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) 2011-2019,2020,2021, 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.
39 * \brief Implements the integrator for normal molecular dynamics simulations
41 * \author David van der Spoel <david.vanderspoel@icm.uu.se>
42 * \ingroup module_mdrun
55 #include "gromacs/applied_forces/awh/awh.h"
56 #include "gromacs/commandline/filenm.h"
57 #include "gromacs/domdec/collect.h"
58 #include "gromacs/domdec/dlbtiming.h"
59 #include "gromacs/domdec/domdec.h"
60 #include "gromacs/domdec/domdec_network.h"
61 #include "gromacs/domdec/domdec_struct.h"
62 #include "gromacs/domdec/gpuhaloexchange.h"
63 #include "gromacs/domdec/mdsetup.h"
64 #include "gromacs/domdec/partition.h"
65 #include "gromacs/essentialdynamics/edsam.h"
66 #include "gromacs/ewald/pme_load_balancing.h"
67 #include "gromacs/ewald/pme_pp.h"
68 #include "gromacs/fileio/trxio.h"
69 #include "gromacs/gmxlib/network.h"
70 #include "gromacs/gmxlib/nrnb.h"
71 #include "gromacs/gpu_utils/device_stream_manager.h"
72 #include "gromacs/gpu_utils/gpu_utils.h"
73 #include "gromacs/imd/imd.h"
74 #include "gromacs/listed_forces/listed_forces.h"
75 #include "gromacs/math/functions.h"
76 #include "gromacs/math/invertmatrix.h"
77 #include "gromacs/math/vec.h"
78 #include "gromacs/math/vectypes.h"
79 #include "gromacs/mdlib/checkpointhandler.h"
80 #include "gromacs/mdlib/compute_io.h"
81 #include "gromacs/mdlib/constr.h"
82 #include "gromacs/mdlib/coupling.h"
83 #include "gromacs/mdlib/ebin.h"
84 #include "gromacs/mdlib/enerdata_utils.h"
85 #include "gromacs/mdlib/energyoutput.h"
86 #include "gromacs/mdlib/expanded.h"
87 #include "gromacs/mdlib/force.h"
88 #include "gromacs/mdlib/force_flags.h"
89 #include "gromacs/mdlib/forcerec.h"
90 #include "gromacs/mdlib/freeenergyparameters.h"
91 #include "gromacs/mdlib/md_support.h"
92 #include "gromacs/mdlib/mdatoms.h"
93 #include "gromacs/mdlib/mdoutf.h"
94 #include "gromacs/mdlib/membed.h"
95 #include "gromacs/mdlib/resethandler.h"
96 #include "gromacs/mdlib/sighandler.h"
97 #include "gromacs/mdlib/simulationsignal.h"
98 #include "gromacs/mdlib/stat.h"
99 #include "gromacs/mdlib/stophandler.h"
100 #include "gromacs/mdlib/tgroup.h"
101 #include "gromacs/mdlib/trajectory_writing.h"
102 #include "gromacs/mdlib/update.h"
103 #include "gromacs/mdlib/update_constrain_gpu.h"
104 #include "gromacs/mdlib/update_vv.h"
105 #include "gromacs/mdlib/vcm.h"
106 #include "gromacs/mdlib/vsite.h"
107 #include "gromacs/mdrunutility/handlerestart.h"
108 #include "gromacs/mdrunutility/multisim.h"
109 #include "gromacs/mdrunutility/printtime.h"
110 #include "gromacs/mdtypes/awh_history.h"
111 #include "gromacs/mdtypes/awh_params.h"
112 #include "gromacs/mdtypes/commrec.h"
113 #include "gromacs/mdtypes/df_history.h"
114 #include "gromacs/mdtypes/energyhistory.h"
115 #include "gromacs/mdtypes/fcdata.h"
116 #include "gromacs/mdtypes/forcebuffers.h"
117 #include "gromacs/mdtypes/forcerec.h"
118 #include "gromacs/mdtypes/group.h"
119 #include "gromacs/mdtypes/inputrec.h"
120 #include "gromacs/mdtypes/interaction_const.h"
121 #include "gromacs/mdtypes/md_enums.h"
122 #include "gromacs/mdtypes/mdatom.h"
123 #include "gromacs/mdtypes/mdrunoptions.h"
124 #include "gromacs/mdtypes/multipletimestepping.h"
125 #include "gromacs/mdtypes/observableshistory.h"
126 #include "gromacs/mdtypes/pullhistory.h"
127 #include "gromacs/mdtypes/simulation_workload.h"
128 #include "gromacs/mdtypes/state.h"
129 #include "gromacs/mdtypes/state_propagator_data_gpu.h"
130 #include "gromacs/modularsimulator/energydata.h"
131 #include "gromacs/nbnxm/gpu_data_mgmt.h"
132 #include "gromacs/nbnxm/nbnxm.h"
133 #include "gromacs/pbcutil/pbc.h"
134 #include "gromacs/pulling/output.h"
135 #include "gromacs/pulling/pull.h"
136 #include "gromacs/swap/swapcoords.h"
137 #include "gromacs/timing/wallcycle.h"
138 #include "gromacs/timing/walltime_accounting.h"
139 #include "gromacs/topology/atoms.h"
140 #include "gromacs/topology/idef.h"
141 #include "gromacs/topology/mtop_util.h"
142 #include "gromacs/topology/topology.h"
143 #include "gromacs/trajectory/trajectoryframe.h"
144 #include "gromacs/utility/basedefinitions.h"
145 #include "gromacs/utility/cstringutil.h"
146 #include "gromacs/utility/fatalerror.h"
147 #include "gromacs/utility/logger.h"
148 #include "gromacs/utility/real.h"
149 #include "gromacs/utility/smalloc.h"
151 #include "legacysimulator.h"
152 #include "replicaexchange.h"
155 using gmx::SimulationSignaller;
157 void gmx::LegacySimulator::do_md()
159 // TODO Historically, the EM and MD "integrators" used different
160 // names for the t_inputrec *parameter, but these must have the
161 // same name, now that it's a member of a struct. We use this ir
162 // alias to avoid a large ripple of nearly useless changes.
163 // t_inputrec is being replaced by IMdpOptionsProvider, so this
164 // will go away eventually.
