include "disre.h"
  57 #include "orires.h"
  58 #include "dihre.h"
  59 #include "pppm.h"
  60 #include "pme.h"
  61 #include "mdatoms.h"
  62 #include "repl_ex.h"
  63 #include "qmmm.h"
  64 #include "mpelogging.h"
  65 #include "domdec.h"
  66 #include "partdec.h"
  67 #include "coulomb.h"
  68 #include "constr.h"
  69 #include "mvdata.h"
  70 #include "checkpoint.h"
  71 #include "mtop_util.h"
  72 #include "sighandler.h"
  73 #include "tpxio.h"
  74 #include "txtdump.h"
  75 #include "pull_rotation.h"
  76 #include "md_openmm.h"
  77
  78 #ifdef GMX_LIB_MPI
  79 #include <mpi.h>
  80 #endif
  81 #ifdef GMX_THREADS
  82 #include "tmpi.h"
  83 #endif
  84
  85 #ifdef GMX_FAHCORE
  86 #include "corewrap.h"
  87 #endif
  88
  89 #ifdef GMX_OPENMM
  90 #include "md_openmm.h"
  91 #endif
  92
  93
  94 typedef struct {
  95     gmx_integrator_t *func;
  96 } gmx_intp_t;
  97
  98 /* The array should match the eI array in include/types/enums.h */
  99 #ifdef GMX_OPENMM  /* FIXME do_md_openmm needs fixing */
 100 const gmx_intp_t integrator[eiNR] = { {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm}, {do_md_openmm},{do_md_openmm}};
 101 #else
 102 const gmx_intp_t integrator[eiNR] = { {do_md}, {do_steep}, {do_cg}, {do_md}, {do_md}, {do_nm}, {do_lbfgs}, {do_tpi}, {do_tpi}, {do_md}, {do_md},{do_md}};
 103 #endif
 104
 105 gmx_large_int_t     deform_init_init_step_tpx;
 106 matrix              deform_init_box_tpx;
 107 #ifdef GMX_THREADS
 108 tMPI_Thread_mutex_t deform_init_box_mutex=TMPI_THREAD_MUTEX_INITIALIZER;
 109 #endif
 110
 111
 112 #ifdef GMX_THREADS
 113 struct mdrunner_arglist
 114 {
 115     FILE *fplog;
 116     t_commrec *cr;
 117     int nfile;
 118     const t_filenm *fnm;
 119     output_env_t oenv;
 120     gmx_bool bVerbose;
 121     gmx_bool bCompact;
 122     int nstglobalcomm;
 123     ivec ddxyz;
 124     int dd_node_order;
 125     real rdd;
 126     real rconstr;
 127     const char *dddlb_opt;
 128     real dlb_scale;
 129     const char *ddcsx;
 130     const char *ddcsy;
 131     const char *ddcsz;
 132     int nstepout;
 133     int resetstep;
 134     int nmultisim;
 135     int repl_ex_nst;
 136     int repl_ex_seed;
 137     real pforce;
 138     real cpt_period;
 139     real max_hours;
 140     const char *deviceOptions;
 141     unsigned long Flags;
 142     int ret; /* return value */
 143 };
 144
 145
 146 /* The function used for spawning threads. Extracts the mdrunner()
 147    arguments from its one argument and calls mdrunner(), after making
 148    a commrec. */
 149 static void mdrunner_start_fn(void *arg)
 150 {
 151     struct mdrunner_arglist *mda=(struct mdrunner_arglist*)arg;
 152     struct mdrunner_arglist mc=*mda; /* copy the arg list to make sure
 153                                         that it's thread-local. This doesn't
 154                                         copy pointed-to items, of course,
 155                                         but those are all const. */
 156     t_commrec *cr;  /* we need a local version of this */
 157     FILE *fplog=NULL;
 158     t_filenm *fnm;
 159
 160     fnm = dup_tfn(mc.nfile, mc.fnm);
 161
 162     cr = init_par_threads(mc.cr);
 163
 164     if (MASTER(cr))
 165     {
 166         fplog=mc.fplog;
 167     }
 168
 169     mda->ret=mdrunner(cr->nnodes, fplog, cr, mc.nfile, fnm, mc.oenv,
 170                       mc.bVerbose, mc.bCompact, mc.nstglobalcomm,
 171                       mc.ddxyz, mc.dd_node_order, mc.rdd,
 172                       mc.rconstr, mc.dddlb_opt, mc.dlb_scale,
 173                       mc.ddcsx, mc.ddcsy, mc.ddcsz, mc.nstepout, mc.resetstep,
 174                       mc.nmultisim, mc.repl_ex_nst, mc.repl_ex_seed, mc.pforce,
 175                       mc.cpt_period, mc.max_hours, mc.deviceOptions, mc.Flags);
 176 }
 177
 178 /* called by mdrunner() to start a specific number of threads (including
 179    the main thread) for thread-parallel runs. This in turn calls mdrunner()
 180    for each thread.
