2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2011,2012,2013,2014, by the GROMACS development team, led by
7 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
8 * and including many others, as listed in the AUTHORS file in the
9 * top-level source directory and at http://www.gromacs.org.
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.
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.
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.
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.
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.
43 #include "gromacs/ewald/pme-load-balancing.h"
44 #include "gromacs/ewald/pme.h"
45 #include "gromacs/fileio/confio.h"
46 #include "gromacs/fileio/mdoutf.h"
47 #include "gromacs/fileio/trajectory_writing.h"
48 #include "gromacs/fileio/trnio.h"
49 #include "gromacs/fileio/trxio.h"
50 #include "gromacs/fileio/xtcio.h"
51 #include "gromacs/gmxpreprocess/compute_io.h"
52 #include "gromacs/imd/imd.h"
53 #include "gromacs/legacyheaders/bonded-threading.h"
54 #include "gromacs/legacyheaders/calcmu.h"
55 #include "gromacs/legacyheaders/checkpoint.h"
56 #include "gromacs/legacyheaders/constr.h"
57 #include "gromacs/legacyheaders/disre.h"
58 #include "gromacs/legacyheaders/domdec.h"
59 #include "gromacs/legacyheaders/domdec_network.h"
60 #include "gromacs/legacyheaders/force.h"
61 #include "gromacs/legacyheaders/md_logging.h"
62 #include "gromacs/legacyheaders/md_support.h"
63 #include "gromacs/legacyheaders/mdatoms.h"
64 #include "gromacs/legacyheaders/mdebin.h"
65 #include "gromacs/legacyheaders/mdrun.h"
66 #include "gromacs/legacyheaders/names.h"
67 #include "gromacs/legacyheaders/network.h"
68 #include "gromacs/legacyheaders/nrnb.h"
69 #include "gromacs/legacyheaders/ns.h"
70 #include "gromacs/legacyheaders/orires.h"
71 #include "gromacs/legacyheaders/qmmm.h"
72 #include "gromacs/legacyheaders/shellfc.h"
73 #include "gromacs/legacyheaders/sighandler.h"
74 #include "gromacs/legacyheaders/txtdump.h"
75 #include "gromacs/legacyheaders/typedefs.h"
76 #include "gromacs/legacyheaders/update.h"
77 #include "gromacs/legacyheaders/vcm.h"
78 #include "gromacs/legacyheaders/vsite.h"
79 #include "gromacs/legacyheaders/types/iteratedconstraints.h"
80 #include "gromacs/legacyheaders/types/nlistheuristics.h"
81 #include "gromacs/math/vec.h"
82 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_data_mgmt.h"
83 #include "gromacs/pbcutil/mshift.h"
84 #include "gromacs/pbcutil/pbc.h"
85 #include "gromacs/pulling/pull.h"
86 #include "gromacs/swap/swapcoords.h"
87 #include "gromacs/timing/wallcycle.h"
88 #include "gromacs/timing/walltime_accounting.h"
89 #include "gromacs/topology/mtop_util.h"
90 #include "gromacs/utility/cstringutil.h"
91 #include "gromacs/utility/gmxmpi.h"
92 #include "gromacs/utility/smalloc.h"
102 static void reset_all_counters(FILE *fplog, t_commrec *cr,
104 gmx_int64_t *step_rel, t_inputrec *ir,
105 gmx_wallcycle_t wcycle, t_nrnb *nrnb,
106 gmx_walltime_accounting_t walltime_accounting,
107 struct nonbonded_verlet_t *nbv)
109 char sbuf[STEPSTRSIZE];
111 /* Reset all the counters related to performance over the run */
112 md_print_warn(cr, fplog, "step %s: resetting all time and cycle counters\n",
113 gmx_step_str(step, sbuf));
115 nbnxn_cuda_reset_timings(nbv);
117 wallcycle_stop(wcycle, ewcRUN);
118 wallcycle_reset_all(wcycle);
119 if (DOMAINDECOMP(cr))
121 reset_dd_statistics_counters(cr->dd);
124 ir->init_step += *step_rel;
125 ir->nsteps -= *step_rel;
127 wallcycle_start(wcycle, ewcRUN);
128 walltime_accounting_start(walltime_accounting);
129 print_date_and_time(fplog, cr->nodeid, "Restarted time", gmx_gettime());
132 double do_md(FILE *fplog, t_commrec *cr, int nfile, const t_filenm fnm[],
133 const output_env_t oenv, gmx_bool bVerbose, gmx_bool bCompact,
135 gmx_vsite_t *vsite, gmx_constr_t constr,
136 int stepout, t_inputrec *ir,
137 gmx_mtop_t *top_global,
139 t_state *state_global,
141 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
142 gmx_edsam_t ed, t_forcerec *fr,
143 int repl_ex_nst, int repl_ex_nex, int repl_ex_seed, gmx_membed_t membed,
144 real cpt_period, real max_hours,
145 const char gmx_unused *deviceOptions,
148 gmx_walltime_accounting_t walltime_accounting)
150 gmx_mdoutf_t outf = NULL;
151 gmx_int64_t step, step_rel;
153 double t, t0, lam0[efptNR];
154 gmx_bool bGStatEveryStep, bGStat, bCalcVir, bCalcEner;
155 gmx_bool bNS, bNStList, bSimAnn, bStopCM, bRerunMD, bNotLastFrame = FALSE,
156 bFirstStep, bStateFromCP, bStateFromTPX, bInitStep, bLastStep,
157 bBornRadii, bStartingFromCpt;
158 gmx_bool bDoDHDL = FALSE, bDoFEP = FALSE, bDoExpanded = FALSE;
159 gmx_bool do_ene, do_log, do_verbose, bRerunWarnNoV = TRUE,
160 bForceUpdate = FALSE, bCPT;
161 gmx_bool bMasterState;
162 int force_flags, cglo_flags;
163 tensor force_vir, shake_vir, total_vir, tmp_vir, pres;
168 t_state *bufstate = NULL;
172 gmx_repl_ex_t repl_ex = NULL;
175 t_mdebin *mdebin = NULL;
176 t_state *state = NULL;
177 rvec *f_global = NULL;
178 gmx_enerdata_t *enerd;
180 gmx_global_stat_t gstat;
181 gmx_update_t upd = NULL;
182 t_graph *graph = NULL;
184 gmx_groups_t *groups;
185 gmx_ekindata_t *ekind, *ekind_save;
186 gmx_shellfc_t shellfc;
187 int count, nconverged = 0;
189 gmx_bool bConverged = TRUE, bSumEkinhOld, bDoReplEx, bExchanged, bNeedRepartition;
190 gmx_bool bResetCountersHalfMaxH = FALSE;
191 gmx_bool bVV, bIterativeCase, bFirstIterate, bTemp, bPres, bTrotter;
192 gmx_bool bUpdateDoLR;
196 real veta_save, scalevir, tracevir;
202 real saved_conserved_quantity = 0;
206 char sbuf[STEPSTRSIZE], sbuf2[STEPSTRSIZE];
207 int handled_stop_condition = gmx_stop_cond_none; /* compare to get_stop_condition*/
208 gmx_iterate_t iterate;
209 gmx_int64_t multisim_nsteps = -1; /* number of steps to do before first multisim
210 simulation stops. If equal to zero, don't
211 communicate any more between multisims.*/
212 /* PME load balancing data for GPU kernels */
213 pme_load_balancing_t pme_loadbal = NULL;
215 gmx_bool bPMETuneTry = FALSE, bPMETuneRunning = FALSE;
218 gmx_bool bIMDstep = FALSE;
221 /* Temporary addition for FAHCORE checkpointing */
225 /* Check for special mdrun options */
226 bRerunMD = (Flags & MD_RERUN);
227 if (Flags & MD_RESETCOUNTERSHALFWAY)
231 /* Signal to reset the counters half the simulation steps. */
232 wcycle_set_reset_counters(wcycle, ir->nsteps/2);
234 /* Signal to reset the counters halfway the simulation time. */
235 bResetCountersHalfMaxH = (max_hours > 0);
238 /* md-vv uses averaged full step velocities for T-control
239 md-vv-avek uses averaged half step velocities for T-control (but full step ekin for P control)
240 md uses averaged half step kinetic energies to determine temperature unless defined otherwise by GMX_EKIN_AVE_VEL; */
242 if (bVV) /* to store the initial velocities while computing virial */
244 snew(cbuf, top_global->natoms);
246 /* all the iteratative cases - only if there are constraints */
247 bIterativeCase = ((IR_NPH_TROTTER(ir) || IR_NPT_TROTTER(ir)) && (constr) && (!bRerunMD));
248 gmx_iterate_init(&iterate, FALSE); /* The default value of iterate->bIterationActive is set to
249 false in this step. The correct value, true or false,
250 is set at each step, as it depends on the frequency of temperature
251 and pressure control.*/
252 bTrotter = (bVV && (IR_NPT_TROTTER(ir) || IR_NPH_TROTTER(ir) || IR_NVT_TROTTER(ir)));
256 /* Since we don't know if the frames read are related in any way,
257 * rebuild the neighborlist at every step.