165 const t_inputrec* ir = inputrec;
167 int64_t step, step_rel;
168 double t, t0 = ir->init_t;
169 gmx_bool bGStatEveryStep, bGStat, bCalcVir, bCalcEnerStep, bCalcEner;
170 gmx_bool bNS = FALSE, bNStList, bStopCM, bFirstStep, bInitStep, bLastStep = FALSE;
171 gmx_bool bDoDHDL = FALSE, bDoFEP = FALSE, bDoExpanded = FALSE;
172 gmx_bool do_ene, do_log, do_verbose;
173 gmx_bool bMasterState;
174 unsigned int force_flags;
175 tensor force_vir = { { 0 } }, shake_vir = { { 0 } }, total_vir = { { 0 } }, pres = { { 0 } };
178 matrix pressureCouplingMu, M;
179 gmx_repl_ex_t repl_ex = nullptr;
180 gmx_global_stat_t gstat;
181 gmx_shellfc_t* shellfc;
182 gmx_bool bSumEkinhOld, bDoReplEx, bExchanged, bNeedRepartition;
185 std::vector<RVec> cbuf;
191 real saved_conserved_quantity = 0;
194 char sbuf[STEPSTRSIZE], sbuf2[STEPSTRSIZE];
196 /* PME load balancing data for GPU kernels */
197 gmx_bool bPMETune = FALSE;
198 gmx_bool bPMETunePrinting = FALSE;
200 bool bInteractiveMDstep = false;
202 /* Domain decomposition could incorrectly miss a bonded
203 interaction, but checking for that requires a global
204 communication stage, which does not otherwise happen in DD
205 code. So we do that alongside the first global energy reduction
206 after a new DD is made. These variables handle whether the
207 check happens, and the result it returns. */
208 bool shouldCheckNumberOfBondedInteractions = false;
209 int totalNumberOfBondedInteractions = -1;
211 SimulationSignals signals;
212 // Most global communnication stages don't propagate mdrun
213 // signals, and will use this object to achieve that.
214 SimulationSignaller nullSignaller(nullptr, nullptr, nullptr, false, false);
216 if (!mdrunOptions.writeConfout)
218 // This is on by default, and the main known use case for
219 // turning it off is for convenience in benchmarking, which is
220 // something that should not show up in the general user
225 "The -noconfout functionality is deprecated, and may be removed in a "
229 /* md-vv uses averaged full step velocities for T-control
230 md-vv-avek uses averaged half step velocities for T-control (but full step ekin for P control)
231 md uses averaged half step kinetic energies to determine temperature unless defined otherwise by GMX_EKIN_AVE_VEL; */
232 bTrotter = (EI_VV(ir->eI)
233 && (inputrecNptTrotter(ir) || inputrecNphTrotter(ir) || inputrecNvtTrotter(ir)));
235 const bool bRerunMD = false;
237 int nstglobalcomm = computeGlobalCommunicationPeriod(mdlog, ir, cr);
238 bGStatEveryStep = (nstglobalcomm == 1);
240 const SimulationGroups* groups = &top_global->groups;
242 std::unique_ptr<EssentialDynamics> ed = nullptr;
243 if (opt2bSet("-ei", nfile, fnm))
245 /* Initialize essential dynamics sampling */
246 ed = init_edsam(mdlog,
247 opt2fn_null("-ei", nfile, fnm),
248 opt2fn("-eo", nfile, fnm),
258 else if (observablesHistory->edsamHistory)
261 "The checkpoint is from a run with essential dynamics sampling, "
262 "but the current run did not specify the -ei option. "
263 "Either specify the -ei option to mdrun, or do not use this checkpoint file.");
266 int* fep_state = MASTER(cr) ? &state_global->fep_state : nullptr;
267 gmx::ArrayRef<real> lambda = MASTER(cr) ? state_global->lambda : gmx::ArrayRef<real>();
268 initialize_lambdas(fplog, *ir, MASTER(cr), fep_state, lambda);
269 Update upd(*ir, deform);
270 bool doSimulatedAnnealing = false;
272 // TODO: Avoid changing inputrec (#3854)
273 // Simulated annealing updates the reference temperature.
274 auto* nonConstInputrec = const_cast<t_inputrec*>(inputrec);
275 doSimulatedAnnealing = initSimulatedAnnealing(nonConstInputrec, &upd);
277 const bool useReplicaExchange = (replExParams.exchangeInterval > 0);
279 const t_fcdata& fcdata = *fr->fcdata;
281 bool simulationsShareState = false;
282 int nstSignalComm = nstglobalcomm;
284 // TODO This implementation of ensemble orientation restraints is nasty because
285 // a user can't just do multi-sim with single-sim orientation restraints.
286 bool usingEnsembleRestraints =
287 (fcdata.disres->nsystems > 1) || ((ms != nullptr) && (fcdata.orires->nr != 0));
288 bool awhUsesMultiSim = (ir->bDoAwh && ir->awhParams->shareBiasMultisim && (ms != nullptr));
290 // Replica exchange, ensemble restraints and AWH need all
291 // simulations to remain synchronized, so they need
292 // checkpoints and stop conditions to act on the same step, so
293 // the propagation of such signals must take place between
294 // simulations, not just within simulations.
295 // TODO: Make algorithm initializers set these flags.
296 simulationsShareState = useReplicaExchange || usingEnsembleRestraints || awhUsesMultiSim;
298 if (simulationsShareState)
300 // Inter-simulation signal communication does not need to happen
301 // often, so we use a minimum of 200 steps to reduce overhead.
302 const int c_minimumInterSimulationSignallingInterval = 200;
303 nstSignalComm = ((c_minimumInterSimulationSignallingInterval + nstglobalcomm - 1) / nstglobalcomm)
308 if (startingBehavior != StartingBehavior::RestartWithAppending)
310 pleaseCiteCouplingAlgorithms(fplog, *ir);
312 gmx_mdoutf* outf = init_mdoutf(fplog,
324 simulationsShareState,
326 gmx::EnergyOutput energyOutput(mdoutf_get_fp_ene(outf),
330 mdoutf_get_fp_dhdl(outf),
333 simulationsShareState,
336 gstat = global_stat_init(ir);
338 const auto& simulationWork = runScheduleWork->simulationWork;
339 const bool useGpuForPme = simulationWork.useGpuPme;
340 const bool useGpuForNonbonded = simulationWork.useGpuNonbonded;
341 const bool useGpuForBufferOps = simulationWork.useGpuBufferOps;
342 const bool useGpuForUpdate = simulationWork.useGpuUpdate;
344 /* Check for polarizable models and flexible constraints */
345 shellfc = init_shell_flexcon(fplog,
347 constr ? constr->numFlexibleConstraints() : 0,
353 double io = compute_io(ir, top_global->natoms, *groups, energyOutput.numEnergyTerms(), 1);
354 if ((io > 2000) && MASTER(cr))
356 fprintf(stderr, "\nWARNING: This run will generate roughly %.0f Mb of data\n\n", io);
360 // Local state only becomes valid now.