 181    All options besides nthreads are the same as for mdrunner(). */
 182 static t_commrec *mdrunner_start_threads(int nthreads,
 183               FILE *fplog,t_commrec *cr,int nfile,
 184               const t_filenm fnm[], const output_env_t oenv, gmx_bool bVerbose,
 185               gmx_bool bCompact, int nstglobalcomm,
 186               ivec ddxyz,int dd_node_order,real rdd,real rconstr,
 187               const char *dddlb_opt,real dlb_scale,
 188               const char *ddcsx,const char *ddcsy,const char *ddcsz,
 189               int nstepout,int resetstep,int nmultisim,int repl_ex_nst,
 190               int repl_ex_seed, real pforce,real cpt_period, real max_hours,
 191               const char *deviceOptions, unsigned long Flags)
 192 {
 193     int ret;
 194     struct mdrunner_arglist *mda;
 195     t_commrec *crn; /* the new commrec */
 196     t_filenm *fnmn;
 197
 198     /* first check whether we even need to start tMPI */
 199     if (nthreads<2)
 200         return cr;
 201
 202     /* a few small, one-time, almost unavoidable memory leaks: */
 203     snew(mda,1);
 204     fnmn=dup_tfn(nfile, fnm);
 205
 206     /* fill the data structure to pass as void pointer to thread start fn */
 207     mda->fplog=fplog;
 208     mda->cr=cr;
 209     mda->nfile=nfile;
 210     mda->fnm=fnmn;
 211     mda->oenv=oenv;
 212     mda->bVerbose=bVerbose;
 213     mda->bCompact=bCompact;
 214     mda->nstglobalcomm=nstglobalcomm;
 215     mda->ddxyz[XX]=ddxyz[XX];
 216     mda->ddxyz[YY]=ddxyz[YY];
 217     mda->ddxyz[ZZ]=ddxyz[ZZ];
 218     mda->dd_node_order=dd_node_order;
 219     mda->rdd=rdd;
 220     mda->rconstr=rconstr;
 221     mda->dddlb_opt=dddlb_opt;
 222     mda->dlb_scale=dlb_scale;
 223     mda->ddcsx=ddcsx;
 224     mda->ddcsy=ddcsy;
 225     mda->ddcsz=ddcsz;
 226     mda->nstepout=nstepout;
 227     mda->resetstep=resetstep;
 228     mda->nmultisim=nmultisim;
 229     mda->repl_ex_nst=repl_ex_nst;
 230     mda->repl_ex_seed=repl_ex_seed;
 231     mda->pforce=pforce;
 232     mda->cpt_period=cpt_period;
 233     mda->max_hours=max_hours;
 234     mda->deviceOptions=deviceOptions;
 235     mda->Flags=Flags;
 236
 237     fprintf(stderr, "Starting %d threads\n",nthreads);
 238     fflush(stderr);
 239     /* now spawn new threads that start mdrunner_start_fn(), while
 240        the main thread returns */
 241     ret=tMPI_Init_fn(TRUE, nthreads, mdrunner_start_fn, (void*)(mda) );
 242     if (ret!=TMPI_SUCCESS)
 243         return NULL;
 244
 245     /* make a new comm_rec to reflect the new situation */
 246     crn=init_par_threads(cr);
 247     return crn;
 248 }
 249
 250
 251 /* get the number of threads based on how many there were requested,
 252    which algorithms we're using, and how many particles there are. */
 253 static int get_nthreads(int nthreads_requested, t_inputrec *inputrec,
 254                         gmx_mtop_t *mtop)
 255 {
 256     int nthreads,nthreads_new;
 257     int min_atoms_per_thread;
 258     char *env;
 259
 260     nthreads = nthreads_requested;
 261
 262     /* determine # of hardware threads. */
 263     if (nthreads_requested < 1)
 264     {
 265         if ((env = getenv("GMX_MAX_THREADS")) != NULL)
 266         {
 267             nthreads = 0;
 268             sscanf(env,"%d",&nthreads);
 269             if (nthreads < 1)
 270             {
 271                 gmx_fatal(FARGS,"GMX_MAX_THREADS (%d) should be larger than 0",
 272                           nthreads);
 273             }
 274         }
 275         else
 276         {
 277             nthreads = tMPI_Get_recommended_nthreads();