260 ir->nstcalcenergy = 1;
264 check_ir_old_tpx_versions(cr, fplog, ir, top_global);
266 nstglobalcomm = check_nstglobalcomm(fplog, cr, nstglobalcomm, ir);
267 bGStatEveryStep = (nstglobalcomm == 1);
269 if (!bGStatEveryStep && ir->nstlist == -1 && fplog != NULL)
272 "To reduce the energy communication with nstlist = -1\n"
273 "the neighbor list validity should not be checked at every step,\n"
274 "this means that exact integration is not guaranteed.\n"
275 "The neighbor list validity is checked after:\n"
276 " <n.list life time> - 2*std.dev.(n.list life time) steps.\n"
277 "In most cases this will result in exact integration.\n"
278 "This reduces the energy communication by a factor of 2 to 3.\n"
279 "If you want less energy communication, set nstlist > 3.\n\n");
284 ir->nstxout_compressed = 0;
286 groups = &top_global->groups;
289 init_md(fplog, cr, ir, oenv, &t, &t0, state_global->lambda,
290 &(state_global->fep_state), lam0,
291 nrnb, top_global, &upd,
292 nfile, fnm, &outf, &mdebin,
293 force_vir, shake_vir, mu_tot, &bSimAnn, &vcm, Flags, wcycle);
295 clear_mat(total_vir);
297 /* Energy terms and groups */
299 init_enerdata(top_global->groups.grps[egcENER].nr, ir->fepvals->n_lambda,
301 if (DOMAINDECOMP(cr))
307 snew(f, top_global->natoms);
310 /* Kinetic energy data */
312 init_ekindata(fplog, top_global, &(ir->opts), ekind);
313 /* needed for iteration of constraints */
315 init_ekindata(fplog, top_global, &(ir->opts), ekind_save);
316 /* Copy the cos acceleration to the groups struct */
317 ekind->cosacc.cos_accel = ir->cos_accel;
319 gstat = global_stat_init(ir);
322 /* Check for polarizable models and flexible constraints */
323 shellfc = init_shell_flexcon(fplog,
324 top_global, n_flexible_constraints(constr),
325 (ir->bContinuation ||
326 (DOMAINDECOMP(cr) && !MASTER(cr))) ?
327 NULL : state_global->x);
328 if (shellfc && ir->nstcalcenergy != 1)
330 gmx_fatal(FARGS, "You have nstcalcenergy set to a value (%d) that is different from 1.\nThis is not supported in combinations with shell particles.\nPlease make a new tpr file.", ir->nstcalcenergy);
332 if (shellfc && DOMAINDECOMP(cr))
334 gmx_fatal(FARGS, "Shell particles are not implemented with domain decomposition, use a single rank");
336 if (shellfc && ir->eI == eiNM)
338 /* Currently shells don't work with Normal Modes */
339 gmx_fatal(FARGS, "Normal Mode analysis is not supported with shells.\nIf you'd like to help with adding support, we have an open discussion at http://redmine.gromacs.org/issues/879\n");
342 if (vsite && ir->eI == eiNM)
344 /* Currently virtual sites don't work with Normal Modes */
345 gmx_fatal(FARGS, "Normal Mode analysis is not supported with virtual sites.\nIf you'd like to help with adding support, we have an open discussion at http://redmine.gromacs.org/issues/879\n");
350 tMPI_Thread_mutex_lock(&deform_init_box_mutex);
351 set_deform_reference_box(upd,
352 deform_init_init_step_tpx,
353 deform_init_box_tpx);
354 tMPI_Thread_mutex_unlock(&deform_init_box_mutex);
358 double io = compute_io(ir, top_global->natoms, groups, mdebin->ebin->nener, 1);
359 if ((io > 2000) && MASTER(cr))
362 "\nWARNING: This run will generate roughly %.0f Mb of data\n\n",
367 if (DOMAINDECOMP(cr))
369 top = dd_init_local_top(top_global);
372 dd_init_local_state(cr->dd, state_global, state);
374 if (DDMASTER(cr->dd) && ir->nstfout)
376 snew(f_global, state_global->natoms);
381 top = gmx_mtop_generate_local_top(top_global, ir);
383 forcerec_set_excl_load(fr, top);
385 state = serial_init_local_state(state_global);
388 atoms2md(top_global, ir, 0, NULL, top_global->natoms, mdatoms);
392 set_vsite_top(vsite, top, mdatoms, cr);
395 if (ir->ePBC != epbcNONE && !fr->bMolPBC)
397 graph = mk_graph(fplog, &(top->idef), 0, top_global->natoms, FALSE, FALSE);
402 make_local_shells(cr, mdatoms, shellfc);
405 setup_bonded_threading(fr, &top->idef);
408 /* Set up interactive MD (IMD) */
409 init_IMD(ir, cr, top_global, fplog, ir->nstcalcenergy, state_global->x,
410 nfile, fnm, oenv, imdport, Flags);
412 if (DOMAINDECOMP(cr))
414 /* Distribute the charge groups over the nodes from the master node */
415 dd_partition_system(fplog, ir->init_step, cr, TRUE, 1,
416 state_global, top_global, ir,
417 state, &f, mdatoms, top, fr,
418 vsite, shellfc, constr,
419 nrnb, wcycle, FALSE);
423 update_mdatoms(mdatoms, state->lambda[efptMASS]);
425 if (opt2bSet("-cpi", nfile, fnm))
427 bStateFromCP = gmx_fexist_master(opt2fn_master("-cpi", nfile, fnm, cr), cr);
431 bStateFromCP = FALSE;
436 init_expanded_ensemble(bStateFromCP, ir, &state->dfhist);
443 /* Update mdebin with energy history if appending to output files */
444 if (Flags & MD_APPENDFILES)
446 restore_energyhistory_from_state(mdebin, &state_global->enerhist);
450 /* We might have read an energy history from checkpoint,
451 * free the allocated memory and reset the counts.
453 done_energyhistory(&state_global->enerhist);
454 init_energyhistory(&state_global->enerhist);
457 /* Set the initial energy history in state by updating once */
458 update_energyhistory(&state_global->enerhist, mdebin);
461 /* Initialize constraints */
462 if (constr && !DOMAINDECOMP(cr))
464 set_constraints(constr, top, ir, mdatoms, cr);
467 if (repl_ex_nst > 0 && MASTER(cr))
469 repl_ex = init_replica_exchange(fplog, cr->ms, state_global, ir,
470 repl_ex_nst, repl_ex_nex, repl_ex_seed);
473 /* PME tuning is only supported with GPUs or PME nodes and not with rerun.
474 * PME tuning is not supported with PME only for LJ and not for Coulomb.