361 std::unique_ptr<t_state> stateInstance;
364 gmx_localtop_t top(top_global->ffparams);
366 auto mdatoms = mdAtoms->mdatoms();
368 ForceBuffers f(fr->useMts,
369 ((useGpuForNonbonded && useGpuForBufferOps) || useGpuForUpdate)
370 ? PinningPolicy::PinnedIfSupported
371 : PinningPolicy::CannotBePinned);
372 if (DOMAINDECOMP(cr))
374 stateInstance = std::make_unique<t_state>();
375 state = stateInstance.get();
376 dd_init_local_state(cr->dd, state_global, state);
378 /* Distribute the charge groups over the nodes from the master node */
379 dd_partition_system(fplog,
400 shouldCheckNumberOfBondedInteractions = true;
401 upd.setNumAtoms(state->natoms);
405 state_change_natoms(state_global, state_global->natoms);
406 /* Copy the pointer to the global state */
407 state = state_global;
409 /* Generate and initialize new topology */
410 mdAlgorithmsSetupAtomData(cr, ir, *top_global, &top, fr, &f, mdAtoms, constr, vsite, shellfc);
412 upd.setNumAtoms(state->natoms);
415 std::unique_ptr<UpdateConstrainGpu> integrator;
417 StatePropagatorDataGpu* stateGpu = fr->stateGpu;
419 // TODO: the assertions below should be handled by UpdateConstraintsBuilder.
422 GMX_RELEASE_ASSERT(!DOMAINDECOMP(cr) || ddUsesUpdateGroups(*cr->dd) || constr == nullptr
423 || constr->numConstraintsTotal() == 0,
424 "Constraints in domain decomposition are only supported with update "
425 "groups if using GPU update.\n");
426 GMX_RELEASE_ASSERT(ir->eConstrAlg != econtSHAKE || constr == nullptr
427 || constr->numConstraintsTotal() == 0,
428 "SHAKE is not supported with GPU update.");
429 GMX_RELEASE_ASSERT(useGpuForPme || (useGpuForNonbonded && simulationWork.useGpuBufferOps),
430 "Either PME or short-ranged non-bonded interaction tasks must run on "
431 "the GPU to use GPU update.\n");
432 GMX_RELEASE_ASSERT(ir->eI == eiMD,
433 "Only the md integrator is supported with the GPU update.\n");
435 ir->etc != etcNOSEHOOVER,
436 "Nose-Hoover temperature coupling is not supported with the GPU update.\n");
438 ir->epc == epcNO || ir->epc == epcPARRINELLORAHMAN || ir->epc == epcBERENDSEN
439 || ir->epc == epcCRESCALE,
440 "Only Parrinello-Rahman, Berendsen, and C-rescale pressure coupling are supported "
441 "with the GPU update.\n");
442 GMX_RELEASE_ASSERT(!mdatoms->haveVsites,
443 "Virtual sites are not supported with the GPU update.\n");
444 GMX_RELEASE_ASSERT(ed == nullptr,
445 "Essential dynamics is not supported with the GPU update.\n");
446 GMX_RELEASE_ASSERT(!ir->bPull || !pull_have_constraint(*ir->pull),
447 "Constraints pulling is not supported with the GPU update.\n");
448 GMX_RELEASE_ASSERT(fcdata.orires->nr == 0,
449 "Orientation restraints are not supported with the GPU update.\n");
452 || (!haveFepPerturbedMasses(*top_global) && !havePerturbedConstraints(*top_global)),
453 "Free energy perturbation of masses and constraints are not supported with the GPU "
456 if (constr != nullptr && constr->numConstraintsTotal() > 0)
460 .appendText("Updating coordinates and applying constraints on the GPU.");
464 GMX_LOG(mdlog.info).asParagraph().appendText("Updating coordinates on the GPU.");
466 GMX_RELEASE_ASSERT(fr->deviceStreamManager != nullptr,
467 "Device stream manager should be initialized in order to use GPU "
468 "update-constraints.");
470 fr->deviceStreamManager->streamIsValid(gmx::DeviceStreamType::UpdateAndConstraints),
471 "Update stream should be initialized in order to use GPU "
472 "update-constraints.");
473 integrator = std::make_unique<UpdateConstrainGpu>(
476 fr->deviceStreamManager->context(),
477 fr->deviceStreamManager->stream(gmx::DeviceStreamType::UpdateAndConstraints),
478 stateGpu->xUpdatedOnDevice(),
481 integrator->setPbc(PbcType::Xyz, state->box);
484 if (useGpuForPme || (useGpuForNonbonded && useGpuForBufferOps) || useGpuForUpdate)
486 changePinningPolicy(&state->x, PinningPolicy::PinnedIfSupported);
490 changePinningPolicy(&state->v, PinningPolicy::PinnedIfSupported);
493 // NOTE: The global state is no longer used at this point.
494 // But state_global is still used as temporary storage space for writing
495 // the global state to file and potentially for replica exchange.
496 // (Global topology should persist.)