 278         }
 279     }
 280
 281     if (inputrec->eI == eiNM || EI_TPI(inputrec->eI))
 282     {
 283         /* Steps are divided over the nodes iso splitting the atoms */
 284         min_atoms_per_thread = 0;
 285     }
 286     else
 287     {
 288         min_atoms_per_thread = MIN_ATOMS_PER_THREAD;
 289     }
 290
 291     /* Check if an algorithm does not support parallel simulation.  */
 292     if (nthreads != 1 &&
 293         ( inputrec->eI == eiLBFGS ||
 294           inputrec->coulombtype == eelEWALD ) )
 295     {
 296         fprintf(stderr,"\nThe integration or electrostatics algorithm doesn't support parallel runs. Not starting any threads.\n");
 297         nthreads = 1;
 298     }
 299     else if (nthreads_requested < 1 &&
 300              mtop->natoms/nthreads < min_atoms_per_thread)
 301     {
 302         /* the thread number was chosen automatically, but there are too many
 303            threads (too few atoms per thread) */
 304         nthreads_new = max(1,mtop->natoms/min_atoms_per_thread);
 305
 306         if (nthreads_new > 8 || (nthreads == 8 && nthreads_new > 4))
 307         {
 308             /* Use only multiples of 4 above 8 threads
 309              * or with an 8-core processor
 310              * (to avoid 6 threads on 8 core processors with 4 real cores).
 311              */
 312             nthreads_new = (nthreads_new/4)*4;
 313         }
 314         else if (nthreads_new > 4)
 315         {
 316             /* Avoid 5 or 7 threads */
 317             nthreads_new = (nthreads_new/2)*2;
 318         }
 319
 320         nthreads = nthreads_new;
 321
 322         fprintf(stderr,"\n");
 323         fprintf(stderr,"NOTE: Parallelization is limited by the small number of atoms,\n");
 324         fprintf(stderr,"      only starting %d threads.\n",nthreads);
 325         fprintf(stderr,"      You can use the -nt option to optimize the number of threads.\n\n");
 326     }
 327     return nthreads;
 328 }
 329 #endif
 330
 331
 332 int mdrunner(int nthreads_requested, FILE *fplog,t_commrec *cr,int nfile,
 333              const t_filenm fnm[], const output_env_t oenv, gmx_bool bVerbose,
 334              gmx_bool bCompact, int nstglobalcomm,
 335              ivec ddxyz,int dd_node_order,real rdd,real rconstr,
 336              const char *dddlb_opt,real dlb_scale,
 337              const char *ddcsx,const char *ddcsy,const char *ddcsz,
 338              int nstepout,int resetstep,int nmultisim,int repl_ex_nst,
 339              int repl_ex_seed, real pforce,real cpt_period,real max_hours,
 340              const char *deviceOptions, unsigned long Flags)
 341 {
 342     double     nodetime=0,realtime;
 343     t_inputrec *inputrec;
 344     t_state    *state=NULL;
 345     matrix     box;
 346     gmx_ddbox_t ddbox={0};
 347     int        npme_major,npme_minor;
 348     real       tmpr1,tmpr2;
 349     t_nrnb     *nrnb;
 350     gmx_mtop_t *mtop=NULL;
 351     t_mdatoms  *mdatoms=NULL;
 352     t_forcerec *fr=NULL;
 353     t_fcdata   *fcd=NULL;
 354     real       ewaldcoeff=0;
 355     gmx_pme_t  *pmedata=NULL;
 356     gmx_vsite_t *vsite=NULL;
 357     gmx_constr_t constr;
 358     int        i,m,nChargePerturbed=-1,status,nalloc;
 359     char       *gro;
 360     gmx_wallcycle_t wcycle;
 361     gmx_bool       bReadRNG,bReadEkin;
 362     int        list;
 363     gmx_runtime_t runtime;
 364     int        rc;
 365     gmx_large_int_t reset_counters;
 366     gmx_edsam_t ed=NULL;
 367     t_commrec   *cr_old=cr;
 368     int         nthreads=1;
 369