476 if ((Flags & MD_TUNEPME) &&
477 EEL_PME(fr->eeltype) &&
478 ( use_GPU(fr->nbv) || !(cr->duty & DUTY_PME)) &&
481 pme_loadbal_init(&pme_loadbal, ir, state->box, fr->ic, fr->pmedata);
483 if (cr->duty & DUTY_PME)
485 /* Start tuning right away, as we can't measure the load */
486 bPMETuneRunning = TRUE;
490 /* Separate PME nodes, we can measure the PP/PME load balance */
495 if (!ir->bContinuation && !bRerunMD)
497 if (mdatoms->cFREEZE && (state->flags & (1<<estV)))
499 /* Set the velocities of frozen particles to zero */
500 for (i = 0; i < mdatoms->homenr; i++)
502 for (m = 0; m < DIM; m++)
504 if (ir->opts.nFreeze[mdatoms->cFREEZE[i]][m])
514 /* Constrain the initial coordinates and velocities */
515 do_constrain_first(fplog, constr, ir, mdatoms, state,
520 /* Construct the virtual sites for the initial configuration */
521 construct_vsites(vsite, state->x, ir->delta_t, NULL,
522 top->idef.iparams, top->idef.il,
523 fr->ePBC, fr->bMolPBC, cr, state->box);
529 /* set free energy calculation frequency as the minimum
530 greatest common denominator of nstdhdl, nstexpanded, and repl_ex_nst*/
531 nstfep = ir->fepvals->nstdhdl;
534 nstfep = gmx_greatest_common_divisor(ir->fepvals->nstdhdl, nstfep);
538 nstfep = gmx_greatest_common_divisor(repl_ex_nst, nstfep);
541 /* I'm assuming we need global communication the first time! MRS */
542 cglo_flags = (CGLO_TEMPERATURE | CGLO_GSTAT
543 | ((ir->comm_mode != ecmNO) ? CGLO_STOPCM : 0)
544 | (bVV ? CGLO_PRESSURE : 0)
545 | (bVV ? CGLO_CONSTRAINT : 0)
546 | (bRerunMD ? CGLO_RERUNMD : 0)
547 | ((Flags & MD_READ_EKIN) ? CGLO_READEKIN : 0));
549 bSumEkinhOld = FALSE;
550 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
551 NULL, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
552 constr, NULL, FALSE, state->box,
553 top_global, &bSumEkinhOld, cglo_flags);
554 if (ir->eI == eiVVAK)
556 /* a second call to get the half step temperature initialized as well */
557 /* we do the same call as above, but turn the pressure off -- internally to
558 compute_globals, this is recognized as a velocity verlet half-step
559 kinetic energy calculation. This minimized excess variables, but
560 perhaps loses some logic?*/
562 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
563 NULL, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
564 constr, NULL, FALSE, state->box,
565 top_global, &bSumEkinhOld,
566 cglo_flags &~(CGLO_STOPCM | CGLO_PRESSURE));
569 /* Calculate the initial half step temperature, and save the ekinh_old */
570 if (!(Flags & MD_STARTFROMCPT))
572 for (i = 0; (i < ir->opts.ngtc); i++)
574 copy_mat(ekind->tcstat[i].ekinh, ekind->tcstat[i].ekinh_old);
579 enerd->term[F_TEMP] *= 2; /* result of averages being done over previous and current step,
580 and there is no previous step */
583 /* if using an iterative algorithm, we need to create a working directory for the state. */
586 bufstate = init_bufstate(state);
589 /* need to make an initiation call to get the Trotter variables set, as well as other constants for non-trotter
590 temperature control */
591 trotter_seq = init_npt_vars(ir, state, &MassQ, bTrotter);
595 if (constr && !ir->bContinuation && ir->eConstrAlg == econtLINCS)
598 "RMS relative constraint deviation after constraining: %.2e\n",
599 constr_rmsd(constr, FALSE));
601 if (EI_STATE_VELOCITY(ir->eI))
603 fprintf(fplog, "Initial temperature: %g K\n", enerd->term[F_TEMP]);
607 fprintf(stderr, "starting md rerun '%s', reading coordinates from"
608 " input trajectory '%s'\n\n",
609 *(top_global->name), opt2fn("-rerun", nfile, fnm));
612 fprintf(stderr, "Calculated time to finish depends on nsteps from "
613 "run input file,\nwhich may not correspond to the time "
614 "needed to process input trajectory.\n\n");
620 fprintf(stderr, "starting mdrun '%s'\n",
621 *(top_global->name));
624 sprintf(tbuf, "%8.1f", (ir->init_step+ir->nsteps)*ir->delta_t);
628 sprintf(tbuf, "%s", "infinite");
630 if (ir->init_step > 0)
632 fprintf(stderr, "%s steps, %s ps (continuing from step %s, %8.1f ps).\n",
633 gmx_step_str(ir->init_step+ir->nsteps, sbuf), tbuf,
634 gmx_step_str(ir->init_step, sbuf2),
635 ir->init_step*ir->delta_t);
639 fprintf(stderr, "%s steps, %s ps.\n",
640 gmx_step_str(ir->nsteps, sbuf), tbuf);
643 fprintf(fplog, "\n");
646 walltime_accounting_start(walltime_accounting);
647 wallcycle_start(wcycle, ewcRUN);
648 print_start(fplog, cr, walltime_accounting, "mdrun");
650 /* safest point to do file checkpointing is here. More general point would be immediately before integrator call */
652 chkpt_ret = fcCheckPointParallel( cr->nodeid,
656 gmx_fatal( 3, __FILE__, __LINE__, "Checkpoint error on step %d\n", 0 );
661 /***********************************************************
665 ************************************************************/
667 /* if rerunMD then read coordinates and velocities from input trajectory */
670 if (getenv("GMX_FORCE_UPDATE"))
678 bNotLastFrame = read_first_frame(oenv, &status,
679 opt2fn("-rerun", nfile, fnm),
680 &rerun_fr, TRX_NEED_X | TRX_READ_V);
681 if (rerun_fr.natoms != top_global->natoms)
684 "Number of atoms in trajectory (%d) does not match the "
685 "run input file (%d)\n",
686 rerun_fr.natoms, top_global->natoms);
688 if (ir->ePBC != epbcNONE)
692 gmx_fatal(FARGS, "Rerun trajectory frame step %d time %f does not contain a box, while pbc is used", rerun_fr.step, rerun_fr.time);
694 if (max_cutoff2(ir->ePBC, rerun_fr.box) < sqr(fr->rlistlong))
696 gmx_fatal(FARGS, "Rerun trajectory frame step %d time %f has too small box dimensions", rerun_fr.step, rerun_fr.time);
703 rerun_parallel_comm(cr, &rerun_fr, &bNotLastFrame);
706 if (ir->ePBC != epbcNONE)
708 /* Set the shift vectors.
709 * Necessary here when have a static box different from the tpr box.
