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.
44 #include "gromacs/math/vec.h"
53 #include "md_support.h"
54 #include "md_logging.h"
66 #include "domdec_network.h"
70 #include "gromacs/gmxpreprocess/compute_io.h"
71 #include "checkpoint.h"
72 #include "gromacs/topology/mtop_util.h"
73 #include "sighandler.h"
75 #include "gromacs/utility/cstringutil.h"
76 #include "pme_loadbal.h"
79 #include "types/nlistheuristics.h"
80 #include "types/iteratedconstraints.h"
81 #include "nbnxn_cuda_data_mgmt.h"
83 #include "gromacs/fileio/confio.h"
84 #include "gromacs/fileio/mdoutf.h"
85 #include "gromacs/fileio/trajectory_writing.h"
86 #include "gromacs/fileio/trnio.h"
87 #include "gromacs/fileio/trxio.h"
88 #include "gromacs/fileio/xtcio.h"
89 #include "gromacs/imd/imd.h"
90 #include "gromacs/pbcutil/mshift.h"
91 #include "gromacs/pbcutil/pbc.h"
92 #include "gromacs/pulling/pull.h"
93 #include "gromacs/swap/swapcoords.h"
94 #include "gromacs/timing/wallcycle.h"
95 #include "gromacs/timing/walltime_accounting.h"
96 #include "gromacs/utility/gmxmpi.h"
97 #include "gromacs/utility/smalloc.h"
100 #include "corewrap.h"
103 static void reset_all_counters(FILE *fplog, t_commrec *cr,
105 gmx_int64_t *step_rel, t_inputrec *ir,
106 gmx_wallcycle_t wcycle, t_nrnb *nrnb,
107 gmx_walltime_accounting_t walltime_accounting,
108 nbnxn_cuda_ptr_t cu_nbv)
110 char sbuf[STEPSTRSIZE];
112 /* Reset all the counters related to performance over the run */
113 md_print_warn(cr, fplog, "step %s: resetting all time and cycle counters\n",
114 gmx_step_str(step, sbuf));
118 nbnxn_cuda_reset_timings(cu_nbv);
121 wallcycle_stop(wcycle, ewcRUN);
122 wallcycle_reset_all(wcycle);
123 if (DOMAINDECOMP(cr))
125 reset_dd_statistics_counters(cr->dd);
128 ir->init_step += *step_rel;
129 ir->nsteps -= *step_rel;
131 wallcycle_start(wcycle, ewcRUN);
132 walltime_accounting_start(walltime_accounting);
133 print_date_and_time(fplog, cr->nodeid, "Restarted time", gmx_gettime());
136 double do_md(FILE *fplog, t_commrec *cr, int nfile, const t_filenm fnm[],
137 const output_env_t oenv, gmx_bool bVerbose, gmx_bool bCompact,
139 gmx_vsite_t *vsite, gmx_constr_t constr,
140 int stepout, t_inputrec *ir,
141 gmx_mtop_t *top_global,
143 t_state *state_global,
145 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
146 gmx_edsam_t ed, t_forcerec *fr,
147 int repl_ex_nst, int repl_ex_nex, int repl_ex_seed, gmx_membed_t membed,
148 real cpt_period, real max_hours,
149 const char gmx_unused *deviceOptions,
152 gmx_walltime_accounting_t walltime_accounting)
154 gmx_mdoutf_t outf = NULL;
155 gmx_int64_t step, step_rel;
157 double t, t0, lam0[efptNR];
158 gmx_bool bGStatEveryStep, bGStat, bCalcVir, bCalcEner;
159 gmx_bool bNS, bNStList, bSimAnn, bStopCM, bRerunMD, bNotLastFrame = FALSE,
160 bFirstStep, bStateFromCP, bStateFromTPX, bInitStep, bLastStep,
161 bBornRadii, bStartingFromCpt;
162 gmx_bool bDoDHDL = FALSE, bDoFEP = FALSE, bDoExpanded = FALSE;
163 gmx_bool do_ene, do_log, do_verbose, bRerunWarnNoV = TRUE,
164 bForceUpdate = FALSE, bCPT;
165 gmx_bool bMasterState;
166 int force_flags, cglo_flags;
167 tensor force_vir, shake_vir, total_vir, tmp_vir, pres;
172 t_state *bufstate = NULL;
176 gmx_repl_ex_t repl_ex = NULL;
179 t_mdebin *mdebin = NULL;
180 t_state *state = NULL;
181 rvec *f_global = NULL;
182 gmx_enerdata_t *enerd;
184 gmx_global_stat_t gstat;
185 gmx_update_t upd = NULL;
186 t_graph *graph = NULL;
188 gmx_groups_t *groups;
189 gmx_ekindata_t *ekind, *ekind_save;
190 gmx_shellfc_t shellfc;
191 int count, nconverged = 0;
193 gmx_bool bConverged = TRUE, bSumEkinhOld, bDoReplEx, bExchanged, bNeedRepartition;
194 gmx_bool bResetCountersHalfMaxH = FALSE;
195 gmx_bool bVV, bIterativeCase, bFirstIterate, bTemp, bPres, bTrotter;
196 gmx_bool bUpdateDoLR;
200 real veta_save, scalevir, tracevir;
206 real saved_conserved_quantity = 0;
210 char sbuf[STEPSTRSIZE], sbuf2[STEPSTRSIZE];
211 int handled_stop_condition = gmx_stop_cond_none; /* compare to get_stop_condition*/
212 gmx_iterate_t iterate;
213 gmx_int64_t multisim_nsteps = -1; /* number of steps to do before first multisim
214 simulation stops. If equal to zero, don't
215 communicate any more between multisims.*/
216 /* PME load balancing data for GPU kernels */
217 pme_load_balancing_t pme_loadbal = NULL;
219 gmx_bool bPMETuneTry = FALSE, bPMETuneRunning = FALSE;
222 gmx_bool bIMDstep = FALSE;
225 /* Temporary addition for FAHCORE checkpointing */
229 /* Check for special mdrun options */
230 bRerunMD = (Flags & MD_RERUN);
231 if (Flags & MD_RESETCOUNTERSHALFWAY)
235 /* Signal to reset the counters half the simulation steps. */
236 wcycle_set_reset_counters(wcycle, ir->nsteps/2);
238 /* Signal to reset the counters halfway the simulation time. */
239 bResetCountersHalfMaxH = (max_hours > 0);
242 /* md-vv uses averaged full step velocities for T-control
243 md-vv-avek uses averaged half step velocities for T-control (but full step ekin for P control)
244 md uses averaged half step kinetic energies to determine temperature unless defined otherwise by GMX_EKIN_AVE_VEL; */
246 if (bVV) /* to store the initial velocities while computing virial */
248 snew(cbuf, top_global->natoms);
250 /* all the iteratative cases - only if there are constraints */
251 bIterativeCase = ((IR_NPH_TROTTER(ir) || IR_NPT_TROTTER(ir)) && (constr) && (!bRerunMD));
252 gmx_iterate_init(&iterate, FALSE); /* The default value of iterate->bIterationActive is set to
253 false in this step. The correct value, true or false,
254 is set at each step, as it depends on the frequency of temperature
255 and pressure control.*/
256 bTrotter = (bVV && (IR_NPT_TROTTER(ir) || IR_NPH_TROTTER(ir) || IR_NVT_TROTTER(ir)));
260 /* Since we don't know if the frames read are related in any way,
261 * rebuild the neighborlist at every step.