 370     /* CAUTION: threads may be started later on in this function, so
 371        cr doesn't reflect the final parallel state right now */
 372     snew(inputrec,1);
 373     snew(mtop,1);
 374
 375     if (bVerbose && SIMMASTER(cr))
 376     {
 377         fprintf(stderr,"Getting Loaded...\n");
 378     }
 379
 380     if (Flags & MD_APPENDFILES)
 381     {
 382         fplog = NULL;
 383     }
 384
 385     snew(state,1);
 386     if (MASTER(cr))
 387     {
 388         /* Read (nearly) all data required for the simulation */
 389         read_tpx_state(ftp2fn(efTPX,nfile,fnm),inputrec,state,NULL,mtop);
 390
 391         /* NOW the threads will be started: */
 392 #ifdef GMX_THREADS
 393         nthreads = get_nthreads(nthreads_requested, inputrec, mtop);
 394
 395         if (nthreads > 1)
 396         {
 397             /* now start the threads. */
 398             cr=mdrunner_start_threads(nthreads, fplog, cr_old, nfile, fnm,
 399                                       oenv, bVerbose, bCompact, nstglobalcomm,
 400                                       ddxyz, dd_node_order, rdd, rconstr,
 401                                       dddlb_opt, dlb_scale, ddcsx, ddcsy, ddcsz,
 402                                       nstepout, resetstep, nmultisim,
 403                                       repl_ex_nst, repl_ex_seed, pforce,
 404                                       cpt_period, max_hours, deviceOptions,
 405                                       Flags);
 406             /* the main thread continues here with a new cr. We don't deallocate
 407                the old cr because other threads may still be reading it. */
 408             if (cr == NULL)
 409             {
 410                 gmx_comm("Failed to spawn threads");
 411             }
 412         }
 413 #endif
 414     }
 415     /* END OF CAUTION: cr is now reliable */
 416
 417     if (PAR(cr))
 418     {
 419         /* now broadcast everything to the non-master nodes/threads: */
 420         init_parallel(fplog, cr, inputrec, mtop);
 421     }
 422     if (fplog != NULL)
 423     {
 424         pr_inputrec(fplog,0,"Input Parameters",inputrec,FALSE);
 425     }
 426
 427     /* now make sure the state is initialized and propagated */
 428     set_state_entries(state,inputrec,cr->nnodes);
 429
 430     /* A parallel command line option consistency check that we can
 431        only do after any threads have started. */
 432     if (!PAR(cr) &&
 433         (ddxyz[XX] > 1 || ddxyz[YY] > 1 || ddxyz[ZZ] > 1 || cr->npmenodes > 0))
 434     {
 435         gmx_fatal(FARGS,
 436                   "The -dd or -npme option request a parallel simulation, "
 437 #ifndef GMX_MPI
 438                   "but mdrun was compiled without threads or MPI enabled"
 439 #else
 440 #ifdef GMX_THREADS
 441                   "but the number of threads (option -nt) is 1"
 442 #else
 443                   "but mdrun was not started through mpirun/mpiexec or only one process was requested through mpirun/mpiexec"
 444 #endif
 445 #endif
 446             );
 447     }
 448
 449     if (can_use_allvsall(inputrec,mtop,TRUE,cr,fplog))
 450     {
 451         /* All-vs-all loops do not work with domain decomposition */
 452         Flags |= MD_PARTDEC;
 453     }
 454
 455     if (!EEL_PME(inputrec->coulombtype) || (Flags & MD_PARTDEC))
 456     {
 457         cr->npmenodes = 0;
 458     }
 459
 460 #ifdef GMX_FAHCORE
 461     fcRegisterSteps(inputrec->nsteps,inputrec->init_step);
 462 #endif
 463
 464     /* NMR restraints must be initialized before load_checkpoint,
 465      * since with time averaging the history is added to t_state.