711 calc_shifts(rerun_fr.box, fr->shift_vec);
715 /* loop over MD steps or if rerunMD to end of input trajectory */
717 /* Skip the first Nose-Hoover integration when we get the state from tpx */
718 bStateFromTPX = !bStateFromCP;
719 bInitStep = bFirstStep && (bStateFromTPX || bVV);
720 bStartingFromCpt = (Flags & MD_STARTFROMCPT) && bInitStep;
721 bSumEkinhOld = FALSE;
723 bNeedRepartition = FALSE;
725 init_global_signals(&gs, cr, ir, repl_ex_nst);
727 step = ir->init_step;
730 init_nlistheuristics(&nlh, bGStatEveryStep, step);
732 if (MULTISIM(cr) && (repl_ex_nst <= 0 ))
734 /* check how many steps are left in other sims */
735 multisim_nsteps = get_multisim_nsteps(cr, ir->nsteps);
739 /* and stop now if we should */
740 bLastStep = (bRerunMD || (ir->nsteps >= 0 && step_rel > ir->nsteps) ||
741 ((multisim_nsteps >= 0) && (step_rel >= multisim_nsteps )));
742 while (!bLastStep || (bRerunMD && bNotLastFrame))
745 wallcycle_start(wcycle, ewcSTEP);
751 step = rerun_fr.step;
752 step_rel = step - ir->init_step;
765 bLastStep = (step_rel == ir->nsteps);
766 t = t0 + step*ir->delta_t;
769 if (ir->efep != efepNO || ir->bSimTemp)
771 /* find and set the current lambdas. If rerunning, we either read in a state, or a lambda value,
772 requiring different logic. */
774 set_current_lambdas(step, ir->fepvals, bRerunMD, &rerun_fr, state_global, state, lam0);
775 bDoDHDL = do_per_step(step, ir->fepvals->nstdhdl);
776 bDoFEP = (do_per_step(step, nstfep) && (ir->efep != efepNO));
777 bDoExpanded = (do_per_step(step, ir->expandedvals->nstexpanded)
778 && (ir->bExpanded) && (step > 0) && (!bStartingFromCpt));
781 bDoReplEx = ((repl_ex_nst > 0) && (step > 0) && !bLastStep &&
782 do_per_step(step, repl_ex_nst));
786 update_annealing_target_temp(&(ir->opts), t);
791 if (!DOMAINDECOMP(cr) || MASTER(cr))
793 for (i = 0; i < state_global->natoms; i++)
795 copy_rvec(rerun_fr.x[i], state_global->x[i]);
799 for (i = 0; i < state_global->natoms; i++)
801 copy_rvec(rerun_fr.v[i], state_global->v[i]);
806 for (i = 0; i < state_global->natoms; i++)
808 clear_rvec(state_global->v[i]);
812 fprintf(stderr, "\nWARNING: Some frames do not contain velocities.\n"
813 " Ekin, temperature and pressure are incorrect,\n"
814 " the virial will be incorrect when constraints are present.\n"
816 bRerunWarnNoV = FALSE;
820 copy_mat(rerun_fr.box, state_global->box);
821 copy_mat(state_global->box, state->box);
823 if (vsite && (Flags & MD_RERUN_VSITE))
825 if (DOMAINDECOMP(cr))
827 gmx_fatal(FARGS, "Vsite recalculation with -rerun is not implemented with domain decomposition, use a single rank");
831 /* Following is necessary because the graph may get out of sync
832 * with the coordinates if we only have every N'th coordinate set
834 mk_mshift(fplog, graph, fr->ePBC, state->box, state->x);
835 shift_self(graph, state->box, state->x);
837 construct_vsites(vsite, state->x, ir->delta_t, state->v,
838 top->idef.iparams, top->idef.il,
839 fr->ePBC, fr->bMolPBC, cr, state->box);
842 unshift_self(graph, state->box, state->x);
847 /* Stop Center of Mass motion */
848 bStopCM = (ir->comm_mode != ecmNO && do_per_step(step, ir->nstcomm));
852 /* for rerun MD always do Neighbour Searching */
853 bNS = (bFirstStep || ir->nstlist != 0);
858 /* Determine whether or not to do Neighbour Searching and LR */
859 bNStList = (ir->nstlist > 0 && step % ir->nstlist == 0);
861 bNS = (bFirstStep || bExchanged || bNeedRepartition || bNStList || bDoFEP ||
862 (ir->nstlist == -1 && nlh.nabnsb > 0));
864 if (bNS && ir->nstlist == -1)
866 set_nlistheuristics(&nlh, bFirstStep || bExchanged || bNeedRepartition || bDoFEP, step);
870 /* check whether we should stop because another simulation has
874 if ( (multisim_nsteps >= 0) && (step_rel >= multisim_nsteps) &&
875 (multisim_nsteps != ir->nsteps) )
882 "Stopping simulation %d because another one has finished\n",
886 gs.sig[eglsCHKPT] = 1;
891 /* < 0 means stop at next step, > 0 means stop at next NS step */
892 if ( (gs.set[eglsSTOPCOND] < 0) ||
893 ( (gs.set[eglsSTOPCOND] > 0) && (bNStList || ir->nstlist == 0) ) )
898 /* Determine whether or not to update the Born radii if doing GB */
899 bBornRadii = bFirstStep;
900 if (ir->implicit_solvent && (step % ir->nstgbradii == 0))
905 do_log = do_per_step(step, ir->nstlog) || bFirstStep || bLastStep;
906 do_verbose = bVerbose &&
907 (step % stepout == 0 || bFirstStep || bLastStep);
909 if (bNS && !(bFirstStep && ir->bContinuation && !bRerunMD))
917 bMasterState = FALSE;
918 /* Correct the new box if it is too skewed */
919 if (DYNAMIC_BOX(*ir))
921 if (correct_box(fplog, step, state->box, graph))
926 if (DOMAINDECOMP(cr) && bMasterState)
928 dd_collect_state(cr->dd, state, state_global);
932 if (DOMAINDECOMP(cr))
934 /* Repartition the domain decomposition */
935 wallcycle_start(wcycle, ewcDOMDEC);
936 dd_partition_system(fplog, step, cr,
937 bMasterState, nstglobalcomm,
938 state_global, top_global, ir,
939 state, &f, mdatoms, top, fr,
940 vsite, shellfc, constr,
942 do_verbose && !bPMETuneRunning);
943 wallcycle_stop(wcycle, ewcDOMDEC);
944 /* If using an iterative integrator, reallocate space to match the decomposition */
948 if (MASTER(cr) && do_log)
950 print_ebin_header(fplog, step, t, state->lambda[efptFEP]); /* can we improve the information printed here? */
953 if (ir->efep != efepNO)
955 update_mdatoms(mdatoms, state->lambda[efptMASS]);
958 if ((bRerunMD && rerun_fr.bV) || bExchanged)
961 /* We need the kinetic energy at minus the half step for determining
962 * the full step kinetic energy and possibly for T-coupling.*/
963 /* This may not be quite working correctly yet . . . . */
964 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
965 wcycle, enerd, NULL, NULL, NULL, NULL, mu_tot,
966 constr, NULL, FALSE, state->box,
967 top_global, &bSumEkinhOld,
968 CGLO_RERUNMD | CGLO_GSTAT | CGLO_TEMPERATURE);
970 clear_mat(force_vir);
972 /* We write a checkpoint at this MD step when:
973 * either at an NS step when we signalled through gs,
974 * or at the last step (but not when we do not want confout),
975 * but never at the first step or with rerun.
977 bCPT = (((gs.set[eglsCHKPT] && (bNS || ir->nstlist == 0)) ||
978 (bLastStep && (Flags & MD_CONFOUT))) &&
979 step > ir->init_step && !bRerunMD);
982 gs.set[eglsCHKPT] = 0;
985 /* Determine the energy and pressure:
986 * at nstcalcenergy steps and at energy output steps (set below).
988 if (EI_VV(ir->eI) && (!bInitStep))
990 /* for vv, the first half of the integration actually corresponds
991 to the previous step. bCalcEner is only required to be evaluated on the 'next' step,
992 but the virial needs to be calculated on both the current step and the 'next' step. Future
993 reorganization may be able to get rid of one of the bCalcVir=TRUE steps. */
995 bCalcEner = do_per_step(step-1, ir->nstcalcenergy);
996 bCalcVir = bCalcEner ||
997 (ir->epc != epcNO && (do_per_step(step, ir->nstpcouple) || do_per_step(step-1, ir->nstpcouple)));
1001 bCalcEner = do_per_step(step, ir->nstcalcenergy);
1002 bCalcVir = bCalcEner ||
1003 (ir->epc != epcNO && do_per_step(step, ir->nstpcouple));
1006 /* Do we need global communication ? */
1007 bGStat = (bCalcVir || bCalcEner || bStopCM ||
1008 do_per_step(step, nstglobalcomm) || (bVV && IR_NVT_TROTTER(ir) && do_per_step(step-1, nstglobalcomm)) ||
1009 (ir->nstlist == -1 && !bRerunMD && step >= nlh.step_nscheck));
1011 do_ene = (do_per_step(step, ir->nstenergy) || bLastStep);
1013 if (do_ene || do_log || bDoReplEx)
1020 /* these CGLO_ options remain the same throughout the iteration */
1021 cglo_flags = ((bRerunMD ? CGLO_RERUNMD : 0) |
1022 (bGStat ? CGLO_GSTAT : 0)
1025 force_flags = (GMX_FORCE_STATECHANGED |
1026 ((DYNAMIC_BOX(*ir) || bRerunMD) ? GMX_FORCE_DYNAMICBOX : 0) |
1027 GMX_FORCE_ALLFORCES |
1029 (bCalcVir ? GMX_FORCE_VIRIAL : 0) |
1030 (bCalcEner ? GMX_FORCE_ENERGY : 0) |
1031 (bDoFEP ? GMX_FORCE_DHDL : 0)
1036 if (do_per_step(step, ir->nstcalclr))
1038 force_flags |= GMX_FORCE_DO_LR;
1044 /* Now is the time to relax the shells */
1045 count = relax_shell_flexcon(fplog, cr, bVerbose, step,
1046 ir, bNS, force_flags,
1049 state, f, force_vir, mdatoms,
1050 nrnb, wcycle, graph, groups,
1051 shellfc, fr, bBornRadii, t, mu_tot,
1053 mdoutf_get_fp_field(outf));
1063 /* The coordinates (x) are shifted (to get whole molecules)
1065 * This is parallellized as well, and does communication too.