498 update_mdatoms(mdatoms, state->lambda[efptMASS]);
502 /* Check nstexpanded here, because the grompp check was broken */
503 if (ir->expandedvals->nstexpanded % ir->nstcalcenergy != 0)
506 "With expanded ensemble, nstexpanded should be a multiple of nstcalcenergy");
508 init_expanded_ensemble(startingBehavior != StartingBehavior::NewSimulation, ir, state->dfhist);
513 EnergyData::initializeEnergyHistory(startingBehavior, observablesHistory, &energyOutput);
516 preparePrevStepPullCom(
517 ir, pull_work, mdatoms->massT, state, state_global, cr, startingBehavior != StartingBehavior::NewSimulation);
519 // TODO: Remove this by converting AWH into a ForceProvider
520 auto awh = prepareAwhModule(fplog,
525 startingBehavior != StartingBehavior::NewSimulation,
527 opt2fn("-awh", nfile, fnm),
530 if (useReplicaExchange && MASTER(cr))
532 repl_ex = init_replica_exchange(fplog, ms, top_global->natoms, ir, replExParams);
534 /* PME tuning is only supported in the Verlet scheme, with PME for
535 * Coulomb. It is not supported with only LJ PME. */
536 bPMETune = (mdrunOptions.tunePme && EEL_PME(fr->ic->eeltype) && !mdrunOptions.reproducible
537 && ir->cutoff_scheme != ecutsGROUP);
539 pme_load_balancing_t* pme_loadbal = nullptr;
543 &pme_loadbal, cr, mdlog, *ir, state->box, *fr->ic, *fr->nbv, fr->pmedata, fr->nbv->useGpu());
546 if (!ir->bContinuation)
548 if (state->flags & (1U << estV))
550 auto v = makeArrayRef(state->v);
551 /* Set the velocities of vsites, shells and frozen atoms to zero */
552 for (i = 0; i < mdatoms->homenr; i++)
554 if (mdatoms->ptype[i] == eptVSite || mdatoms->ptype[i] == eptShell)
558 else if (mdatoms->cFREEZE)
560 for (m = 0; m < DIM; m++)
562 if (ir->opts.nFreeze[mdatoms->cFREEZE[i]][m])
573 /* Constrain the initial coordinates and velocities */
574 do_constrain_first(fplog,
579 state->x.arrayRefWithPadding(),
580 state->v.arrayRefWithPadding(),
582 state->lambda[efptBONDED]);
586 /* Construct the virtual sites for the initial configuration */
587 vsite->construct(state->x, ir->delta_t, {}, state->box);
591 if (ir->efep != efepNO)
593 /* Set free energy calculation frequency as the greatest common
594 * denominator of nstdhdl and repl_ex_nst. */
595 nstfep = ir->fepvals->nstdhdl;
598 nstfep = std::gcd(ir->expandedvals->nstexpanded, nstfep);
600 if (useReplicaExchange)
602 nstfep = std::gcd(replExParams.exchangeInterval, nstfep);
606 nstfep = std::gcd(ir->awhParams->nstSampleCoord, nstfep);
610 /* Be REALLY careful about what flags you set here. You CANNOT assume
611 * this is the first step, since we might be restarting from a checkpoint,
612 * and in that case we should not do any modifications to the state.
614 bStopCM = (ir->comm_mode != ecmNO && !ir->bContinuation);
616 // When restarting from a checkpoint, it can be appropriate to
617 // initialize ekind from quantities in the checkpoint. Otherwise,
618 // compute_globals must initialize ekind before the simulation
619 // starts/restarts. However, only the master rank knows what was
620 // found in the checkpoint file, so we have to communicate in
621 // order to coordinate the restart.
623 // TODO Consider removing this communication if/when checkpoint
624 // reading directly follows .tpr reading, because all ranks can
625 // agree on hasReadEkinState at that time.
626 bool hasReadEkinState = MASTER(cr) ? state_global->ekinstate.hasReadEkinState : false;
629 gmx_bcast(sizeof(hasReadEkinState), &hasReadEkinState, cr->mpi_comm_mygroup);
631 if (hasReadEkinState)
633 restore_ekinstate_from_state(cr, ekind, &state_global->ekinstate);
636 unsigned int cglo_flags =
637 (CGLO_TEMPERATURE | CGLO_GSTAT | (EI_VV(ir->eI) ? CGLO_PRESSURE : 0)
638 | (EI_VV(ir->eI) ? CGLO_CONSTRAINT : 0) | (hasReadEkinState ? CGLO_READEKIN : 0));
640 bSumEkinhOld = FALSE;
642 t_vcm vcm(top_global->groups, *ir);
643 reportComRemovalInfo(fplog, vcm);
645 /* To minimize communication, compute_globals computes the COM velocity
646 * and the kinetic energy for the velocities without COM motion removed.
647 * Thus to get the kinetic energy without the COM contribution, we need
648 * to call compute_globals twice.
650 for (int cgloIteration = 0; cgloIteration < (bStopCM ? 2 : 1); cgloIteration++)
652 unsigned int cglo_flags_iteration = cglo_flags;
653 if (bStopCM && cgloIteration == 0)
655 cglo_flags_iteration |= CGLO_STOPCM;
656 cglo_flags_iteration &= ~CGLO_TEMPERATURE;
658 compute_globals(gstat,
663 makeConstArrayRef(state->x),
664 makeConstArrayRef(state->v),
678 &totalNumberOfBondedInteractions,
681 | (shouldCheckNumberOfBondedInteractions ? CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS
683 if (cglo_flags_iteration & CGLO_STOPCM)
685 /* At initialization, do not pass x with acceleration-correction mode
686 * to avoid (incorrect) correction of the initial coordinates.
688 auto x = (vcm.mode == ecmLINEAR_ACCELERATION_CORRECTION) ? ArrayRef<RVec>()
689 : makeArrayRef(state->x);
690 process_and_stopcm_grp(fplog, &vcm, *mdatoms, x, makeArrayRef(state->v));
691 inc_nrnb(nrnb, eNR_STOPCM, mdatoms->homenr);
694 checkNumberOfBondedInteractions(mdlog,
696 totalNumberOfBondedInteractions,
699 makeConstArrayRef(state->x),
701 &shouldCheckNumberOfBondedInteractions);
702 if (ir->eI == eiVVAK)
704 /* a second call to get the half step temperature initialized as well */
705 /* we do the same call as above, but turn the pressure off -- internally to
706 compute_globals, this is recognized as a velocity verlet half-step
707 kinetic energy calculation. This minimized excess variables, but
708 perhaps loses some logic?*/
710 compute_globals(gstat,
715 makeConstArrayRef(state->x),
716 makeConstArrayRef(state->v),
732 cglo_flags & ~CGLO_PRESSURE);
735 /* Calculate the initial half step temperature, and save the ekinh_old */
736 if (startingBehavior == StartingBehavior::NewSimulation)
738 for (i = 0; (i < ir->opts.ngtc); i++)
740 copy_mat(ekind->tcstat[i].ekinh, ekind->tcstat[i].ekinh_old);
744 /* need to make an initiation call to get the Trotter variables set, as well as other constants
745 for non-trotter temperature control */
746 auto trotter_seq = init_npt_vars(ir, state, &MassQ, bTrotter);
750 if (!ir->bContinuation)
752 if (constr && ir->eConstrAlg == econtLINCS)
755 "RMS relative constraint deviation after constraining: %.2e\n",
758 if (EI_STATE_VELOCITY(ir->eI))
760 real temp = enerd->term[F_TEMP];
763 /* Result of Ekin averaged over velocities of -half
764 * and +half step, while we only have -half step here.