264 ir->nstcalcenergy = 1;
268 check_ir_old_tpx_versions(cr, fplog, ir, top_global);
270 nstglobalcomm = check_nstglobalcomm(fplog, cr, nstglobalcomm, ir);
271 bGStatEveryStep = (nstglobalcomm == 1);
273 if (!bGStatEveryStep && ir->nstlist == -1 && fplog != NULL)
276 "To reduce the energy communication with nstlist = -1\n"
277 "the neighbor list validity should not be checked at every step,\n"
278 "this means that exact integration is not guaranteed.\n"
279 "The neighbor list validity is checked after:\n"
280 " <n.list life time> - 2*std.dev.(n.list life time) steps.\n"
281 "In most cases this will result in exact integration.\n"
282 "This reduces the energy communication by a factor of 2 to 3.\n"
283 "If you want less energy communication, set nstlist > 3.\n\n");
288 ir->nstxout_compressed = 0;
290 groups = &top_global->groups;
293 init_md(fplog, cr, ir, oenv, &t, &t0, state_global->lambda,
294 &(state_global->fep_state), lam0,
295 nrnb, top_global, &upd,
296 nfile, fnm, &outf, &mdebin,
297 force_vir, shake_vir, mu_tot, &bSimAnn, &vcm, Flags);
299 clear_mat(total_vir);
301 /* Energy terms and groups */
303 init_enerdata(top_global->groups.grps[egcENER].nr, ir->fepvals->n_lambda,
305 if (DOMAINDECOMP(cr))
311 snew(f, top_global->natoms);
314 /* Kinetic energy data */
316 init_ekindata(fplog, top_global, &(ir->opts), ekind);
317 /* needed for iteration of constraints */
319 init_ekindata(fplog, top_global, &(ir->opts), ekind_save);
320 /* Copy the cos acceleration to the groups struct */
321 ekind->cosacc.cos_accel = ir->cos_accel;
323 gstat = global_stat_init(ir);
326 /* Check for polarizable models and flexible constraints */
327 shellfc = init_shell_flexcon(fplog,
328 top_global, n_flexible_constraints(constr),
329 (ir->bContinuation ||
330 (DOMAINDECOMP(cr) && !MASTER(cr))) ?
331 NULL : state_global->x);
332 if (shellfc && ir->nstcalcenergy != 1)
334 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);
336 if (shellfc && DOMAINDECOMP(cr))
338 gmx_fatal(FARGS, "Shell particles are not implemented with domain decomposition, use a single rank");
340 if (shellfc && ir->eI == eiNM)
342 /* Currently shells don't work with Normal Modes */
343 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");
346 if (vsite && ir->eI == eiNM)
348 /* Currently virtual sites don't work with Normal Modes */
349 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");
354 tMPI_Thread_mutex_lock(&deform_init_box_mutex);
355 set_deform_reference_box(upd,
356 deform_init_init_step_tpx,
357 deform_init_box_tpx);
358 tMPI_Thread_mutex_unlock(&deform_init_box_mutex);
362 double io = compute_io(ir, top_global->natoms, groups, mdebin->ebin->nener, 1);
363 if ((io > 2000) && MASTER(cr))
366 "\nWARNING: This run will generate roughly %.0f Mb of data\n\n",
371 if (DOMAINDECOMP(cr))
373 top = dd_init_local_top(top_global);
376 dd_init_local_state(cr->dd, state_global, state);
378 if (DDMASTER(cr->dd) && ir->nstfout)
380 snew(f_global, state_global->natoms);
385 top = gmx_mtop_generate_local_top(top_global, ir);
387 forcerec_set_excl_load(fr, top);
389 state = serial_init_local_state(state_global);
392 atoms2md(top_global, ir, 0, NULL, top_global->natoms, mdatoms);
396 set_vsite_top(vsite, top, mdatoms, cr);
399 if (ir->ePBC != epbcNONE && !fr->bMolPBC)
401 graph = mk_graph(fplog, &(top->idef), 0, top_global->natoms, FALSE, FALSE);
406 make_local_shells(cr, mdatoms, shellfc);
409 setup_bonded_threading(fr, &top->idef);
412 /* Set up interactive MD (IMD) */
413 init_IMD(ir, cr, top_global, fplog, ir->nstcalcenergy, state_global->x,
414 nfile, fnm, oenv, imdport, Flags);
416 if (DOMAINDECOMP(cr))
418 /* Distribute the charge groups over the nodes from the master node */
419 dd_partition_system(fplog, ir->init_step, cr, TRUE, 1,
420 state_global, top_global, ir,
421 state, &f, mdatoms, top, fr,
422 vsite, shellfc, constr,
423 nrnb, wcycle, FALSE);
427 update_mdatoms(mdatoms, state->lambda[efptMASS]);
429 if (opt2bSet("-cpi", nfile, fnm))
431 bStateFromCP = gmx_fexist_master(opt2fn_master("-cpi", nfile, fnm, cr), cr);
435 bStateFromCP = FALSE;
440 init_expanded_ensemble(bStateFromCP, ir, &state->dfhist);
447 /* Update mdebin with energy history if appending to output files */
448 if (Flags & MD_APPENDFILES)
450 restore_energyhistory_from_state(mdebin, &state_global->enerhist);
454 /* We might have read an energy history from checkpoint,
455 * free the allocated memory and reset the counts.
457 done_energyhistory(&state_global->enerhist);
458 init_energyhistory(&state_global->enerhist);
461 /* Set the initial energy history in state by updating once */
462 update_energyhistory(&state_global->enerhist, mdebin);
465 /* Initialize constraints */
466 if (constr && !DOMAINDECOMP(cr))
468 set_constraints(constr, top, ir, mdatoms, cr);
471 if (repl_ex_nst > 0 && MASTER(cr))
473 repl_ex = init_replica_exchange(fplog, cr->ms, state_global, ir,
474 repl_ex_nst, repl_ex_nex, repl_ex_seed);
477 /* PME tuning is only supported with GPUs or PME nodes and not with rerun.
478 * PME tuning is not supported with PME only for LJ and not for Coulomb.