1066 * Check comments in sim_util.c
1068 do_force(fplog, cr, ir, step, nrnb, wcycle, top, groups,
1069 state->box, state->x, &state->hist,
1070 f, force_vir, mdatoms, enerd, fcd,
1071 state->lambda, graph,
1072 fr, vsite, mu_tot, t, mdoutf_get_fp_field(outf), ed, bBornRadii,
1073 (bNS ? GMX_FORCE_NS : 0) | force_flags);
1076 if (bVV && !bStartingFromCpt && !bRerunMD)
1077 /* ############### START FIRST UPDATE HALF-STEP FOR VV METHODS############### */
1079 wallcycle_start(wcycle, ewcUPDATE);
1080 if (ir->eI == eiVV && bInitStep)
1082 /* if using velocity verlet with full time step Ekin,
1083 * take the first half step only to compute the
1084 * virial for the first step. From there,
1085 * revert back to the initial coordinates
1086 * so that the input is actually the initial step.
1088 copy_rvecn(state->v, cbuf, 0, state->natoms); /* should make this better for parallelizing? */
1092 /* this is for NHC in the Ekin(t+dt/2) version of vv */
1093 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ1);
1096 /* If we are using twin-range interactions where the long-range component
1097 * is only evaluated every nstcalclr>1 steps, we should do a special update
1098 * step to combine the long-range forces on these steps.
1099 * For nstcalclr=1 this is not done, since the forces would have been added
1100 * directly to the short-range forces already.
1102 * TODO Remove various aspects of VV+twin-range in master
1103 * branch, because VV integrators did not ever support
1104 * twin-range multiple time stepping with constraints.
1106 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1108 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC,
1109 f, bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1110 ekind, M, upd, bInitStep, etrtVELOCITY1,
1111 cr, nrnb, constr, &top->idef);
1113 if (bIterativeCase && do_per_step(step-1, ir->nstpcouple) && !bInitStep)
1115 gmx_iterate_init(&iterate, TRUE);
1117 /* for iterations, we save these vectors, as we will be self-consistently iterating
1120 /*#### UPDATE EXTENDED VARIABLES IN TROTTER FORMULATION */
1122 /* save the state */
1123 if (iterate.bIterationActive)
1125 copy_coupling_state(state, bufstate, ekind, ekind_save, &(ir->opts));
1128 bFirstIterate = TRUE;
1129 while (bFirstIterate || iterate.bIterationActive)
1131 if (iterate.bIterationActive)
1133 copy_coupling_state(bufstate, state, ekind_save, ekind, &(ir->opts));
1134 if (bFirstIterate && bTrotter)
1136 /* The first time through, we need a decent first estimate
1137 of veta(t+dt) to compute the constraints. Do
1138 this by computing the box volume part of the
1139 trotter integration at this time. Nothing else
1140 should be changed by this routine here. If
1141 !(first time), we start with the previous value
1144 veta_save = state->veta;
1145 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ0);
1146 vetanew = state->veta;
1147 state->veta = veta_save;
1151 if (!bRerunMD || rerun_fr.bV || bForceUpdate) /* Why is rerun_fr.bV here? Unclear. */
1153 wallcycle_stop(wcycle, ewcUPDATE);
1154 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1155 state, fr->bMolPBC, graph, f,
1156 &top->idef, shake_vir,
1157 cr, nrnb, wcycle, upd, constr,
1158 TRUE, bCalcVir, vetanew);
1159 wallcycle_start(wcycle, ewcUPDATE);
1161 if (bCalcVir && bUpdateDoLR && ir->nstcalclr > 1)
1163 /* Correct the virial for multiple time stepping */
1164 m_sub(shake_vir, fr->vir_twin_constr, shake_vir);
1169 /* Need to unshift here if a do_force has been
1170 called in the previous step */
1171 unshift_self(graph, state->box, state->x);
1174 /* if VV, compute the pressure and constraints */
1175 /* For VV2, we strictly only need this if using pressure
1176 * control, but we really would like to have accurate pressures
1178 * Think about ways around this in the future?
1179 * For now, keep this choice in comments.
1181 /*bPres = (ir->eI==eiVV || IR_NPT_TROTTER(ir)); */
1182 /*bTemp = ((ir->eI==eiVV &&(!bInitStep)) || (ir->eI==eiVVAK && IR_NPT_TROTTER(ir)));*/
1184 bTemp = ((ir->eI == eiVV && (!bInitStep)) || (ir->eI == eiVVAK));
1185 if (bCalcEner && ir->eI == eiVVAK) /*MRS: 7/9/2010 -- this still doesn't fix it?*/
1187 bSumEkinhOld = TRUE;
1189 /* for vv, the first half of the integration actually corresponds to the previous step.
1190 So we need information from the last step in the first half of the integration */
1191 if (bGStat || do_per_step(step-1, nstglobalcomm))
1193 wallcycle_stop(wcycle, ewcUPDATE);
1194 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1195 wcycle, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
1196 constr, NULL, FALSE, state->box,
1197 top_global, &bSumEkinhOld,
1200 | (bTemp ? CGLO_TEMPERATURE : 0)
1201 | (bPres ? CGLO_PRESSURE : 0)
1202 | (bPres ? CGLO_CONSTRAINT : 0)
1203 | ((iterate.bIterationActive) ? CGLO_ITERATE : 0)
1204 | (bFirstIterate ? CGLO_FIRSTITERATE : 0)
1207 /* explanation of above:
1208 a) We compute Ekin at the full time step
1209 if 1) we are using the AveVel Ekin, and it's not the
1210 initial step, or 2) if we are using AveEkin, but need the full
1211 time step kinetic energy for the pressure (always true now, since we want accurate statistics).
1212 b) If we are using EkinAveEkin for the kinetic energy for the temperature control, we still feed in
1213 EkinAveVel because it's needed for the pressure */
1214 wallcycle_start(wcycle, ewcUPDATE);
1216 /* temperature scaling and pressure scaling to produce the extended variables at t+dt */
1221 m_add(force_vir, shake_vir, total_vir); /* we need the un-dispersion corrected total vir here */
1222 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ2);
1228 wallcycle_stop(wcycle, ewcUPDATE);
1229 /* We need the kinetic energy at minus the half step for determining
1230 * the full step kinetic energy and possibly for T-coupling.*/
1231 /* This may not be quite working correctly yet . . . . */
1232 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1233 wcycle, enerd, NULL, NULL, NULL, NULL, mu_tot,
1234 constr, NULL, FALSE, state->box,
1235 top_global, &bSumEkinhOld,
1236 CGLO_RERUNMD | CGLO_GSTAT | CGLO_TEMPERATURE);
1237 wallcycle_start(wcycle, ewcUPDATE);
1242 if (iterate.bIterationActive &&
1243 done_iterating(cr, fplog, step, &iterate, bFirstIterate,
1244 state->veta, &vetanew))
1248 bFirstIterate = FALSE;
1251 if (bTrotter && !bInitStep)
1253 copy_mat(shake_vir, state->svir_prev);
1254 copy_mat(force_vir, state->fvir_prev);
1255 if (IR_NVT_TROTTER(ir) && ir->eI == eiVV)
1257 /* update temperature and kinetic energy now that step is over - this is the v(t+dt) point */
1258 enerd->term[F_TEMP] = sum_ekin(&(ir->opts), ekind, NULL, (ir->eI == eiVV), FALSE);
1259 enerd->term[F_EKIN] = trace(ekind->ekin);
1262 /* if it's the initial step, we performed this first step just to get the constraint virial */
1263 if (bInitStep && ir->eI == eiVV)
1265 copy_rvecn(cbuf, state->v, 0, state->natoms);
1267 wallcycle_stop(wcycle, ewcUPDATE);
1270 /* MRS -- now done iterating -- compute the conserved quantity */
1273 saved_conserved_quantity = compute_conserved_from_auxiliary(ir, state, &MassQ);
1276 last_ekin = enerd->term[F_EKIN];
1278 if ((ir->eDispCorr != edispcEnerPres) && (ir->eDispCorr != edispcAllEnerPres))