768 fprintf(fplog, "Initial temperature: %g K\n", temp);
773 fprintf(stderr, "starting mdrun '%s'\n", *(top_global->name));
776 sprintf(tbuf, "%8.1f", (ir->init_step + ir->nsteps) * ir->delta_t);
780 sprintf(tbuf, "%s", "infinite");
782 if (ir->init_step > 0)
785 "%s steps, %s ps (continuing from step %s, %8.1f ps).\n",
786 gmx_step_str(ir->init_step + ir->nsteps, sbuf),
788 gmx_step_str(ir->init_step, sbuf2),
789 ir->init_step * ir->delta_t);
793 fprintf(stderr, "%s steps, %s ps.\n", gmx_step_str(ir->nsteps, sbuf), tbuf);
795 fprintf(fplog, "\n");
798 walltime_accounting_start_time(walltime_accounting);
799 wallcycle_start(wcycle, ewcRUN);
800 print_start(fplog, cr, walltime_accounting, "mdrun");
802 /***********************************************************
806 ************************************************************/
809 /* Skip the first Nose-Hoover integration when we get the state from tpx */
810 bInitStep = startingBehavior == StartingBehavior::NewSimulation || EI_VV(ir->eI);
811 bSumEkinhOld = FALSE;
813 bNeedRepartition = FALSE;
815 step = ir->init_step;
818 auto stopHandler = stopHandlerBuilder->getStopHandlerMD(
819 compat::not_null<SimulationSignal*>(&signals[eglsSTOPCOND]),
820 simulationsShareState,
823 mdrunOptions.reproducible,
825 mdrunOptions.maximumHoursToRun,
830 walltime_accounting);
832 auto checkpointHandler = std::make_unique<CheckpointHandler>(
833 compat::make_not_null<SimulationSignal*>(&signals[eglsCHKPT]),
834 simulationsShareState,
837 mdrunOptions.writeConfout,
838 mdrunOptions.checkpointOptions.period);
840 const bool resetCountersIsLocal = true;
841 auto resetHandler = std::make_unique<ResetHandler>(
842 compat::make_not_null<SimulationSignal*>(&signals[eglsRESETCOUNTERS]),
843 !resetCountersIsLocal,
846 mdrunOptions.timingOptions.resetHalfway,
847 mdrunOptions.maximumHoursToRun,
850 walltime_accounting);
852 const DDBalanceRegionHandler ddBalanceRegionHandler(cr);
854 if (MASTER(cr) && isMultiSim(ms) && !useReplicaExchange)
856 logInitialMultisimStatus(ms, cr, mdlog, simulationsShareState, ir->nsteps, ir->init_step);
859 /* and stop now if we should */
860 bLastStep = (bLastStep || (ir->nsteps >= 0 && step_rel > ir->nsteps));
864 /* Determine if this is a neighbor search step */
865 bNStList = (ir->nstlist > 0 && step % ir->nstlist == 0);
867 if (bPMETune && bNStList)
869 // This has to be here because PME load balancing is called so early.
870 // TODO: Move to after all booleans are defined.
871 if (useGpuForUpdate && !bFirstStep)
873 stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
874 stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
876 /* PME grid + cut-off optimization with GPUs or PME nodes */
877 pme_loadbal_do(pme_loadbal,
879 (mdrunOptions.verbose && MASTER(cr)) ? stderr : nullptr,
890 simulationWork.useGpuPmePpCommunication);
893 wallcycle_start(wcycle, ewcSTEP);
895 bLastStep = (step_rel == ir->nsteps);
896 t = t0 + step * ir->delta_t;
898 // TODO Refactor this, so that nstfep does not need a default value of zero
899 if (ir->efep != efepNO || ir->bSimTemp)
901 /* find and set the current lambdas */
902 state->lambda = currentLambdas(step, *(ir->fepvals), state->fep_state);
904 bDoDHDL = do_per_step(step, ir->fepvals->nstdhdl);
905 bDoFEP = ((ir->efep != efepNO) && do_per_step(step, nstfep));
906 bDoExpanded = (do_per_step(step, ir->expandedvals->nstexpanded) && (ir->bExpanded)
910 bDoReplEx = (useReplicaExchange && (step > 0) && !bLastStep
911 && do_per_step(step, replExParams.exchangeInterval));
913 if (doSimulatedAnnealing)
915 // TODO: Avoid changing inputrec (#3854)
916 // Simulated annealing updates the reference temperature.
917 auto* nonConstInputrec = const_cast<t_inputrec*>(inputrec);
918 update_annealing_target_temp(nonConstInputrec, t, &upd);
921 /* Stop Center of Mass motion */
922 bStopCM = (ir->comm_mode != ecmNO && do_per_step(step, ir->nstcomm));
924 /* Determine whether or not to do Neighbour Searching */
925 bNS = (bFirstStep || bNStList || bExchanged || bNeedRepartition);
927 /* Note that the stopHandler will cause termination at nstglobalcomm
928 * steps. Since this concides with nstcalcenergy, nsttcouple and/or
929 * nstpcouple steps, we have computed the half-step kinetic energy
930 * of the previous step and can always output energies at the last step.
932 bLastStep = bLastStep || stopHandler->stoppingAfterCurrentStep(bNS);
934 /* do_log triggers energy and virial calculation. Because this leads
935 * to different code paths, forces can be different. Thus for exact
936 * continuation we should avoid extra log output.
937 * Note that the || bLastStep can result in non-exact continuation
938 * beyond the last step. But we don't consider that to be an issue.
940 do_log = (do_per_step(step, ir->nstlog)
941 || (bFirstStep && startingBehavior == StartingBehavior::NewSimulation) || bLastStep);
942 do_verbose = mdrunOptions.verbose
943 && (step % mdrunOptions.verboseStepPrintInterval == 0 || bFirstStep || bLastStep);
945 if (useGpuForUpdate && !bFirstStep && bNS)
947 // Copy velocities from the GPU on search steps to keep a copy on host (device buffers are reinitialized).
948 stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
949 stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
950 // Copy coordinate from the GPU when needed at the search step.
951 // NOTE: The cases when coordinates needed on CPU for force evaluation are handled in sim_utils.
952 // NOTE: If the coordinates are to be written into output file they are also copied separately before the output.