480 if ((Flags & MD_TUNEPME) &&
481 EEL_PME(fr->eeltype) &&
482 ( (fr->cutoff_scheme == ecutsVERLET && fr->nbv->bUseGPU) || !(cr->duty & DUTY_PME)) &&
485 pme_loadbal_init(&pme_loadbal, ir, state->box, fr->ic, fr->pmedata);
487 if (cr->duty & DUTY_PME)
489 /* Start tuning right away, as we can't measure the load */
490 bPMETuneRunning = TRUE;
494 /* Separate PME nodes, we can measure the PP/PME load balance */
499 if (!ir->bContinuation && !bRerunMD)
501 if (mdatoms->cFREEZE && (state->flags & (1<<estV)))
503 /* Set the velocities of frozen particles to zero */
504 for (i = 0; i < mdatoms->homenr; i++)
506 for (m = 0; m < DIM; m++)
508 if (ir->opts.nFreeze[mdatoms->cFREEZE[i]][m])
518 /* Constrain the initial coordinates and velocities */
519 do_constrain_first(fplog, constr, ir, mdatoms, state,
524 /* Construct the virtual sites for the initial configuration */
525 construct_vsites(vsite, state->x, ir->delta_t, NULL,
526 top->idef.iparams, top->idef.il,
527 fr->ePBC, fr->bMolPBC, cr, state->box);
533 /* set free energy calculation frequency as the minimum
534 greatest common denominator of nstdhdl, nstexpanded, and repl_ex_nst*/
535 nstfep = ir->fepvals->nstdhdl;
538 nstfep = gmx_greatest_common_divisor(ir->fepvals->nstdhdl, nstfep);
542 nstfep = gmx_greatest_common_divisor(repl_ex_nst, nstfep);
545 /* I'm assuming we need global communication the first time! MRS */
546 cglo_flags = (CGLO_TEMPERATURE | CGLO_GSTAT
547 | ((ir->comm_mode != ecmNO) ? CGLO_STOPCM : 0)
548 | (bVV ? CGLO_PRESSURE : 0)
549 | (bVV ? CGLO_CONSTRAINT : 0)
550 | (bRerunMD ? CGLO_RERUNMD : 0)
551 | ((Flags & MD_READ_EKIN) ? CGLO_READEKIN : 0));
553 bSumEkinhOld = FALSE;
554 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
555 NULL, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
556 constr, NULL, FALSE, state->box,
557 top_global, &bSumEkinhOld, cglo_flags);
558 if (ir->eI == eiVVAK)
560 /* a second call to get the half step temperature initialized as well */
561 /* we do the same call as above, but turn the pressure off -- internally to
562 compute_globals, this is recognized as a velocity verlet half-step
563 kinetic energy calculation. This minimized excess variables, but
564 perhaps loses some logic?*/
566 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
567 NULL, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
568 constr, NULL, FALSE, state->box,
569 top_global, &bSumEkinhOld,
570 cglo_flags &~(CGLO_STOPCM | CGLO_PRESSURE));
573 /* Calculate the initial half step temperature, and save the ekinh_old */
574 if (!(Flags & MD_STARTFROMCPT))
576 for (i = 0; (i < ir->opts.ngtc); i++)
578 copy_mat(ekind->tcstat[i].ekinh, ekind->tcstat[i].ekinh_old);
583 enerd->term[F_TEMP] *= 2; /* result of averages being done over previous and current step,
584 and there is no previous step */
587 /* if using an iterative algorithm, we need to create a working directory for the state. */
590 bufstate = init_bufstate(state);
593 /* need to make an initiation call to get the Trotter variables set, as well as other constants for non-trotter
594 temperature control */
595 trotter_seq = init_npt_vars(ir, state, &MassQ, bTrotter);
599 if (constr && !ir->bContinuation && ir->eConstrAlg == econtLINCS)
602 "RMS relative constraint deviation after constraining: %.2e\n",
603 constr_rmsd(constr, FALSE));
605 if (EI_STATE_VELOCITY(ir->eI))
607 fprintf(fplog, "Initial temperature: %g K\n", enerd->term[F_TEMP]);
611 fprintf(stderr, "starting md rerun '%s', reading coordinates from"
612 " input trajectory '%s'\n\n",
613 *(top_global->name), opt2fn("-rerun", nfile, fnm));
616 fprintf(stderr, "Calculated time to finish depends on nsteps from "
617 "run input file,\nwhich may not correspond to the time "
618 "needed to process input trajectory.\n\n");
624 fprintf(stderr, "starting mdrun '%s'\n",
625 *(top_global->name));
628 sprintf(tbuf, "%8.1f", (ir->init_step+ir->nsteps)*ir->delta_t);
632 sprintf(tbuf, "%s", "infinite");
634 if (ir->init_step > 0)
636 fprintf(stderr, "%s steps, %s ps (continuing from step %s, %8.1f ps).\n",
637 gmx_step_str(ir->init_step+ir->nsteps, sbuf), tbuf,
638 gmx_step_str(ir->init_step, sbuf2),
639 ir->init_step*ir->delta_t);
643 fprintf(stderr, "%s steps, %s ps.\n",
644 gmx_step_str(ir->nsteps, sbuf), tbuf);
647 fprintf(fplog, "\n");
650 walltime_accounting_start(walltime_accounting);
651 wallcycle_start(wcycle, ewcRUN);
652 print_start(fplog, cr, walltime_accounting, "mdrun");
654 /* safest point to do file checkpointing is here. More general point would be immediately before integrator call */
656 chkpt_ret = fcCheckPointParallel( cr->nodeid,
660 gmx_fatal( 3, __FILE__, __LINE__, "Checkpoint error on step %d\n", 0 );
665 /***********************************************************
669 ************************************************************/
671 /* if rerunMD then read coordinates and velocities from input trajectory */
674 if (getenv("GMX_FORCE_UPDATE"))
682 bNotLastFrame = read_first_frame(oenv, &status,
683 opt2fn("-rerun", nfile, fnm),
684 &rerun_fr, TRX_NEED_X | TRX_READ_V);
685 if (rerun_fr.natoms != top_global->natoms)
688 "Number of atoms in trajectory (%d) does not match the "
689 "run input file (%d)\n",
690 rerun_fr.natoms, top_global->natoms);
692 if (ir->ePBC != epbcNONE)
696 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);
698 if (max_cutoff2(ir->ePBC, rerun_fr.box) < sqr(fr->rlistlong))
700 gmx_fatal(FARGS, "Rerun trajectory frame step %d time %f has too small box dimensions", rerun_fr.step, rerun_fr.time);
707 rerun_parallel_comm(cr, &rerun_fr, &bNotLastFrame);
710 if (ir->ePBC != epbcNONE)
712 /* Set the shift vectors.
713 * Necessary here when have a static box different from the tpr box.
715 calc_shifts(rerun_fr.box, fr->shift_vec);
719 /* loop over MD steps or if rerunMD to end of input trajectory */
721 /* Skip the first Nose-Hoover integration when we get the state from tpx */
722 bStateFromTPX = !bStateFromCP;
723 bInitStep = bFirstStep && (bStateFromTPX || bVV);
724 bStartingFromCpt = (Flags & MD_STARTFROMCPT) && bInitStep;
725 bSumEkinhOld = FALSE;
727 bNeedRepartition = FALSE;
729 init_global_signals(&gs, cr, ir, repl_ex_nst);
731 step = ir->init_step;
734 init_nlistheuristics(&nlh, bGStatEveryStep, step);
736 if (MULTISIM(cr) && (repl_ex_nst <= 0 ))
738 /* check how many steps are left in other sims */
739 multisim_nsteps = get_multisim_nsteps(cr, ir->nsteps);
743 /* and stop now if we should */
744 bLastStep = (bRerunMD || (ir->nsteps >= 0 && step_rel > ir->nsteps) ||
745 ((multisim_nsteps >= 0) && (step_rel >= multisim_nsteps )));
746 while (!bLastStep || (bRerunMD && bNotLastFrame))
749 wallcycle_start(wcycle, ewcSTEP);
755 step = rerun_fr.step;
756 step_rel = step - ir->init_step;
769 bLastStep = (step_rel == ir->nsteps);
770 t = t0 + step*ir->delta_t;
773 if (ir->efep != efepNO || ir->bSimTemp)
775 /* find and set the current lambdas. If rerunning, we either read in a state, or a lambda value,
776 requiring different logic. */
778 set_current_lambdas(step, ir->fepvals, bRerunMD, &rerun_fr, state_global, state, lam0);