 466      * For proper consistency check we therefore need to extend
 467      * t_state here.
 468      * So the PME-only nodes (if present) will also initialize
 469      * the distance restraints.
 470      */
 471     snew(fcd,1);
 472
 473     /* This needs to be called before read_checkpoint to extend the state */
 474     init_disres(fplog,mtop,inputrec,cr,Flags & MD_PARTDEC,fcd,state);
 475
 476     if (gmx_mtop_ftype_count(mtop,F_ORIRES) > 0)
 477     {
 478         if (PAR(cr) && !(Flags & MD_PARTDEC))
 479         {
 480             gmx_fatal(FARGS,"Orientation restraints do not work (yet) with domain decomposition, use particle decomposition (mdrun option -pd)");
 481         }
 482         /* Orientation restraints */
 483         if (MASTER(cr))
 484         {
 485             init_orires(fplog,mtop,state->x,inputrec,cr->ms,&(fcd->orires),
 486                         state);
 487         }
 488     }
 489
 490     if (DEFORM(*inputrec))
 491     {
 492         /* Store the deform reference box before reading the checkpoint */
 493         if (SIMMASTER(cr))
 494         {
 495             copy_mat(state->box,box);
 496         }
 497         if (PAR(cr))
 498         {
 499             gmx_bcast(sizeof(box),box,cr);
 500         }
 501         /* Because we do not have the update struct available yet
 502          * in which the reference values should be stored,
 503          * we store them temporarily in static variables.
 504          * This should be thread safe, since they are only written once
 505          * and with identical values.
 506          */
 507 #ifdef GMX_THREADS
 508         tMPI_Thread_mutex_lock(&deform_init_box_mutex);
 509 #endif
 510         deform_init_init_step_tpx = inputrec->init_step;
 511         copy_mat(box,deform_init_box_tpx);
 512 #ifdef GMX_THREADS
 513         tMPI_Thread_mutex_unlock(&deform_init_box_mutex);
 514 #endif
 515     }
 516
 517     if (opt2bSet("-cpi",nfile,fnm))
 518     {
 519         /* Check if checkpoint file exists before doing continuation.
 520          * This way we can use identical input options for the first and subsequent runs...
 521          */
 522         if( gmx_fexist_master(opt2fn_master("-cpi",nfile,fnm,cr),cr) )
 523         {
 524             load_checkpoint(opt2fn_master("-cpi",nfile,fnm,cr),&fplog,
 525                             cr,Flags & MD_PARTDEC,ddxyz,
 526                             inputrec,state,&bReadRNG,&bReadEkin,
 527                             (Flags & MD_APPENDFILES));
 528
 529             if (bReadRNG)
 530             {
 531                 Flags |= MD_READ_RNG;
 532             }
 533             if (bReadEkin)
 534             {
 535                 Flags |= MD_READ_EKIN;
 536             }
 537         }
 538     }
 539
 540     if (((MASTER(cr) || (Flags & MD_SEPPOT)) && (Flags & MD_APPENDFILES))
 541 #ifdef GMX_THREADS
 542         /* With thread MPI only the master node/thread exists in mdrun.c,
 543          * therefore non-master nodes need to open the "seppot" log file here.