953 stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
954 stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
957 if (bNS && !(bFirstStep && ir->bContinuation))
959 bMasterState = FALSE;
960 /* Correct the new box if it is too skewed */
961 if (inputrecDynamicBox(ir))
963 if (correct_box(fplog, step, state->box))
966 // If update is offloaded, it should be informed about the box size change
969 integrator->setPbc(PbcType::Xyz, state->box);
973 if (DOMAINDECOMP(cr) && bMasterState)
975 dd_collect_state(cr->dd, state, state_global);
978 if (DOMAINDECOMP(cr))
980 /* Repartition the domain decomposition */
981 dd_partition_system(fplog,
1001 do_verbose && !bPMETunePrinting);
1002 shouldCheckNumberOfBondedInteractions = true;
1003 upd.setNumAtoms(state->natoms);
1007 // Allocate or re-size GPU halo exchange object, if necessary
1008 if (bNS && havePPDomainDecomposition(cr) && simulationWork.useGpuHaloExchange)
1010 GMX_RELEASE_ASSERT(fr->deviceStreamManager != nullptr,
1011 "GPU device manager has to be initialized to use GPU "
1012 "version of halo exchange.");
1013 constructGpuHaloExchange(mdlog, *cr, *fr->deviceStreamManager, wcycle);
1016 if (MASTER(cr) && do_log)
1018 gmx::EnergyOutput::printHeader(
1019 fplog, step, t); /* can we improve the information printed here? */
1022 if (ir->efep != efepNO)
1024 update_mdatoms(mdatoms, state->lambda[efptMASS]);
1029 /* We need the kinetic energy at minus the half step for determining
1030 * the full step kinetic energy and possibly for T-coupling.*/
1031 /* This may not be quite working correctly yet . . . . */
1032 compute_globals(gstat,
1037 makeConstArrayRef(state->x),
1038 makeConstArrayRef(state->v),
1052 &totalNumberOfBondedInteractions,
1054 CGLO_GSTAT | CGLO_TEMPERATURE | CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS);
1055 checkNumberOfBondedInteractions(mdlog,
1057 totalNumberOfBondedInteractions,
1060 makeConstArrayRef(state->x),
1062 &shouldCheckNumberOfBondedInteractions);
1064 clear_mat(force_vir);
1066 checkpointHandler->decideIfCheckpointingThisStep(bNS, bFirstStep, bLastStep);
1068 /* Determine the energy and pressure:
1069 * at nstcalcenergy steps and at energy output steps (set below).
1071 if (EI_VV(ir->eI) && (!bInitStep))
1073 bCalcEnerStep = do_per_step(step, ir->nstcalcenergy);
1074 bCalcVir = bCalcEnerStep
1075 || (ir->epc != epcNO
1076 && (do_per_step(step, ir->nstpcouple) || do_per_step(step - 1, ir->nstpcouple)));
1080 bCalcEnerStep = do_per_step(step, ir->nstcalcenergy);
1081 bCalcVir = bCalcEnerStep || (ir->epc != epcNO && do_per_step(step, ir->nstpcouple));
1083 bCalcEner = bCalcEnerStep;
1085 do_ene = (do_per_step(step, ir->nstenergy) || bLastStep);
1087 if (do_ene || do_log || bDoReplEx)
1093 /* Do we need global communication ? */
1094 bGStat = (bCalcVir || bCalcEner || bStopCM || do_per_step(step, nstglobalcomm)
1095 || (EI_VV(ir->eI) && inputrecNvtTrotter(ir) && do_per_step(step - 1, nstglobalcomm)));
1097 force_flags = (GMX_FORCE_STATECHANGED | ((inputrecDynamicBox(ir)) ? GMX_FORCE_DYNAMICBOX : 0)
1098 | GMX_FORCE_ALLFORCES | (bCalcVir ? GMX_FORCE_VIRIAL : 0)
1099 | (bCalcEner ? GMX_FORCE_ENERGY : 0) | (bDoFEP ? GMX_FORCE_DHDL : 0));
1100 if (fr->useMts && !do_per_step(step, ir->nstfout))
1102 force_flags |= GMX_FORCE_DO_NOT_NEED_NORMAL_FORCE;
1107 /* Now is the time to relax the shells */
1108 relax_shell_flexcon(fplog,
1111 mdrunOptions.verbose,
1123 state->x.arrayRefWithPadding(),
1124 state->v.arrayRefWithPadding(),
1139 ddBalanceRegionHandler);
1143 /* The AWH history need to be saved _before_ doing force calculations where the AWH bias
1144 is updated (or the AWH update will be performed twice for one step when continuing).
1145 It would be best to call this update function from do_md_trajectory_writing but that
1146 would occur after do_force. One would have to divide the update_awh function into one
1147 function applying the AWH force and one doing the AWH bias update. The update AWH
1148 bias function could then be called after do_md_trajectory_writing (then containing
1149 update_awh_history). The checkpointing will in the future probably moved to the start
1150 of the md loop which will rid of this issue. */
1151 if (awh && checkpointHandler->isCheckpointingStep() && MASTER(cr))
1153 awh->updateHistory(state_global->awhHistory.get());
1156 /* The coordinates (x) are shifted (to get whole molecules)
1158 * This is parallellized as well, and does communication too.
1159 * Check comments in sim_util.c
1174 state->x.arrayRefWithPadding(),
1186 ed ? ed->getLegacyED() : nullptr,
1187 (bNS ? GMX_FORCE_NS : 0) | force_flags,
1188 ddBalanceRegionHandler);
1191 // VV integrators do not need the following velocity half step
1192 // if it is the first step after starting from a checkpoint.
1193 // That is, the half step is needed on all other steps, and
1194 // also the first step when starting from a .tpr file.
1197 integrateVVFirstStep(step,
1230 &shouldCheckNumberOfBondedInteractions,
1231 &saved_conserved_quantity,
1243 /* ######## END FIRST UPDATE STEP ############## */
1244 /* ######## If doing VV, we now have v(dt) ###### */
1247 /* perform extended ensemble sampling in lambda - we don't
1248 actually move to the new state before outputting
1249 statistics, but if performing simulated tempering, we
1250 do update the velocities and the tau_t. */
1251 // TODO: Avoid changing inputrec (#3854)
1252 // Simulated tempering updates the reference temperature.
1253 // Expanded ensemble without simulated tempering does not change the inputrec.
1254 auto* nonConstInputrec = const_cast<t_inputrec*>(inputrec);
1255 lamnew = ExpandedEnsembleDynamics(fplog,
1263 state->v.rvec_array(),
1265 /* history is maintained in state->dfhist, but state_global is what is sent to trajectory and log output */
1268 copy_df_history(state_global->dfhist, state->dfhist);
1272 // Copy coordinate from the GPU for the output/checkpointing if the update is offloaded and
1273 // coordinates have not already been copied for i) search or ii) CPU force tasks.
1274 if (useGpuForUpdate && !bNS && !runScheduleWork->domainWork.haveCpuLocalForceWork
1275 && (do_per_step(step, ir->nstxout) || do_per_step(step, ir->nstxout_compressed)
1276 || checkpointHandler->isCheckpointingStep()))
1278 stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
1279 stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
1281 // Copy velocities if needed for the output/checkpointing.