779 bDoDHDL = do_per_step(step, ir->fepvals->nstdhdl);
780 bDoFEP = (do_per_step(step, nstfep) && (ir->efep != efepNO));
781 bDoExpanded = (do_per_step(step, ir->expandedvals->nstexpanded)
782 && (ir->bExpanded) && (step > 0) && (!bStartingFromCpt));
785 bDoReplEx = ((repl_ex_nst > 0) && (step > 0) && !bLastStep &&
786 do_per_step(step, repl_ex_nst));
790 update_annealing_target_temp(&(ir->opts), t);
795 if (!DOMAINDECOMP(cr) || MASTER(cr))
797 for (i = 0; i < state_global->natoms; i++)
799 copy_rvec(rerun_fr.x[i], state_global->x[i]);
803 for (i = 0; i < state_global->natoms; i++)
805 copy_rvec(rerun_fr.v[i], state_global->v[i]);
810 for (i = 0; i < state_global->natoms; i++)
812 clear_rvec(state_global->v[i]);
816 fprintf(stderr, "\nWARNING: Some frames do not contain velocities.\n"
817 " Ekin, temperature and pressure are incorrect,\n"
818 " the virial will be incorrect when constraints are present.\n"
820 bRerunWarnNoV = FALSE;
824 copy_mat(rerun_fr.box, state_global->box);
825 copy_mat(state_global->box, state->box);
827 if (vsite && (Flags & MD_RERUN_VSITE))
829 if (DOMAINDECOMP(cr))
831 gmx_fatal(FARGS, "Vsite recalculation with -rerun is not implemented with domain decomposition, use a single rank");
835 /* Following is necessary because the graph may get out of sync
836 * with the coordinates if we only have every N'th coordinate set
838 mk_mshift(fplog, graph, fr->ePBC, state->box, state->x);
839 shift_self(graph, state->box, state->x);
841 construct_vsites(vsite, state->x, ir->delta_t, state->v,
842 top->idef.iparams, top->idef.il,
843 fr->ePBC, fr->bMolPBC, cr, state->box);
846 unshift_self(graph, state->box, state->x);
851 /* Stop Center of Mass motion */
852 bStopCM = (ir->comm_mode != ecmNO && do_per_step(step, ir->nstcomm));
856 /* for rerun MD always do Neighbour Searching */
857 bNS = (bFirstStep || ir->nstlist != 0);
862 /* Determine whether or not to do Neighbour Searching and LR */
863 bNStList = (ir->nstlist > 0 && step % ir->nstlist == 0);
865 bNS = (bFirstStep || bExchanged || bNeedRepartition || bNStList || bDoFEP ||
866 (ir->nstlist == -1 && nlh.nabnsb > 0));
868 if (bNS && ir->nstlist == -1)
870 set_nlistheuristics(&nlh, bFirstStep || bExchanged || bNeedRepartition || bDoFEP, step);
874 /* check whether we should stop because another simulation has
878 if ( (multisim_nsteps >= 0) && (step_rel >= multisim_nsteps) &&
879 (multisim_nsteps != ir->nsteps) )
886 "Stopping simulation %d because another one has finished\n",
890 gs.sig[eglsCHKPT] = 1;
895 /* < 0 means stop at next step, > 0 means stop at next NS step */
896 if ( (gs.set[eglsSTOPCOND] < 0) ||
897 ( (gs.set[eglsSTOPCOND] > 0) && (bNStList || ir->nstlist == 0) ) )
902 /* Determine whether or not to update the Born radii if doing GB */
903 bBornRadii = bFirstStep;
904 if (ir->implicit_solvent && (step % ir->nstgbradii == 0))
909 do_log = do_per_step(step, ir->nstlog) || bFirstStep || bLastStep;
910 do_verbose = bVerbose &&
911 (step % stepout == 0 || bFirstStep || bLastStep);
913 if (bNS && !(bFirstStep && ir->bContinuation && !bRerunMD))
921 bMasterState = FALSE;
922 /* Correct the new box if it is too skewed */
923 if (DYNAMIC_BOX(*ir))
925 if (correct_box(fplog, step, state->box, graph))
930 if (DOMAINDECOMP(cr) && bMasterState)
932 dd_collect_state(cr->dd, state, state_global);
936 if (DOMAINDECOMP(cr))
938 /* Repartition the domain decomposition */
939 wallcycle_start(wcycle, ewcDOMDEC);
940 dd_partition_system(fplog, step, cr,
941 bMasterState, nstglobalcomm,
942 state_global, top_global, ir,
943 state, &f, mdatoms, top, fr,
944 vsite, shellfc, constr,
946 do_verbose && !bPMETuneRunning);
947 wallcycle_stop(wcycle, ewcDOMDEC);
948 /* If using an iterative integrator, reallocate space to match the decomposition */
952 if (MASTER(cr) && do_log)
954 print_ebin_header(fplog, step, t, state->lambda[efptFEP]); /* can we improve the information printed here? */
957 if (ir->efep != efepNO)
959 update_mdatoms(mdatoms, state->lambda[efptMASS]);
962 if ((bRerunMD && rerun_fr.bV) || bExchanged)
965 /* We need the kinetic energy at minus the half step for determining
966 * the full step kinetic energy and possibly for T-coupling.*/
967 /* This may not be quite working correctly yet . . . . */
968 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
969 wcycle, enerd, NULL, NULL, NULL, NULL, mu_tot,
970 constr, NULL, FALSE, state->box,
971 top_global, &bSumEkinhOld,
972 CGLO_RERUNMD | CGLO_GSTAT | CGLO_TEMPERATURE);
974 clear_mat(force_vir);
976 /* We write a checkpoint at this MD step when:
977 * either at an NS step when we signalled through gs,
978 * or at the last step (but not when we do not want confout),
979 * but never at the first step or with rerun.
981 bCPT = (((gs.set[eglsCHKPT] && (bNS || ir->nstlist == 0)) ||
982 (bLastStep && (Flags & MD_CONFOUT))) &&
983 step > ir->init_step && !bRerunMD);
986 gs.set[eglsCHKPT] = 0;
989 /* Determine the energy and pressure:
990 * at nstcalcenergy steps and at energy output steps (set below).
992 if (EI_VV(ir->eI) && (!bInitStep))
994 /* for vv, the first half of the integration actually corresponds
995 to the previous step. bCalcEner is only required to be evaluated on the 'next' step,
996 but the virial needs to be calculated on both the current step and the 'next' step. Future
997 reorganization may be able to get rid of one of the bCalcVir=TRUE steps. */
999 bCalcEner = do_per_step(step-1, ir->nstcalcenergy);
1000 bCalcVir = bCalcEner ||
1001 (ir->epc != epcNO && (do_per_step(step, ir->nstpcouple) || do_per_step(step-1, ir->nstpcouple)));
1005 bCalcEner = do_per_step(step, ir->nstcalcenergy);
1006 bCalcVir = bCalcEner ||
1007 (ir->epc != epcNO && do_per_step(step, ir->nstpcouple));
1010 /* Do we need global communication ? */
1011 bGStat = (bCalcVir || bCalcEner || bStopCM ||
1012 do_per_step(step, nstglobalcomm) || (bVV && IR_NVT_TROTTER(ir) && do_per_step(step-1, nstglobalcomm)) ||
1013 (ir->nstlist == -1 && !bRerunMD && step >= nlh.step_nscheck));
1015 do_ene = (do_per_step(step, ir->nstenergy) || bLastStep);
1017 if (do_ene || do_log || bDoReplEx)
1024 /* these CGLO_ options remain the same throughout the iteration */
1025 cglo_flags = ((bRerunMD ? CGLO_RERUNMD : 0) |
1026 (bGStat ? CGLO_GSTAT : 0)
1029 force_flags = (GMX_FORCE_STATECHANGED |
1030 ((DYNAMIC_BOX(*ir) || bRerunMD) ? GMX_FORCE_DYNAMICBOX : 0) |
1031 GMX_FORCE_ALLFORCES |
1033 (bCalcVir ? GMX_FORCE_VIRIAL : 0) |
1034 (bCalcEner ? GMX_FORCE_ENERGY : 0) |
1035 (bDoFEP ? GMX_FORCE_DHDL : 0)
1040 if (do_per_step(step, ir->nstcalclr))
1042 force_flags |= GMX_FORCE_DO_LR;
1048 /* Now is the time to relax the shells */
1049 count = relax_shell_flexcon(fplog, cr, bVerbose, step,
1050 ir, bNS, force_flags,
1053 state, f, force_vir, mdatoms,
1054 nrnb, wcycle, graph, groups,
1055 shellfc, fr, bBornRadii, t, mu_tot,
1057 mdoutf_get_fp_field(outf));
1067 /* The coordinates (x) are shifted (to get whole molecules)
1069 * This is parallellized as well, and does communication too.