 544          */
 545         || (!MASTER(cr) && (Flags & MD_SEPPOT))
 546 #endif
 547         )
 548     {
 549         gmx_log_open(ftp2fn(efLOG,nfile,fnm),cr,!(Flags & MD_SEPPOT),
 550                              Flags,&fplog);
 551     }
 552
 553     if (SIMMASTER(cr))
 554     {
 555         copy_mat(state->box,box);
 556     }
 557
 558     if (PAR(cr))
 559     {
 560         gmx_bcast(sizeof(box),box,cr);
 561     }
 562
 563     /* Essential dynamics */
 564     if (opt2bSet("-ei",nfile,fnm))
 565     {
 566         /* Open input and output files, allocate space for ED data structure */
 567         ed = ed_open(nfile,fnm,Flags,cr);
 568     }
 569
 570     if (bVerbose && SIMMASTER(cr))
 571     {
 572         fprintf(stderr,"Loaded with Money\n\n");
 573     }
 574
 575     if (PAR(cr) && !((Flags & MD_PARTDEC) ||
 576                      EI_TPI(inputrec->eI) ||
 577                      inputrec->eI == eiNM))
 578     {
 579         cr->dd = init_domain_decomposition(fplog,cr,Flags,ddxyz,rdd,rconstr,
 580                                            dddlb_opt,dlb_scale,
 581                                            ddcsx,ddcsy,ddcsz,
 582                                            mtop,inputrec,
 583                                            box,state->x,
 584                                            &ddbox,&npme_major,&npme_minor);
 585
 586         make_dd_communicators(fplog,cr,dd_node_order);
 587
 588         /* Set overallocation to avoid frequent reallocation of arrays */
 589         set_over_alloc_dd(TRUE);
 590     }
 591     else
 592     {
 593         /* PME, if used, is done on all nodes with 1D decomposition */
 594         cr->npmenodes = 0;
 595         cr->duty = (DUTY_PP | DUTY_PME);
 596         npme_major = 1;
 597         npme_minor = 1;
 598         if (!EI_TPI(inputrec->eI))
 599         {
 600             npme_major = cr->nnodes;
 601         }
 602
 603         if (inputrec->ePBC == epbcSCREW)
 604         {
 605             gmx_fatal(FARGS,
 606                       "pbc=%s is only implemented with domain decomposition",
 607                       epbc_names[inputrec->ePBC]);
 608         }
 609     }
 610
 611     if (PAR(cr))
 612     {
 613         /* After possible communicator splitting in make_dd_communicators.
 614          * we can set up the intra/inter node communication.
 615          */
 616         gmx_setup_nodecomm(fplog,cr);
 617     }
 618
 619     wcycle = wallcycle_init(fplog,resetstep,cr);
 620     if (PAR(cr))
 621     {
 622         /* Master synchronizes its value of reset_counters with all nodes
 623          * including PME only nodes */
 624         reset_counters = wcycle_get_reset_counters(wcycle);
 625         gmx_bcast_sim(sizeof(reset_counters),&reset_counters,cr);
 626         wcycle_set_reset_counters(wcycle, reset_counters);
 627     }
 628
 629
 630     snew(nrnb,1);
 631     if (cr->duty & DUTY_PP)
 632     {
 633         /* For domain decomposition we allocate dynamically
 634          * in dd_partition_system.
 635          */
 636         if (DOMAINDECOMP(cr))
 637         {
 638             bcast_state_setup(cr,state);
 639         }
 640         else
 641         {
 642             if (PAR(cr))
 643             {
 644                 if (!MASTER(cr))
 645                 {
 646                     snew(state,1);
 647                 }
 648                 bcast_state(cr,state,TRUE);
 649             }
 650         }
 651
 652         /* Dihedral Restraints */
 653         if (gmx_mtop_ftype_count(mtop,F_DIHRES) > 0)
 654         {
 655             init_dihres(fplog,mtop,inputrec,fcd);
 656         }
 657
 658         /* Initiate forcerecord */
 659         fr = mk_forcerec();
 660         init_forcerec(fplog,oenv,fr,fcd,inputrec,mtop,cr,box,FALSE,
 661                       opt2fn("-table",nfile,fnm),
 662                       opt2fn("-tablep",nfile,fnm),
 663                       opt2fn("-tableb",nfile,fnm),FALSE,pforce);
 664
 665         /* version for PCA_NOT_READ_NODE (see md.c) */
 666         /*init_forcerec(fplog,fr,fcd,inputrec,mtop,cr,box,FALSE,
 667           "nofile","nofile","nofile",FALSE,pforce);
 668           */
 669         fr->bSepDVDL = ((Flags & MD_SEPPOT) == MD_SEPPOT);
 670
 671         /* Initialize QM-MM */
 672         if(fr->bQMMM)
 673         {
 674             init_QMMMrec(cr,box,mtop,inputrec,fr);
 675         }
 676
 677         /* Initialize the mdatoms structure.