1282 // NOTE: Copy on the search steps is done at the beginning of the step.
1283 if (useGpuForUpdate && !bNS
1284 && (do_per_step(step, ir->nstvout) || checkpointHandler->isCheckpointingStep()))
1286 stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
1287 stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
1289 // Copy forces for the output if the forces were reduced on the GPU (not the case on virial steps)
1290 // and update is offloaded hence forces are kept on the GPU for update and have not been
1291 // already transferred in do_force().
1292 // TODO: There should be an improved, explicit mechanism that ensures this copy is only executed
1293 // when the forces are ready on the GPU -- the same synchronizer should be used as the one
1294 // prior to GPU update.
1295 // TODO: When the output flags will be included in step workload, this copy can be combined with the
1296 // copy call in do_force(...).
1297 // NOTE: The forces should not be copied here if the vsites are present, since they were modified
1298 // on host after the D2H copy in do_force(...).
1299 if (runScheduleWork->stepWork.useGpuFBufferOps && (simulationWork.useGpuUpdate && !vsite)
1300 && do_per_step(step, ir->nstfout))
1302 stateGpu->copyForcesFromGpu(f.view().force(), AtomLocality::Local);
1303 stateGpu->waitForcesReadyOnHost(AtomLocality::Local);
1305 /* Now we have the energies and forces corresponding to the
1306 * coordinates at time t. We must output all of this before
1309 do_md_trajectory_writing(fplog,
1326 checkpointHandler->isCheckpointingStep(),
1329 mdrunOptions.writeConfout,
1331 /* Check if IMD step and do IMD communication, if bIMD is TRUE. */
1332 bInteractiveMDstep = imdSession->run(step, bNS, state->box, state->x.rvec_array(), t);
1334 /* kludge -- virial is lost with restart for MTTK NPT control. Must reload (saved earlier). */
1335 if (startingBehavior != StartingBehavior::NewSimulation && bFirstStep
1336 && (inputrecNptTrotter(ir) || inputrecNphTrotter(ir)))
1338 copy_mat(state->svir_prev, shake_vir);
1339 copy_mat(state->fvir_prev, force_vir);
1342 stopHandler->setSignal();
1343 resetHandler->setSignal(walltime_accounting);
1345 if (bGStat || !PAR(cr))
1347 /* In parallel we only have to check for checkpointing in steps
1348 * where we do global communication,
1349 * otherwise the other nodes don't know.
1351 checkpointHandler->setSignal(walltime_accounting);
1354 /* ######### START SECOND UPDATE STEP ################# */
1356 /* at the start of step, randomize or scale the velocities ((if vv. Restriction of Andersen
1357 controlled in preprocessing */
1359 if (ETC_ANDERSEN(ir->etc)) /* keep this outside of update_tcouple because of the extra info required to pass */
1361 gmx_bool bIfRandomize;
1362 bIfRandomize = update_randomize_velocities(ir, step, cr, mdatoms, state->v, &upd, constr);
1363 /* if we have constraints, we have to remove the kinetic energy parallel to the bonds */
1364 if (constr && bIfRandomize)
1366 constrain_velocities(constr, do_log, do_ene, step, state, nullptr, false, nullptr);
1369 /* Box is changed in update() when we do pressure coupling,
1370 * but we should still use the old box for energy corrections and when
1371 * writing it to the energy file, so it matches the trajectory files for
1372 * the same timestep above. Make a copy in a separate array.
1374 copy_mat(state->box, lastbox);
1378 if (!useGpuForUpdate)
1380 wallcycle_start(wcycle, ewcUPDATE);
1382 /* UPDATE PRESSURE VARIABLES IN TROTTER FORMULATION WITH CONSTRAINTS */
1385 trotter_update(ir, step, ekind, enerd, state, total_vir, mdatoms, &MassQ, trotter_seq, ettTSEQ3);
1386 /* We can only do Berendsen coupling after we have summed
1387 * the kinetic energy or virial. Since the happens
1388 * in global_state after update, we should only do it at
1389 * step % nstlist = 1 with bGStatEveryStep=FALSE.
1394 update_tcouple(step, ir, state, ekind, &MassQ, mdatoms);
1395 update_pcouple_before_coordinates(fplog, step, ir, state, pressureCouplingMu, M, bInitStep);
1398 /* With leap-frog type integrators we compute the kinetic energy
1399 * at a whole time step as the average of the half-time step kinetic
1400 * energies of two subsequent steps. Therefore we need to compute the
1401 * half step kinetic energy also if we need energies at the next step.
1403 const bool needHalfStepKineticEnergy =
1404 (!EI_VV(ir->eI) && (do_per_step(step + 1, nstglobalcomm) || step_rel + 1 == ir->nsteps));
1406 // Parrinello-Rahman requires the pressure to be availible before the update to compute
1407 // the velocity scaling matrix. Hence, it runs one step after the nstpcouple step.
1408 const bool doParrinelloRahman = (ir->epc == epcPARRINELLORAHMAN
1409 && do_per_step(step + ir->nstpcouple - 1, ir->nstpcouple));
1413 GMX_ASSERT(!useGpuForUpdate, "GPU update is not supported with VVAK integrator.");
1415 integrateVVSecondStep(step,
1451 if (useGpuForUpdate)
1454 wallcycle_stop(wcycle, ewcUPDATE);
1456 if (bNS && (bFirstStep || DOMAINDECOMP(cr)))
1458 integrator->set(stateGpu->getCoordinates(),
1459 stateGpu->getVelocities(),
1460 stateGpu->getForces(),
1465 // Copy data to the GPU after buffers might have being reinitialized
1466 stateGpu->copyVelocitiesToGpu(state->v, AtomLocality::Local);
1467 stateGpu->copyCoordinatesToGpu(state->x, AtomLocality::Local);
1470 if (simulationWork.useGpuPme && !runScheduleWork->simulationWork.useGpuPmePpCommunication
1471 && !thisRankHasDuty(cr, DUTY_PME))
1473 // The PME forces were recieved to the host, so have to be copied
1474 stateGpu->copyForcesToGpu(f.view().force(), AtomLocality::All);
1476 else if (!runScheduleWork->stepWork.useGpuFBufferOps)
1478 // The buffer ops were not offloaded this step, so the forces are on the
1479 // host and have to be copied
1480 stateGpu->copyForcesToGpu(f.view().force(), AtomLocality::Local);
1483 const bool doTemperatureScaling =
1484 (ir->etc != etcNO && do_per_step(step + ir->nsttcouple - 1, ir->nsttcouple));
1486 // This applies Leap-Frog, LINCS and SETTLE in succession
1487 integrator->integrate(
1488 stateGpu->getForcesReadyOnDeviceEvent(
1489 AtomLocality::Local, runScheduleWork->stepWork.useGpuFBufferOps),
1494 doTemperatureScaling,
1497 ir->nstpcouple * ir->delta_t,
1500 // Copy velocities D2H after update if:
1501 // - Globals are computed this step (includes the energy output steps).