1070 * Check comments in sim_util.c
1072 do_force(fplog, cr, ir, step, nrnb, wcycle, top, groups,
1073 state->box, state->x, &state->hist,
1074 f, force_vir, mdatoms, enerd, fcd,
1075 state->lambda, graph,
1076 fr, vsite, mu_tot, t, mdoutf_get_fp_field(outf), ed, bBornRadii,
1077 (bNS ? GMX_FORCE_NS : 0) | force_flags);
1080 if (bVV && !bStartingFromCpt && !bRerunMD)
1081 /* ############### START FIRST UPDATE HALF-STEP FOR VV METHODS############### */
1083 if (ir->eI == eiVV && bInitStep)
1085 /* if using velocity verlet with full time step Ekin,
1086 * take the first half step only to compute the
1087 * virial for the first step. From there,
1088 * revert back to the initial coordinates
1089 * so that the input is actually the initial step.
1091 copy_rvecn(state->v, cbuf, 0, state->natoms); /* should make this better for parallelizing? */
1095 /* this is for NHC in the Ekin(t+dt/2) version of vv */
1096 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ1);
1099 /* If we are using twin-range interactions where the long-range component
1100 * is only evaluated every nstcalclr>1 steps, we should do a special update
1101 * step to combine the long-range forces on these steps.
1102 * For nstcalclr=1 this is not done, since the forces would have been added
1103 * directly to the short-range forces already.
1105 * TODO Remove various aspects of VV+twin-range in master
1106 * branch, because VV integrators did not ever support
1107 * twin-range multiple time stepping with constraints.
1109 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1111 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC,
1112 f, bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1113 ekind, M, upd, bInitStep, etrtVELOCITY1,
1114 cr, nrnb, constr, &top->idef);
1116 if (bIterativeCase && do_per_step(step-1, ir->nstpcouple) && !bInitStep)
1118 gmx_iterate_init(&iterate, TRUE);
1120 /* for iterations, we save these vectors, as we will be self-consistently iterating
1123 /*#### UPDATE EXTENDED VARIABLES IN TROTTER FORMULATION */
1125 /* save the state */
1126 if (iterate.bIterationActive)
1128 copy_coupling_state(state, bufstate, ekind, ekind_save, &(ir->opts));
1131 bFirstIterate = TRUE;
1132 while (bFirstIterate || iterate.bIterationActive)
1134 if (iterate.bIterationActive)
1136 copy_coupling_state(bufstate, state, ekind_save, ekind, &(ir->opts));
1137 if (bFirstIterate && bTrotter)
1139 /* The first time through, we need a decent first estimate
1140 of veta(t+dt) to compute the constraints. Do
1141 this by computing the box volume part of the
1142 trotter integration at this time. Nothing else
1143 should be changed by this routine here. If
1144 !(first time), we start with the previous value
1147 veta_save = state->veta;
1148 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ0);
1149 vetanew = state->veta;
1150 state->veta = veta_save;
1154 if (!bRerunMD || rerun_fr.bV || bForceUpdate) /* Why is rerun_fr.bV here? Unclear. */
1156 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1157 state, fr->bMolPBC, graph, f,
1158 &top->idef, shake_vir,
1159 cr, nrnb, wcycle, upd, constr,
1160 TRUE, bCalcVir, vetanew);
1162 if (bCalcVir && bUpdateDoLR && ir->nstcalclr > 1)
1164 /* Correct the virial for multiple time stepping */
1165 m_sub(shake_vir, fr->vir_twin_constr, shake_vir);
1170 /* Need to unshift here if a do_force has been
1171 called in the previous step */
1172 unshift_self(graph, state->box, state->x);
1175 /* if VV, compute the pressure and constraints */
1176 /* For VV2, we strictly only need this if using pressure
1177 * control, but we really would like to have accurate pressures
1179 * Think about ways around this in the future?
1180 * For now, keep this choice in comments.
1182 /*bPres = (ir->eI==eiVV || IR_NPT_TROTTER(ir)); */
1183 /*bTemp = ((ir->eI==eiVV &&(!bInitStep)) || (ir->eI==eiVVAK && IR_NPT_TROTTER(ir)));*/
1185 bTemp = ((ir->eI == eiVV && (!bInitStep)) || (ir->eI == eiVVAK));
1186 if (bCalcEner && ir->eI == eiVVAK) /*MRS: 7/9/2010 -- this still doesn't fix it?*/
1188 bSumEkinhOld = TRUE;
1190 /* for vv, the first half of the integration actually corresponds to the previous step.
1191 So we need information from the last step in the first half of the integration */
1192 if (bGStat || do_per_step(step-1, nstglobalcomm))
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 */
1215 /* temperature scaling and pressure scaling to produce the extended variables at t+dt */
1220 m_add(force_vir, shake_vir, total_vir); /* we need the un-dispersion corrected total vir here */
1221 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ2);
1228 /* We need the kinetic energy at minus the half step for determining
1229 * the full step kinetic energy and possibly for T-coupling.*/
1230 /* This may not be quite working correctly yet . . . . */
1231 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1232 wcycle, enerd, NULL, NULL, NULL, NULL, mu_tot,
1233 constr, NULL, FALSE, state->box,
1234 top_global, &bSumEkinhOld,
1235 CGLO_RERUNMD | CGLO_GSTAT | CGLO_TEMPERATURE);
1240 if (iterate.bIterationActive &&
1241 done_iterating(cr, fplog, step, &iterate, bFirstIterate,
1242 state->veta, &vetanew))
1246 bFirstIterate = FALSE;
1249 if (bTrotter && !bInitStep)
1251 copy_mat(shake_vir, state->svir_prev);
1252 copy_mat(force_vir, state->fvir_prev);
1253 if (IR_NVT_TROTTER(ir) && ir->eI == eiVV)
1255 /* update temperature and kinetic energy now that step is over - this is the v(t+dt) point */
1256 enerd->term[F_TEMP] = sum_ekin(&(ir->opts), ekind, NULL, (ir->eI == eiVV), FALSE);
1257 enerd->term[F_EKIN] = trace(ekind->ekin);
1260 /* if it's the initial step, we performed this first step just to get the constraint virial */
1261 if (bInitStep && ir->eI == eiVV)
1263 copy_rvecn(cbuf, state->v, 0, state->natoms);
1267 /* MRS -- now done iterating -- compute the conserved quantity */
1270 saved_conserved_quantity = compute_conserved_from_auxiliary(ir, state, &MassQ);
1273 last_ekin = enerd->term[F_EKIN];
1275 if ((ir->eDispCorr != edispcEnerPres) && (ir->eDispCorr != edispcAllEnerPres))
1277 saved_conserved_quantity -= enerd->term[F_DISPCORR];
1279 /* sum up the foreign energy and dhdl terms for vv. currently done every step so that dhdl is correct in the .edr */
1282 sum_dhdl(enerd, state->lambda, ir->fepvals);
1286 /* ######## END FIRST UPDATE STEP ############## */
1287 /* ######## If doing VV, we now have v(dt) ###### */
1290 /* perform extended ensemble sampling in lambda - we don't
1291 actually move to the new state before outputting
1292 statistics, but if performing simulated tempering, we
1293 do update the velocities and the tau_t. */
1295 lamnew = ExpandedEnsembleDynamics(fplog, ir, enerd, state, &MassQ, state->fep_state, &state->dfhist, step, state->v, mdatoms);
1296 /* history is maintained in state->dfhist, but state_global is what is sent to trajectory and log output */
1297 copy_df_history(&state_global->dfhist, &state->dfhist);
1300 /* Now we have the energies and forces corresponding to the
1301 * coordinates at time t. We must output all of this before
1304 do_md_trajectory_writing(fplog, cr, nfile, fnm, step, step_rel, t,
1305 ir, state, state_global, top_global, fr,
1306 outf, mdebin, ekind, f, f_global,
1308 bCPT, bRerunMD, bLastStep, (Flags & MD_CONFOUT),
1310 /* Check if IMD step and do IMD communication, if bIMD is TRUE. */
1311 bIMDstep = do_IMD(ir->bIMD, step, cr, bNS, state->box, state->x, ir, t, wcycle);
1313 /* kludge -- virial is lost with restart for NPT control. Must restart */
1314 if (bStartingFromCpt && bVV)
1316 copy_mat(state->svir_prev, shake_vir);
1317 copy_mat(state->fvir_prev, force_vir);