 678          * mdatoms is not filled with atom data,
 679          * as this can not be done now with domain decomposition.
 680          */
 681         mdatoms = init_mdatoms(fplog,mtop,inputrec->efep!=efepNO);
 682
 683         /* Initialize the virtual site communication */
 684         vsite = init_vsite(mtop,cr);
 685
 686         calc_shifts(box,fr->shift_vec);
 687
 688         /* With periodic molecules the charge groups should be whole at start up
 689          * and the virtual sites should not be far from their proper positions.
 690          */
 691         if (!inputrec->bContinuation && MASTER(cr) &&
 692             !(inputrec->ePBC != epbcNONE && inputrec->bPeriodicMols))
 693         {
 694             /* Make molecules whole at start of run */
 695             if (fr->ePBC != epbcNONE)
 696             {
 697                 do_pbc_first_mtop(fplog,inputrec->ePBC,box,mtop,state->x);
 698             }
 699             if (vsite)
 700             {
 701                 /* Correct initial vsite positions are required
 702                  * for the initial distribution in the domain decomposition
 703                  * and for the initial shell prediction.
 704                  */
 705                 construct_vsites_mtop(fplog,vsite,mtop,state->x);
 706             }
 707         }
 708
 709         /* Initiate PPPM if necessary */
 710         if (fr->eeltype == eelPPPM)
 711         {
 712             if (mdatoms->nChargePerturbed)
 713             {
 714                 gmx_fatal(FARGS,"Free energy with %s is not implemented",
 715                           eel_names[fr->eeltype]);
 716             }
 717             status = gmx_pppm_init(fplog,cr,oenv,FALSE,TRUE,box,
 718                                    getenv("GMXGHAT"),inputrec, (Flags & MD_REPRODUCIBLE));
 719             if (status != 0)
 720             {
 721                 gmx_fatal(FARGS,"Error %d initializing PPPM",status);
 722             }
 723         }
 724
 725         if (EEL_PME(fr->eeltype))
 726         {
 727             ewaldcoeff = fr->ewaldcoeff;
 728             pmedata = &fr->pmedata;
 729         }
 730         else
 731         {
 732             pmedata = NULL;
 733         }
 734     }
 735     else
 736     {
 737         /* This is a PME only node */
 738
 739         /* We don't need the state */
 740         done_state(state);
 741
 742         ewaldcoeff = calc_ewaldcoeff(inputrec->rcoulomb, inputrec->ewald_rtol);
 743         snew(pmedata,1);
 744     }
 745
 746     /* Initiate PME if necessary,
 747      * either on all nodes or on dedicated PME nodes only. */
 748     if (EEL_PME(inputrec->coulombtype))
 749     {
 750         if (mdatoms)
 751         {
 752             nChargePerturbed = mdatoms->nChargePerturbed;
 753         }
 754         if (cr->npmenodes > 0)
 755         {
 756             /* The PME only nodes need to know nChargePerturbed */
 757             gmx_bcast_sim(sizeof(nChargePerturbed),&nChargePerturbed,cr);
 758         }
 759         if (cr->duty & DUTY_PME)
 760         {
 761             status = gmx_pme_init(pmedata,cr,npme_major,npme_minor,inputrec,
 762                                   mtop ? mtop->natoms : 0,nChargePerturbed,
 763                                   (Flags & MD_REPRODUCIBLE));
 764             if (status != 0)
 765             {
 766                 gmx_fatal(FARGS,"Error %d initializing PME",status);
 767             }
 768         }
 769     }
 770
 771
 772     if (integrator[inputrec->eI].func == do_md
 773 #ifdef GMX_OPENMM
 774         ||
 775         integrator[inputrec->eI].func == do_md_openmm
 776 #endif
 777         )
 778     {
 779         /* Turn on signal handling on all nodes */
 780         /*
 781          * (A user signal from the PME nodes (if any)
 782          * is communicated to the PP nodes.