1280 saved_conserved_quantity -= enerd->term[F_DISPCORR];
1282 /* sum up the foreign energy and dhdl terms for vv. currently done every step so that dhdl is correct in the .edr */
1285 sum_dhdl(enerd, state->lambda, ir->fepvals);
1289 /* ######## END FIRST UPDATE STEP ############## */
1290 /* ######## If doing VV, we now have v(dt) ###### */
1293 /* perform extended ensemble sampling in lambda - we don't
1294 actually move to the new state before outputting
1295 statistics, but if performing simulated tempering, we
1296 do update the velocities and the tau_t. */
1298 lamnew = ExpandedEnsembleDynamics(fplog, ir, enerd, state, &MassQ, state->fep_state, &state->dfhist, step, state->v, mdatoms);
1299 /* history is maintained in state->dfhist, but state_global is what is sent to trajectory and log output */
1300 copy_df_history(&state_global->dfhist, &state->dfhist);
1303 /* Now we have the energies and forces corresponding to the
1304 * coordinates at time t. We must output all of this before
1307 do_md_trajectory_writing(fplog, cr, nfile, fnm, step, step_rel, t,
1308 ir, state, state_global, top_global, fr,
1309 outf, mdebin, ekind, f, f_global,
1311 bCPT, bRerunMD, bLastStep, (Flags & MD_CONFOUT),
1313 /* Check if IMD step and do IMD communication, if bIMD is TRUE. */
1314 bIMDstep = do_IMD(ir->bIMD, step, cr, bNS, state->box, state->x, ir, t, wcycle);
1316 /* kludge -- virial is lost with restart for NPT control. Must restart */
1317 if (bStartingFromCpt && bVV)
1319 copy_mat(state->svir_prev, shake_vir);
1320 copy_mat(state->fvir_prev, force_vir);
1323 elapsed_time = walltime_accounting_get_current_elapsed_time(walltime_accounting);
1325 /* Check whether everything is still allright */
1326 if (((int)gmx_get_stop_condition() > handled_stop_condition)
1327 #ifdef GMX_THREAD_MPI
1332 /* this is just make gs.sig compatible with the hack
1333 of sending signals around by MPI_Reduce with together with
1335 if (gmx_get_stop_condition() == gmx_stop_cond_next_ns)
1337 gs.sig[eglsSTOPCOND] = 1;
1339 if (gmx_get_stop_condition() == gmx_stop_cond_next)
1341 gs.sig[eglsSTOPCOND] = -1;
1343 /* < 0 means stop at next step, > 0 means stop at next NS step */
1347 "\n\nReceived the %s signal, stopping at the next %sstep\n\n",
1348 gmx_get_signal_name(),
1349 gs.sig[eglsSTOPCOND] == 1 ? "NS " : "");
1353 "\n\nReceived the %s signal, stopping at the next %sstep\n\n",
1354 gmx_get_signal_name(),
1355 gs.sig[eglsSTOPCOND] == 1 ? "NS " : "");
1357 handled_stop_condition = (int)gmx_get_stop_condition();
1359 else if (MASTER(cr) && (bNS || ir->nstlist <= 0) &&
1360 (max_hours > 0 && elapsed_time > max_hours*60.0*60.0*0.99) &&
1361 gs.sig[eglsSTOPCOND] == 0 && gs.set[eglsSTOPCOND] == 0)
1363 /* Signal to terminate the run */
1364 gs.sig[eglsSTOPCOND] = 1;
1367 fprintf(fplog, "\nStep %s: Run time exceeded %.3f hours, will terminate the run\n", gmx_step_str(step, sbuf), max_hours*0.99);
1369 fprintf(stderr, "\nStep %s: Run time exceeded %.3f hours, will terminate the run\n", gmx_step_str(step, sbuf), max_hours*0.99);
1372 if (bResetCountersHalfMaxH && MASTER(cr) &&
1373 elapsed_time > max_hours*60.0*60.0*0.495)
1375 gs.sig[eglsRESETCOUNTERS] = 1;
1378 if (ir->nstlist == -1 && !bRerunMD)
1380 /* When bGStatEveryStep=FALSE, global_stat is only called
1381 * when we check the atom displacements, not at NS steps.
1382 * This means that also the bonded interaction count check is not
1383 * performed immediately after NS. Therefore a few MD steps could
1384 * be performed with missing interactions.
1385 * But wrong energies are never written to file,
1386 * since energies are only written after global_stat
1389 if (step >= nlh.step_nscheck)
1391 nlh.nabnsb = natoms_beyond_ns_buffer(ir, fr, &top->cgs,
1392 nlh.scale_tot, state->x);
1396 /* This is not necessarily true,
1397 * but step_nscheck is determined quite conservatively.
1403 /* In parallel we only have to check for checkpointing in steps
1404 * where we do global communication,
1405 * otherwise the other nodes don't know.
1407 if (MASTER(cr) && ((bGStat || !PAR(cr)) &&
1410 elapsed_time >= nchkpt*cpt_period*60.0)) &&
1411 gs.set[eglsCHKPT] == 0)
1413 gs.sig[eglsCHKPT] = 1;
1416 /* at the start of step, randomize or scale the velocities (trotter done elsewhere) */
1421 update_tcouple(step, ir, state, ekind, &MassQ, mdatoms);
1423 if (ETC_ANDERSEN(ir->etc)) /* keep this outside of update_tcouple because of the extra info required to pass */
1425 gmx_bool bIfRandomize;
1426 bIfRandomize = update_randomize_velocities(ir, step, cr, mdatoms, state, upd, constr);
1427 /* if we have constraints, we have to remove the kinetic energy parallel to the bonds */
1428 if (constr && bIfRandomize)
1430 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1431 state, fr->bMolPBC, graph, f,
1432 &top->idef, tmp_vir,
1433 cr, nrnb, wcycle, upd, constr,
1434 TRUE, bCalcVir, vetanew);
1439 if (bIterativeCase && do_per_step(step, ir->nstpcouple))
1441 gmx_iterate_init(&iterate, TRUE);
1442 /* for iterations, we save these vectors, as we will be redoing the calculations */
1443 copy_coupling_state(state, bufstate, ekind, ekind_save, &(ir->opts));
1446 bFirstIterate = TRUE;
1447 while (bFirstIterate || iterate.bIterationActive)
1449 /* We now restore these vectors to redo the calculation with improved extended variables */
1450 if (iterate.bIterationActive)
1452 copy_coupling_state(bufstate, state, ekind_save, ekind, &(ir->opts));
1455 /* We make the decision to break or not -after- the calculation of Ekin and Pressure,
1456 so scroll down for that logic */
1458 /* ######### START SECOND UPDATE STEP ################# */
1459 /* Box is changed in update() when we do pressure coupling,
1460 * but we should still use the old box for energy corrections and when
1461 * writing it to the energy file, so it matches the trajectory files for
1462 * the same timestep above. Make a copy in a separate array.
1464 copy_mat(state->box, lastbox);
1468 if (!(bRerunMD && !rerun_fr.bV && !bForceUpdate))
1470 wallcycle_start(wcycle, ewcUPDATE);
1471 /* UPDATE PRESSURE VARIABLES IN TROTTER FORMULATION WITH CONSTRAINTS */
1474 if (iterate.bIterationActive)
1482 /* we use a new value of scalevir to converge the iterations faster */
1483 scalevir = tracevir/trace(shake_vir);
1485 msmul(shake_vir, scalevir, shake_vir);
1486 m_add(force_vir, shake_vir, total_vir);
1487 clear_mat(shake_vir);
1489 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ3);
1490 /* We can only do Berendsen coupling after we have summed
1491 * the kinetic energy or virial. Since the happens
1492 * in global_state after update, we should only do it at
1493 * step % nstlist = 1 with bGStatEveryStep=FALSE.
1498 update_tcouple(step, ir, state, ekind, &MassQ, mdatoms);
1499 update_pcouple(fplog, step, ir, state, pcoupl_mu, M, bInitStep);
1504 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1506 /* velocity half-step update */
1507 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1508 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1509 ekind, M, upd, FALSE, etrtVELOCITY2,
1510 cr, nrnb, constr, &top->idef);
1513 /* Above, initialize just copies ekinh into ekin,
1514 * it doesn't copy position (for VV),
1515 * and entire integrator for MD.