1320 elapsed_time = walltime_accounting_get_current_elapsed_time(walltime_accounting);
1322 /* Check whether everything is still allright */
1323 if (((int)gmx_get_stop_condition() > handled_stop_condition)
1324 #ifdef GMX_THREAD_MPI
1329 /* this is just make gs.sig compatible with the hack
1330 of sending signals around by MPI_Reduce with together with
1332 if (gmx_get_stop_condition() == gmx_stop_cond_next_ns)
1334 gs.sig[eglsSTOPCOND] = 1;
1336 if (gmx_get_stop_condition() == gmx_stop_cond_next)
1338 gs.sig[eglsSTOPCOND] = -1;
1340 /* < 0 means stop at next step, > 0 means stop at next NS step */
1344 "\n\nReceived the %s signal, stopping at the next %sstep\n\n",
1345 gmx_get_signal_name(),
1346 gs.sig[eglsSTOPCOND] == 1 ? "NS " : "");
1350 "\n\nReceived the %s signal, stopping at the next %sstep\n\n",
1351 gmx_get_signal_name(),
1352 gs.sig[eglsSTOPCOND] == 1 ? "NS " : "");
1354 handled_stop_condition = (int)gmx_get_stop_condition();
1356 else if (MASTER(cr) && (bNS || ir->nstlist <= 0) &&
1357 (max_hours > 0 && elapsed_time > max_hours*60.0*60.0*0.99) &&
1358 gs.sig[eglsSTOPCOND] == 0 && gs.set[eglsSTOPCOND] == 0)
1360 /* Signal to terminate the run */
1361 gs.sig[eglsSTOPCOND] = 1;
1364 fprintf(fplog, "\nStep %s: Run time exceeded %.3f hours, will terminate the run\n", gmx_step_str(step, sbuf), max_hours*0.99);
1366 fprintf(stderr, "\nStep %s: Run time exceeded %.3f hours, will terminate the run\n", gmx_step_str(step, sbuf), max_hours*0.99);
1369 if (bResetCountersHalfMaxH && MASTER(cr) &&
1370 elapsed_time > max_hours*60.0*60.0*0.495)
1372 gs.sig[eglsRESETCOUNTERS] = 1;
1375 if (ir->nstlist == -1 && !bRerunMD)
1377 /* When bGStatEveryStep=FALSE, global_stat is only called
1378 * when we check the atom displacements, not at NS steps.
1379 * This means that also the bonded interaction count check is not
1380 * performed immediately after NS. Therefore a few MD steps could
1381 * be performed with missing interactions.
1382 * But wrong energies are never written to file,
1383 * since energies are only written after global_stat
1386 if (step >= nlh.step_nscheck)
1388 nlh.nabnsb = natoms_beyond_ns_buffer(ir, fr, &top->cgs,
1389 nlh.scale_tot, state->x);
1393 /* This is not necessarily true,
1394 * but step_nscheck is determined quite conservatively.
1400 /* In parallel we only have to check for checkpointing in steps
1401 * where we do global communication,
1402 * otherwise the other nodes don't know.
1404 if (MASTER(cr) && ((bGStat || !PAR(cr)) &&
1407 elapsed_time >= nchkpt*cpt_period*60.0)) &&
1408 gs.set[eglsCHKPT] == 0)
1410 gs.sig[eglsCHKPT] = 1;
1413 /* at the start of step, randomize or scale the velocities (trotter done elsewhere) */
1418 update_tcouple(step, ir, state, ekind, &MassQ, mdatoms);
1420 if (ETC_ANDERSEN(ir->etc)) /* keep this outside of update_tcouple because of the extra info required to pass */
1422 gmx_bool bIfRandomize;
1423 bIfRandomize = update_randomize_velocities(ir, step, cr, mdatoms, state, upd, constr);
1424 /* if we have constraints, we have to remove the kinetic energy parallel to the bonds */
1425 if (constr && bIfRandomize)
1427 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1428 state, fr->bMolPBC, graph, f,
1429 &top->idef, tmp_vir,
1430 cr, nrnb, wcycle, upd, constr,
1431 TRUE, bCalcVir, vetanew);
1436 if (bIterativeCase && do_per_step(step, ir->nstpcouple))
1438 gmx_iterate_init(&iterate, TRUE);
1439 /* for iterations, we save these vectors, as we will be redoing the calculations */
1440 copy_coupling_state(state, bufstate, ekind, ekind_save, &(ir->opts));
1443 bFirstIterate = TRUE;
1444 while (bFirstIterate || iterate.bIterationActive)
1446 /* We now restore these vectors to redo the calculation with improved extended variables */
1447 if (iterate.bIterationActive)
1449 copy_coupling_state(bufstate, state, ekind_save, ekind, &(ir->opts));
1452 /* We make the decision to break or not -after- the calculation of Ekin and Pressure,
1453 so scroll down for that logic */
1455 /* ######### START SECOND UPDATE STEP ################# */
1456 /* Box is changed in update() when we do pressure coupling,
1457 * but we should still use the old box for energy corrections and when
1458 * writing it to the energy file, so it matches the trajectory files for
1459 * the same timestep above. Make a copy in a separate array.
1461 copy_mat(state->box, lastbox);
1465 if (!(bRerunMD && !rerun_fr.bV && !bForceUpdate))
1467 wallcycle_start(wcycle, ewcUPDATE);
1468 /* UPDATE PRESSURE VARIABLES IN TROTTER FORMULATION WITH CONSTRAINTS */
1471 if (iterate.bIterationActive)
1479 /* we use a new value of scalevir to converge the iterations faster */
1480 scalevir = tracevir/trace(shake_vir);
1482 msmul(shake_vir, scalevir, shake_vir);
1483 m_add(force_vir, shake_vir, total_vir);
1484 clear_mat(shake_vir);
1486 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ3);
1487 /* We can only do Berendsen coupling after we have summed
1488 * the kinetic energy or virial. Since the happens
1489 * in global_state after update, we should only do it at
1490 * step % nstlist = 1 with bGStatEveryStep=FALSE.
1495 update_tcouple(step, ir, state, ekind, &MassQ, mdatoms);
1496 update_pcouple(fplog, step, ir, state, pcoupl_mu, M, bInitStep);
1501 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1503 /* velocity half-step update */
1504 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1505 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1506 ekind, M, upd, FALSE, etrtVELOCITY2,
1507 cr, nrnb, constr, &top->idef);
1510 /* Above, initialize just copies ekinh into ekin,
1511 * it doesn't copy position (for VV),
1512 * and entire integrator for MD.
1515 if (ir->eI == eiVVAK)
1517 copy_rvecn(state->x, cbuf, 0, state->natoms);
1519 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1521 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1522 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1523 ekind, M, upd, bInitStep, etrtPOSITION, cr, nrnb, constr, &top->idef);
1524 wallcycle_stop(wcycle, ewcUPDATE);
1526 update_constraints(fplog, step, &dvdl_constr, ir, ekind, mdatoms, state,
1527 fr->bMolPBC, graph, f,
1528 &top->idef, shake_vir,
1529 cr, nrnb, wcycle, upd, constr,
1530 FALSE, bCalcVir, state->veta);
1532 if (bCalcVir && bUpdateDoLR && ir->nstcalclr > 1)
1534 /* Correct the virial for multiple time stepping */
1535 m_sub(shake_vir, fr->vir_twin_constr, shake_vir);
1538 if (ir->eI == eiVVAK)
1540 /* erase F_EKIN and F_TEMP here? */
1541 /* just compute the kinetic energy at the half step to perform a trotter step */
1542 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1543 wcycle, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
1544 constr, NULL, FALSE, lastbox,
1545 top_global, &bSumEkinhOld,
1546 cglo_flags | CGLO_TEMPERATURE
1548 wallcycle_start(wcycle, ewcUPDATE);
1549 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ4);
1550 /* now we know the scaling, we can compute the positions again again */
1551 copy_rvecn(cbuf, state->x, 0, state->natoms);
1553 bUpdateDoLR = (fr->bTwinRange && do_per_step(step, ir->nstcalclr));
1555 update_coords(fplog, step, ir, mdatoms, state, fr->bMolPBC, f,
1556 bUpdateDoLR, fr->f_twin, bCalcVir ? &fr->vir_twin_constr : NULL, fcd,
1557 ekind, M, upd, bInitStep, etrtPOSITION, cr, nrnb, constr, &top->idef);
1558 wallcycle_stop(wcycle, ewcUPDATE);
1560 /* do we need an extra constraint here? just need to copy out of state->v to upd->xp? */
1561 /* are the small terms in the shake_vir here due
1562 * to numerical errors, or are they important
1563 * physically? I'm thinking they are just errors, but not completely sure.