 783          */
 784         signal_handler_install();
 785     }
 786
 787     if (cr->duty & DUTY_PP)
 788     {
 789         if (inputrec->ePull != epullNO)
 790         {
 791             /* Initialize pull code */
 792             init_pull(fplog,inputrec,nfile,fnm,mtop,cr,oenv,
 793                       EI_DYNAMICS(inputrec->eI) && MASTER(cr),Flags);
 794         }
 795
 796         if (inputrec->bRot)
 797         {
 798            /* Initialize enforced rotation code */
 799            init_rot(fplog,inputrec,nfile,fnm,cr,state->x,state->box,mtop,oenv,
 800                     bVerbose,Flags);
 801         }
 802
 803         constr = init_constraints(fplog,mtop,inputrec,ed,state,cr);
 804
 805         if (DOMAINDECOMP(cr))
 806         {
 807             dd_init_bondeds(fplog,cr->dd,mtop,vsite,constr,inputrec,
 808                             Flags & MD_DDBONDCHECK,fr->cginfo_mb);
 809
 810             set_dd_parameters(fplog,cr->dd,dlb_scale,inputrec,fr,&ddbox);
 811
 812             setup_dd_grid(fplog,cr->dd);
 813         }
 814
 815         /* Now do whatever the user wants us to do (how flexible...) */
 816         integrator[inputrec->eI].func(fplog,cr,nfile,fnm,
 817                                       oenv,bVerbose,bCompact,
 818                                       nstglobalcomm,
 819                                       vsite,constr,
 820                                       nstepout,inputrec,mtop,
 821                                       fcd,state,
 822                                       mdatoms,nrnb,wcycle,ed,fr,
 823                                       repl_ex_nst,repl_ex_seed,
 824                                       cpt_period,max_hours,
 825                                       deviceOptions,
 826                                       Flags,
 827                                       &runtime);
 828
 829         if (inputrec->ePull != epullNO)
 830         {
 831             finish_pull(fplog,inputrec->pull);
 832         }
 833
 834         if (inputrec->bRot)
 835         {
 836             finish_rot(fplog,inputrec->rot);
 837         }
 838
 839     }
 840     else
 841     {
 842         /* do PME only */
 843         gmx_pmeonly(*pmedata,cr,nrnb,wcycle,ewaldcoeff,FALSE,inputrec);
 844     }
 845
 846     if (EI_DYNAMICS(inputrec->eI) || EI_TPI(inputrec->eI))
 847     {
 848         /* Some timing stats */
 849         if (SIMMASTER(cr))
 850         {
 851             if (runtime.proc == 0)
 852             {
 853                 runtime.proc = runtime.real;
 854             }
 855         }
 856         else
 857         {
 858             runtime.real = 0;
 859         }
 860     }
 861
 862     wallcycle_stop(wcycle,ewcRUN);
 863
 864     /* Finish up, write some stuff
 865      * if rerunMD, don't write last frame again
 866      */
 867     finish_run(fplog,cr,ftp2fn(efSTO,nfile,fnm),
 868                inputrec,nrnb,wcycle,&runtime,
 869                EI_DYNAMICS(inputrec->eI) && !MULTISIM(cr));
 870
 871     /* Does what it says */
 872     print_date_and_time(fplog,cr->nodeid,"Finished mdrun",&runtime);
 873
 874     /* Close logfile already here if we were appending to it */
 875     if (MASTER(cr) && (Flags & MD_APPENDFILES))
 876     {
 877         gmx_log_close(fplog);
 878     }
 879
 880     rc=(int)gmx_get_stop_condition();
 881
 882 #ifdef GMX_THREADS
 883     /* we need to join all threads. The sub-threads join when they
 884        exit this function, but the master thread needs to be told to
 885        wait for that. */
 886     if (PAR(cr) && MASTER(cr))
 887     {
 888         tMPI_Finalize();
 889     }
 890 #endif
 891
 892     return rc;
 893 }
 894
 895 void md_print_warning(const t_commrec *cr,FILE *fplog,const char *buf)
 896 {
 897     if (MASTER(cr))
 898     {
 899         fprintf(stderr,"\n%s\n",buf);
 900     }
 901     if (fplog)
 902     {
 903         fprintf(fplog,"\n%s\n",buf);
 904     }
 905 }