1518 if (ir->eI == eiVVAK)
1520 copy_rvecn(state->x, cbuf, 0, state->natoms);
1522 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1524 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1525 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1526 ekind, M, upd, bInitStep, etrtPOSITION, cr, nrnb, constr, &top->idef);
1527 wallcycle_stop(wcycle, ewcUPDATE);
1529 update_constraints(fplog, step, &dvdl_constr, ir, ekind, mdatoms, state,
1530 fr->bMolPBC, graph, f,
1531 &top->idef, shake_vir,
1532 cr, nrnb, wcycle, upd, constr,
1533 FALSE, bCalcVir, state->veta);
1535 if (bCalcVir && bUpdateDoLR && ir->nstcalclr > 1)
1537 /* Correct the virial for multiple time stepping */
1538 m_sub(shake_vir, fr->vir_twin_constr, shake_vir);
1541 if (ir->eI == eiVVAK)
1543 /* erase F_EKIN and F_TEMP here? */
1544 /* just compute the kinetic energy at the half step to perform a trotter step */
1545 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1546 wcycle, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
1547 constr, NULL, FALSE, lastbox,
1548 top_global, &bSumEkinhOld,
1549 cglo_flags | CGLO_TEMPERATURE
1551 wallcycle_start(wcycle, ewcUPDATE);
1552 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ4);
1553 /* now we know the scaling, we can compute the positions again again */
1554 copy_rvecn(cbuf, state->x, 0, state->natoms);
1556 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1558 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1559 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1560 ekind, M, upd, bInitStep, etrtPOSITION, cr, nrnb, constr, &top->idef);
1561 wallcycle_stop(wcycle, ewcUPDATE);
1563 /* do we need an extra constraint here? just need to copy out of state->v to upd->xp? */
1564 /* are the small terms in the shake_vir here due
1565 * to numerical errors, or are they important
1566 * physically? I'm thinking they are just errors, but not completely sure.
1567 * For now, will call without actually constraining, constr=NULL*/
1568 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1569 state, fr->bMolPBC, graph, f,
1570 &top->idef, tmp_vir,
1571 cr, nrnb, wcycle, upd, NULL,
1578 /* this factor or 2 correction is necessary
1579 because half of the constraint force is removed
1580 in the vv step, so we have to double it. See
1581 the Redmine issue #1255. It is not yet clear
1582 if the factor of 2 is exact, or just a very
1583 good approximation, and this will be
1584 investigated. The next step is to see if this
1585 can be done adding a dhdl contribution from the
1586 rattle step, but this is somewhat more
1587 complicated with the current code. Will be
1588 investigated, hopefully for 4.6.3. However,
1589 this current solution is much better than
1590 having it completely wrong.
1592 enerd->term[F_DVDL_CONSTR] += 2*dvdl_constr;
1596 enerd->term[F_DVDL_CONSTR] += dvdl_constr;
1601 /* Need to unshift here */
1602 unshift_self(graph, state->box, state->x);
1607 wallcycle_start(wcycle, ewcVSITECONSTR);
1610 shift_self(graph, state->box, state->x);
1612 construct_vsites(vsite, state->x, ir->delta_t, state->v,
1613 top->idef.iparams, top->idef.il,
1614 fr->ePBC, fr->bMolPBC, cr, state->box);
1618 unshift_self(graph, state->box, state->x);
1620 wallcycle_stop(wcycle, ewcVSITECONSTR);
1623 /* ############## IF NOT VV, Calculate globals HERE, also iterate constraints ############ */
1624 /* With Leap-Frog we can skip compute_globals at
1625 * non-communication steps, but we need to calculate
1626 * the kinetic energy one step before communication.
1628 if (bGStat || (!EI_VV(ir->eI) && do_per_step(step+1, nstglobalcomm)))
1630 if (ir->nstlist == -1 && bFirstIterate)
1632 gs.sig[eglsNABNSB] = nlh.nabnsb;
1634 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1635 wcycle, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
1637 bFirstIterate ? &gs : NULL,
1638 (step_rel % gs.nstms == 0) &&
1639 (multisim_nsteps < 0 || (step_rel < multisim_nsteps)),
1641 top_global, &bSumEkinhOld,
1643 | (!EI_VV(ir->eI) || bRerunMD ? CGLO_ENERGY : 0)
1644 | (!EI_VV(ir->eI) && bStopCM ? CGLO_STOPCM : 0)
1645 | (!EI_VV(ir->eI) ? CGLO_TEMPERATURE : 0)
1646 | (!EI_VV(ir->eI) || bRerunMD ? CGLO_PRESSURE : 0)
1647 | (iterate.bIterationActive ? CGLO_ITERATE : 0)
1648 | (bFirstIterate ? CGLO_FIRSTITERATE : 0)
1651 if (ir->nstlist == -1 && bFirstIterate)
1653 nlh.nabnsb = gs.set[eglsNABNSB];
1654 gs.set[eglsNABNSB] = 0;
1657 /* bIterate is set to keep it from eliminating the old ekin kinetic energy terms */
1658 /* ############# END CALC EKIN AND PRESSURE ################# */
1660 /* Note: this is OK, but there are some numerical precision issues with using the convergence of
1661 the virial that should probably be addressed eventually. state->veta has better properies,
1662 but what we actually need entering the new cycle is the new shake_vir value. Ideally, we could
1663 generate the new shake_vir, but test the veta value for convergence. This will take some thought. */
1665 if (iterate.bIterationActive &&
1666 done_iterating(cr, fplog, step, &iterate, bFirstIterate,
1667 trace(shake_vir), &tracevir))
1671 bFirstIterate = FALSE;
1674 if (!bVV || bRerunMD)
1676 /* sum up the foreign energy and dhdl terms for md and sd. currently done every step so that dhdl is correct in the .edr */
1677 sum_dhdl(enerd, state->lambda, ir->fepvals);
1679 update_box(fplog, step, ir, mdatoms, state, f,
1680 ir->nstlist == -1 ? &nlh.scale_tot : NULL, pcoupl_mu, nrnb, upd);
1682 /* ################# END UPDATE STEP 2 ################# */
1683 /* #### We now have r(t+dt) and v(t+dt/2) ############# */
1685 /* The coordinates (x) were unshifted in update */
1688 /* We will not sum ekinh_old,
1689 * so signal that we still have to do it.
1691 bSumEkinhOld = TRUE;
1694 /* ######### BEGIN PREPARING EDR OUTPUT ########### */
1696 /* use the directly determined last velocity, not actually the averaged half steps */
1697 if (bTrotter && ir->eI == eiVV)
1699 enerd->term[F_EKIN] = last_ekin;
1701 enerd->term[F_ETOT] = enerd->term[F_EPOT] + enerd->term[F_EKIN];
1705 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + saved_conserved_quantity;
1709 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + compute_conserved_from_auxiliary(ir, state, &MassQ);
1711 /* ######### END PREPARING EDR OUTPUT ########### */
1716 gmx_bool do_dr, do_or;
1718 if (fplog && do_log && bDoExpanded)
1720 /* only needed if doing expanded ensemble */
1721 PrintFreeEnergyInfoToFile(fplog, ir->fepvals, ir->expandedvals, ir->bSimTemp ? ir->simtempvals : NULL,
1722 &state_global->dfhist, state->fep_state, ir->nstlog, step);
1724 if (!(bStartingFromCpt && (EI_VV(ir->eI))))
1728 upd_mdebin(mdebin, bDoDHDL, TRUE,
1729 t, mdatoms->tmass, enerd, state,
1730 ir->fepvals, ir->expandedvals, lastbox,
1731 shake_vir, force_vir, total_vir, pres,
1732 ekind, mu_tot, constr);
1736 upd_mdebin_step(mdebin);
1739 do_dr = do_per_step(step, ir->nstdisreout);
1740 do_or = do_per_step(step, ir->nstorireout);
1742 print_ebin(mdoutf_get_fp_ene(outf), do_ene, do_dr, do_or, do_log ? fplog : NULL,
1744 eprNORMAL, bCompact, mdebin, fcd, groups, &(ir->opts));
1746 if (ir->ePull != epullNO)
1748 pull_print_output(ir->pull, step, t);
1751 if (do_per_step(step, ir->nstlog))
1753 if (fflush(fplog) != 0)
1755 gmx_fatal(FARGS, "Cannot flush logfile - maybe you are out of disk space?");
1761 /* Have to do this part _after_ outputting the logfile and the edr file */
1762 /* Gets written into the state at the beginning of next loop*/
1763 state->fep_state = lamnew;
1765 /* Print the remaining wall clock time for the run */
1766 if (MULTIMASTER(cr) && (do_verbose || gmx_got_usr_signal()) && !bPMETuneRunning)
1770 fprintf(stderr, "\n");
1772 print_time(stderr, walltime_accounting, step, ir, cr);
1775 /* Ion/water position swapping.