1564 * For now, will call without actually constraining, constr=NULL*/
1565 update_constraints(fplog, step, NULL, ir, ekind, mdatoms,
1566 state, fr->bMolPBC, graph, f,
1567 &top->idef, tmp_vir,
1568 cr, nrnb, wcycle, upd, NULL,
1573 if (fr->bSepDVDL && fplog && do_log)
1575 gmx_print_sepdvdl(fplog, "Constraint dV/dl", 0.0, dvdl_constr);
1579 /* this factor or 2 correction is necessary
1580 because half of the constraint force is removed
1581 in the vv step, so we have to double it. See
1582 the Redmine issue #1255. It is not yet clear
1583 if the factor of 2 is exact, or just a very
1584 good approximation, and this will be
1585 investigated. The next step is to see if this
1586 can be done adding a dhdl contribution from the
1587 rattle step, but this is somewhat more
1588 complicated with the current code. Will be
1589 investigated, hopefully for 4.6.3. However,
1590 this current solution is much better than
1591 having it completely wrong.
1593 enerd->term[F_DVDL_CONSTR] += 2*dvdl_constr;
1597 enerd->term[F_DVDL_CONSTR] += dvdl_constr;
1602 /* Need to unshift here */
1603 unshift_self(graph, state->box, state->x);
1608 wallcycle_start(wcycle, ewcVSITECONSTR);
1611 shift_self(graph, state->box, state->x);
1613 construct_vsites(vsite, state->x, ir->delta_t, state->v,
1614 top->idef.iparams, top->idef.il,
1615 fr->ePBC, fr->bMolPBC, cr, state->box);
1619 unshift_self(graph, state->box, state->x);
1621 wallcycle_stop(wcycle, ewcVSITECONSTR);
1624 /* ############## IF NOT VV, Calculate globals HERE, also iterate constraints ############ */
1625 /* With Leap-Frog we can skip compute_globals at
1626 * non-communication steps, but we need to calculate
1627 * the kinetic energy one step before communication.
1629 if (bGStat || (!EI_VV(ir->eI) && do_per_step(step+1, nstglobalcomm)))
1631 if (ir->nstlist == -1 && bFirstIterate)
1633 gs.sig[eglsNABNSB] = nlh.nabnsb;
1635 compute_globals(fplog, gstat, cr, ir, fr, ekind, state, state_global, mdatoms, nrnb, vcm,
1636 wcycle, enerd, force_vir, shake_vir, total_vir, pres, mu_tot,
1638 bFirstIterate ? &gs : NULL,
1639 (step_rel % gs.nstms == 0) &&
1640 (multisim_nsteps < 0 || (step_rel < multisim_nsteps)),
1642 top_global, &bSumEkinhOld,
1644 | (!EI_VV(ir->eI) || bRerunMD ? CGLO_ENERGY : 0)
1645 | (!EI_VV(ir->eI) && bStopCM ? CGLO_STOPCM : 0)
1646 | (!EI_VV(ir->eI) ? CGLO_TEMPERATURE : 0)
1647 | (!EI_VV(ir->eI) || bRerunMD ? CGLO_PRESSURE : 0)
1648 | (iterate.bIterationActive ? CGLO_ITERATE : 0)
1649 | (bFirstIterate ? CGLO_FIRSTITERATE : 0)
1652 if (ir->nstlist == -1 && bFirstIterate)
1654 nlh.nabnsb = gs.set[eglsNABNSB];
1655 gs.set[eglsNABNSB] = 0;
1658 /* bIterate is set to keep it from eliminating the old ekin kinetic energy terms */
1659 /* ############# END CALC EKIN AND PRESSURE ################# */
1661 /* Note: this is OK, but there are some numerical precision issues with using the convergence of
1662 the virial that should probably be addressed eventually. state->veta has better properies,
1663 but what we actually need entering the new cycle is the new shake_vir value. Ideally, we could
1664 generate the new shake_vir, but test the veta value for convergence. This will take some thought. */
1666 if (iterate.bIterationActive &&
1667 done_iterating(cr, fplog, step, &iterate, bFirstIterate,
1668 trace(shake_vir), &tracevir))
1672 bFirstIterate = FALSE;
1675 if (!bVV || bRerunMD)
1677 /* sum up the foreign energy and dhdl terms for md and sd. currently done every step so that dhdl is correct in the .edr */
1678 sum_dhdl(enerd, state->lambda, ir->fepvals);
1680 update_box(fplog, step, ir, mdatoms, state, f,
1681 ir->nstlist == -1 ? &nlh.scale_tot : NULL, pcoupl_mu, nrnb, upd);
1683 /* ################# END UPDATE STEP 2 ################# */
1684 /* #### We now have r(t+dt) and v(t+dt/2) ############# */
1686 /* The coordinates (x) were unshifted in update */
1689 /* We will not sum ekinh_old,
1690 * so signal that we still have to do it.
1692 bSumEkinhOld = TRUE;
1695 /* ######### BEGIN PREPARING EDR OUTPUT ########### */
1697 /* use the directly determined last velocity, not actually the averaged half steps */
1698 if (bTrotter && ir->eI == eiVV)
1700 enerd->term[F_EKIN] = last_ekin;
1702 enerd->term[F_ETOT] = enerd->term[F_EPOT] + enerd->term[F_EKIN];
1706 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + saved_conserved_quantity;
1710 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + compute_conserved_from_auxiliary(ir, state, &MassQ);
1712 /* ######### END PREPARING EDR OUTPUT ########### */
1717 gmx_bool do_dr, do_or;
1719 if (fplog && do_log && bDoExpanded)
1721 /* only needed if doing expanded ensemble */
1722 PrintFreeEnergyInfoToFile(fplog, ir->fepvals, ir->expandedvals, ir->bSimTemp ? ir->simtempvals : NULL,
1723 &state_global->dfhist, state->fep_state, ir->nstlog, step);
1725 if (!(bStartingFromCpt && (EI_VV(ir->eI))))
1729 upd_mdebin(mdebin, bDoDHDL, TRUE,
1730 t, mdatoms->tmass, enerd, state,
1731 ir->fepvals, ir->expandedvals, lastbox,
1732 shake_vir, force_vir, total_vir, pres,
1733 ekind, mu_tot, constr);
1737 upd_mdebin_step(mdebin);
1740 do_dr = do_per_step(step, ir->nstdisreout);
1741 do_or = do_per_step(step, ir->nstorireout);
1743 print_ebin(mdoutf_get_fp_ene(outf), do_ene, do_dr, do_or, do_log ? fplog : NULL,
1745 eprNORMAL, bCompact, mdebin, fcd, groups, &(ir->opts));
1747 if (ir->ePull != epullNO)
1749 pull_print_output(ir->pull, step, t);
1752 if (do_per_step(step, ir->nstlog))
1754 if (fflush(fplog) != 0)
1756 gmx_fatal(FARGS, "Cannot flush logfile - maybe you are out of disk space?");
1762 /* Have to do this part _after_ outputting the logfile and the edr file */
1763 /* Gets written into the state at the beginning of next loop*/
1764 state->fep_state = lamnew;
1766 /* Print the remaining wall clock time for the run */
1767 if (MULTIMASTER(cr) && (do_verbose || gmx_got_usr_signal()) && !bPMETuneRunning)
1771 fprintf(stderr, "\n");
1773 print_time(stderr, walltime_accounting, step, ir, cr);
1776 /* Ion/water position swapping.
1777 * Not done in last step since trajectory writing happens before this call
1778 * in the MD loop and exchanges would be lost anyway. */
1779 bNeedRepartition = FALSE;
1780 if ((ir->eSwapCoords != eswapNO) && (step > 0) && !bLastStep &&
1781 do_per_step(step, ir->swap->nstswap))
1783 bNeedRepartition = do_swapcoords(cr, step, t, ir, wcycle,
1784 bRerunMD ? rerun_fr.x : state->x,
1785 bRerunMD ? rerun_fr.box : state->box,
1786 top_global, MASTER(cr) && bVerbose, bRerunMD);
1788 if (bNeedRepartition && DOMAINDECOMP(cr))
1790 dd_collect_state(cr->dd, state, state_global);
1794 /* Replica exchange */
1798 bExchanged = replica_exchange(fplog, cr, repl_ex,
1799 state_global, enerd,
1803 if ( (bExchanged || bNeedRepartition) && DOMAINDECOMP(cr) )
1805 dd_partition_system(fplog, step, cr, TRUE, 1,
1806 state_global, top_global, ir,
1807 state, &f, mdatoms, top, fr,
1808 vsite, shellfc, constr,
1809 nrnb, wcycle, FALSE);
1814 bStartingFromCpt = FALSE;
1816 /* ####### SET VARIABLES FOR NEXT ITERATION IF THEY STILL NEED IT ###### */
1817 /* With all integrators, except VV, we need to retain the pressure
1818 * at the current step for coupling at the next step.
1820 if ((state->flags & (1<<estPRES_PREV)) &&
1822 (ir->nstpcouple > 0 && step % ir->nstpcouple == 0)))
1824 /* Store the pressure in t_state for pressure coupling
1825 * at the next MD step.
1827 copy_mat(pres, state->pres_prev);
1830 /* ####### END SET VARIABLES FOR NEXT ITERATION ###### */
1832 if ( (membed != NULL) && (!bLastStep) )
1834 rescale_membed(step_rel, membed, state_global->x);
1841 /* read next frame from input trajectory */
1842 bNotLastFrame = read_next_frame(oenv, status, &rerun_fr);
1847 rerun_parallel_comm(cr, &rerun_fr, &bNotLastFrame);
1851 if (!bRerunMD || !rerun_fr.bStep)
1853 /* increase the MD step number */
1858 cycles = wallcycle_stop(wcycle, ewcSTEP);
1859 if (DOMAINDECOMP(cr) && wcycle)
1861 dd_cycles_add(cr->dd, cycles, ddCyclStep);
1864 if (bPMETuneRunning || bPMETuneTry)
1866 /* PME grid + cut-off optimization with GPUs or PME nodes */
1868 /* Count the total cycles over the last steps */
1869 cycles_pmes += cycles;
1871 /* We can only switch cut-off at NS steps */
1872 if (step % ir->nstlist == 0)
1874 /* PME grid + cut-off optimization with GPUs or PME nodes */
1877 if (DDMASTER(cr->dd))
1879 /* PME node load is too high, start tuning */
1880 bPMETuneRunning = (dd_pme_f_ratio(cr->dd) >= 1.05);
1882 dd_bcast(cr->dd, sizeof(gmx_bool), &bPMETuneRunning);
1884 if (bPMETuneRunning || step_rel > ir->nstlist*50)
1886 bPMETuneTry = FALSE;
1889 if (bPMETuneRunning)
1891 /* init_step might not be a multiple of nstlist,
1892 * but the first cycle is always skipped anyhow.
1895 pme_load_balance(pme_loadbal, cr,
1896 (bVerbose && MASTER(cr)) ? stderr : NULL,
1898 ir, state, cycles_pmes,
1899 fr->ic, fr->nbv, &fr->pmedata,
1902 /* Update constants in forcerec/inputrec to keep them in sync with fr->ic */
1903 fr->ewaldcoeff_q = fr->ic->ewaldcoeff_q;
1904 fr->ewaldcoeff_lj = fr->ic->ewaldcoeff_lj;
1905 fr->rlist = fr->ic->rlist;
1906 fr->rlistlong = fr->ic->rlistlong;
1907 fr->rcoulomb = fr->ic->rcoulomb;
1908 fr->rvdw = fr->ic->rvdw;
1910 if (ir->eDispCorr != edispcNO)
1912 calc_enervirdiff(NULL, ir->eDispCorr, fr);
1919 if (step_rel == wcycle_get_reset_counters(wcycle) ||
1920 gs.set[eglsRESETCOUNTERS] != 0)
1922 /* Reset all the counters related to performance over the run */
1923 reset_all_counters(fplog, cr, step, &step_rel, ir, wcycle, nrnb, walltime_accounting,
1924 fr->nbv != NULL && fr->nbv->bUseGPU ? fr->nbv->cu_nbv : NULL);
1925 wcycle_set_reset_counters(wcycle, -1);
1926 if (!(cr->duty & DUTY_PME))
1928 /* Tell our PME node to reset its counters */
1929 gmx_pme_send_resetcounters(cr, step);
1931 /* Correct max_hours for the elapsed time */
1932 max_hours -= elapsed_time/(60.0*60.0);
1933 bResetCountersHalfMaxH = FALSE;
1934 gs.set[eglsRESETCOUNTERS] = 0;
1937 /* If bIMD is TRUE, the master updates the IMD energy record and sends positions to VMD client */
1938 IMD_prep_energies_send_positions(ir->bIMD && MASTER(cr), bIMDstep, ir->imd, enerd, step, bCalcEner, wcycle);
1941 /* End of main MD loop */
1944 /* Stop measuring walltime */
1945 walltime_accounting_end(walltime_accounting);
1947 if (bRerunMD && MASTER(cr))
1952 if (!(cr->duty & DUTY_PME))
1954 /* Tell the PME only node to finish */
1955 gmx_pme_send_finish(cr);
1960 if (ir->nstcalcenergy > 0 && !bRerunMD)
1962 print_ebin(mdoutf_get_fp_ene(outf), FALSE, FALSE, FALSE, fplog, step, t,
1963 eprAVER, FALSE, mdebin, fcd, groups, &(ir->opts));
1970 if (ir->nstlist == -1 && nlh.nns > 0 && fplog)
1972 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)));
1973 fprintf(fplog, "Average number of atoms that crossed the half buffer length: %.1f\n\n", nlh.ab/nlh.nns);
1976 if (pme_loadbal != NULL)
1978 pme_loadbal_done(pme_loadbal, cr, fplog,
1979 fr->nbv != NULL && fr->nbv->bUseGPU);
1982 if (shellfc && fplog)
1984 fprintf(fplog, "Fraction of iterations that converged: %.2f %%\n",
1985 (nconverged*100.0)/step_rel);
1986 fprintf(fplog, "Average number of force evaluations per MD step: %.2f\n\n",
1990 if (repl_ex_nst > 0 && MASTER(cr))
1992 print_replica_exchange_statistics(fplog, repl_ex);
1995 /* IMD cleanup, if bIMD is TRUE. */
1996 IMD_finalize(ir->bIMD, ir->imd);
1998 walltime_accounting_set_nsteps_done(walltime_accounting, step_rel);