1776 * Not done in last step since trajectory writing happens before this call
1777 * in the MD loop and exchanges would be lost anyway. */
1778 bNeedRepartition = FALSE;
1779 if ((ir->eSwapCoords != eswapNO) && (step > 0) && !bLastStep &&
1780 do_per_step(step, ir->swap->nstswap))
1782 bNeedRepartition = do_swapcoords(cr, step, t, ir, wcycle,
1783 bRerunMD ? rerun_fr.x : state->x,
1784 bRerunMD ? rerun_fr.box : state->box,
1785 top_global, MASTER(cr) && bVerbose, bRerunMD);
1787 if (bNeedRepartition && DOMAINDECOMP(cr))
1789 dd_collect_state(cr->dd, state, state_global);
1793 /* Replica exchange */
1797 bExchanged = replica_exchange(fplog, cr, repl_ex,
1798 state_global, enerd,
1802 if ( (bExchanged || bNeedRepartition) && DOMAINDECOMP(cr) )
1804 dd_partition_system(fplog, step, cr, TRUE, 1,
1805 state_global, top_global, ir,
1806 state, &f, mdatoms, top, fr,
1807 vsite, shellfc, constr,
1808 nrnb, wcycle, FALSE);
1813 bStartingFromCpt = FALSE;
1815 /* ####### SET VARIABLES FOR NEXT ITERATION IF THEY STILL NEED IT ###### */
1816 /* With all integrators, except VV, we need to retain the pressure
1817 * at the current step for coupling at the next step.
1819 if ((state->flags & (1<<estPRES_PREV)) &&
1821 (ir->nstpcouple > 0 && step % ir->nstpcouple == 0)))
1823 /* Store the pressure in t_state for pressure coupling
1824 * at the next MD step.
1826 copy_mat(pres, state->pres_prev);
1829 /* ####### END SET VARIABLES FOR NEXT ITERATION ###### */
1831 if ( (membed != NULL) && (!bLastStep) )
1833 rescale_membed(step_rel, membed, state_global->x);
1840 /* read next frame from input trajectory */
1841 bNotLastFrame = read_next_frame(oenv, status, &rerun_fr);
1846 rerun_parallel_comm(cr, &rerun_fr, &bNotLastFrame);
1850 if (!bRerunMD || !rerun_fr.bStep)
1852 /* increase the MD step number */
1857 cycles = wallcycle_stop(wcycle, ewcSTEP);
1858 if (DOMAINDECOMP(cr) && wcycle)
1860 dd_cycles_add(cr->dd, cycles, ddCyclStep);
1863 if (bPMETuneRunning || bPMETuneTry)
1865 /* PME grid + cut-off optimization with GPUs or PME nodes */
1867 /* Count the total cycles over the last steps */
1868 cycles_pmes += cycles;
1870 /* We can only switch cut-off at NS steps */
1871 if (step % ir->nstlist == 0)
1873 /* PME grid + cut-off optimization with GPUs or PME nodes */
1876 if (DDMASTER(cr->dd))
1878 /* PME node load is too high, start tuning */
1879 bPMETuneRunning = (dd_pme_f_ratio(cr->dd) >= 1.05);
1881 dd_bcast(cr->dd, sizeof(gmx_bool), &bPMETuneRunning);
1883 if (bPMETuneRunning &&
1884 use_GPU(fr->nbv) && DOMAINDECOMP(cr) &&
1885 !(cr->duty & DUTY_PME))
1887 /* Lock DLB=auto to off (does nothing when DLB=yes/no).
1888 * With GPUs + separate PME ranks, we don't want DLB.
1889 * This could happen when we scan coarse grids and
1890 * it would then never be turned off again.
1891 * This would hurt performance at the final, optimal
1892 * grid spacing, where DLB almost never helps.
1893 * Also, DLB can limit the cut-off for PME tuning.
1895 dd_dlb_set_lock(cr->dd, TRUE);
1898 if (bPMETuneRunning || step_rel > ir->nstlist*50)
1900 bPMETuneTry = FALSE;
1903 if (bPMETuneRunning)
1905 /* init_step might not be a multiple of nstlist,
1906 * but the first cycle is always skipped anyhow.
1909 pme_load_balance(pme_loadbal, cr,
1910 (bVerbose && MASTER(cr)) ? stderr : NULL,
1912 ir, state, cycles_pmes,
1913 fr->ic, fr->nbv, &fr->pmedata,
1916 /* Update constants in forcerec/inputrec to keep them in sync with fr->ic */
1917 fr->ewaldcoeff_q = fr->ic->ewaldcoeff_q;
1918 fr->ewaldcoeff_lj = fr->ic->ewaldcoeff_lj;
1919 fr->rlist = fr->ic->rlist;
1920 fr->rlistlong = fr->ic->rlistlong;
1921 fr->rcoulomb = fr->ic->rcoulomb;
1922 fr->rvdw = fr->ic->rvdw;
1924 if (ir->eDispCorr != edispcNO)
1926 calc_enervirdiff(NULL, ir->eDispCorr, fr);
1929 if (!bPMETuneRunning &&
1931 dd_dlb_is_locked(cr->dd))
1933 /* Unlock the DLB=auto, DLB is allowed to activate
1934 * (but we don't expect it to activate in most cases).
1936 dd_dlb_set_lock(cr->dd, FALSE);
1943 if (step_rel == wcycle_get_reset_counters(wcycle) ||
1944 gs.set[eglsRESETCOUNTERS] != 0)
1946 /* Reset all the counters related to performance over the run */
1947 reset_all_counters(fplog, cr, step, &step_rel, ir, wcycle, nrnb, walltime_accounting,
1948 use_GPU(fr->nbv) ? fr->nbv : NULL);
1949 wcycle_set_reset_counters(wcycle, -1);
1950 if (!(cr->duty & DUTY_PME))
1952 /* Tell our PME node to reset its counters */
1953 gmx_pme_send_resetcounters(cr, step);
1955 /* Correct max_hours for the elapsed time */
1956 max_hours -= elapsed_time/(60.0*60.0);
1957 bResetCountersHalfMaxH = FALSE;
1958 gs.set[eglsRESETCOUNTERS] = 0;
1961 /* If bIMD is TRUE, the master updates the IMD energy record and sends positions to VMD client */
1962 IMD_prep_energies_send_positions(ir->bIMD && MASTER(cr), bIMDstep, ir->imd, enerd, step, bCalcEner, wcycle);
1965 /* End of main MD loop */
1968 /* Closing TNG files can include compressing data. Therefore it is good to do that
1969 * before stopping the time measurements. */
1970 mdoutf_tng_close(outf);
1972 /* Stop measuring walltime */
1973 walltime_accounting_end(walltime_accounting);
1975 if (bRerunMD && MASTER(cr))
1980 if (!(cr->duty & DUTY_PME))
1982 /* Tell the PME only node to finish */
1983 gmx_pme_send_finish(cr);
1988 if (ir->nstcalcenergy > 0 && !bRerunMD)
1990 print_ebin(mdoutf_get_fp_ene(outf), FALSE, FALSE, FALSE, fplog, step, t,
1991 eprAVER, FALSE, mdebin, fcd, groups, &(ir->opts));
1998 if (ir->nstlist == -1 && nlh.nns > 0 && fplog)
2000 fprintf(fplog, "Average neighborlist lifetime: %.1f steps, std.dev.: %.1f steps\n", nlh.s1/nlh.nns, sqrt(nlh.s2/nlh.nns - sqr(nlh.s1/nlh.nns)));
2001 fprintf(fplog, "Average number of atoms that crossed the half buffer length: %.1f\n\n", nlh.ab/nlh.nns);
2004 if (pme_loadbal != NULL)
2006 pme_loadbal_done(pme_loadbal, cr, fplog,
2010 if (shellfc && fplog)
2012 fprintf(fplog, "Fraction of iterations that converged: %.2f %%\n",
2013 (nconverged*100.0)/step_rel);
2014 fprintf(fplog, "Average number of force evaluations per MD step: %.2f\n\n",
2018 if (repl_ex_nst > 0 && MASTER(cr))
2020 print_replica_exchange_statistics(fplog, repl_ex);
2023 /* IMD cleanup, if bIMD is TRUE. */
2024 IMD_finalize(ir->bIMD, ir->imd);
2026 walltime_accounting_set_nsteps_done(walltime_accounting, step_rel);