1502 // - Temperature is needed for the next step.
1503 if (bGStat || needHalfStepKineticEnergy)
1505 stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
1506 stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
1511 /* With multiple time stepping we need to do an additional normal
1512 * update step to obtain the virial, as the actual MTS integration
1513 * using an acceleration where the slow forces are multiplied by mtsFactor.
1514 * Using that acceleration would result in a virial with the slow
1515 * force contribution would be a factor mtsFactor too large.
1517 if (fr->useMts && bCalcVir && constr != nullptr)
1519 upd.update_for_constraint_virial(
1520 *ir, *mdatoms, *state, f.view().forceWithPadding(), *ekind);
1522 constrain_coordinates(constr,
1527 upd.xp()->arrayRefWithPadding(),
1533 ArrayRefWithPadding<const RVec> forceCombined =
1534 (fr->useMts && step % ir->mtsLevels[1].stepFactor == 0)
1535 ? f.view().forceMtsCombinedWithPadding()
1536 : f.view().forceWithPadding();
1538 *ir, step, mdatoms, state, forceCombined, fcdata, ekind, M, etrtPOSITION, cr, constr != nullptr);
1540 wallcycle_stop(wcycle, ewcUPDATE);
1542 constrain_coordinates(constr,
1547 upd.xp()->arrayRefWithPadding(),
1549 bCalcVir && !fr->useMts,
1552 upd.update_sd_second_half(
1553 *ir, step, &dvdl_constr, mdatoms, state, cr, nrnb, wcycle, constr, do_log, do_ene);
1554 upd.finish_update(*ir, mdatoms, state, wcycle, constr != nullptr);
1557 if (ir->bPull && ir->pull->bSetPbcRefToPrevStepCOM)
1559 updatePrevStepPullCom(pull_work, state);
1562 enerd->term[F_DVDL_CONSTR] += dvdl_constr;
1565 if (vsite != nullptr)
1567 wallcycle_start(wcycle, ewcVSITECONSTR);
1568 vsite->construct(state->x, ir->delta_t, state->v, state->box);
1569 wallcycle_stop(wcycle, ewcVSITECONSTR);
1572 /* ############## IF NOT VV, Calculate globals HERE ############ */
1573 /* With Leap-Frog we can skip compute_globals at
1574 * non-communication steps, but we need to calculate
1575 * the kinetic energy one step before communication.
1578 // Organize to do inter-simulation signalling on steps if
1579 // and when algorithms require it.
1580 const bool doInterSimSignal = (simulationsShareState && do_per_step(step, nstSignalComm));
1582 if (bGStat || needHalfStepKineticEnergy || doInterSimSignal)
1584 // Copy coordinates when needed to stop the CM motion.
1585 if (useGpuForUpdate && !EI_VV(ir->eI) && bStopCM)
1587 stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
1588 stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
1590 // Since we're already communicating at this step, we
1591 // can propagate intra-simulation signals. Note that
1592 // check_nstglobalcomm has the responsibility for
1593 // choosing the value of nstglobalcomm that is one way
1594 // bGStat becomes true, so we can't get into a
1595 // situation where e.g. checkpointing can't be
1597 bool doIntraSimSignal = true;
1598 SimulationSignaller signaller(&signals, cr, ms, doInterSimSignal, doIntraSimSignal);
1600 compute_globals(gstat,
1605 makeConstArrayRef(state->x),
1606 makeConstArrayRef(state->v),
1620 &totalNumberOfBondedInteractions,
1622 (bGStat ? CGLO_GSTAT : 0) | (!EI_VV(ir->eI) && bCalcEner ? CGLO_ENERGY : 0)
1623 | (!EI_VV(ir->eI) && bStopCM ? CGLO_STOPCM : 0)
1624 | (!EI_VV(ir->eI) ? CGLO_TEMPERATURE : 0)
1625 | (!EI_VV(ir->eI) ? CGLO_PRESSURE : 0) | CGLO_CONSTRAINT
1626 | (shouldCheckNumberOfBondedInteractions ? CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS
1628 checkNumberOfBondedInteractions(mdlog,
1630 totalNumberOfBondedInteractions,
1633 makeConstArrayRef(state->x),
1635 &shouldCheckNumberOfBondedInteractions);
1636 if (!EI_VV(ir->eI) && bStopCM)
1638 process_and_stopcm_grp(
1639 fplog, &vcm, *mdatoms, makeArrayRef(state->x), makeArrayRef(state->v));
1640 inc_nrnb(nrnb, eNR_STOPCM, mdatoms->homenr);
1642 // TODO: The special case of removing CM motion should be dealt more gracefully
1643 if (useGpuForUpdate)
1645 stateGpu->copyCoordinatesToGpu(state->x, AtomLocality::Local);
1646 // Here we block until the H2D copy completes because event sync with the
1647 // force kernels that use the coordinates on the next steps is not implemented
1648 // (not because of a race on state->x being modified on the CPU while H2D is in progress).
1649 stateGpu->waitCoordinatesCopiedToDevice(AtomLocality::Local);
1650 // If the COM removal changed the velocities on the CPU, this has to be accounted for.
1651 if (vcm.mode != ecmNO)
1653 stateGpu->copyVelocitiesToGpu(state->v, AtomLocality::Local);
1660 /* ############# END CALC EKIN AND PRESSURE ################# */
1662 /* Note: this is OK, but there are some numerical precision issues with using the convergence of
1663 the virial that should probably be addressed eventually. state->veta has better properies,
1664 but what we actually need entering the new cycle is the new shake_vir value. Ideally, we could
1665 generate the new shake_vir, but test the veta value for convergence. This will take some thought. */
1667 if (ir->efep != efepNO && !EI_VV(ir->eI))
1669 /* Sum up the foreign energy and dK/dl terms for md and sd.
1670 Currently done every step so that dH/dl is correct in the .edr */
1671 accumulateKineticLambdaComponents(enerd, state->lambda, *ir->fepvals);
1674 update_pcouple_after_coordinates(
1675 fplog, step, ir, mdatoms, pres, force_vir, shake_vir, pressureCouplingMu, state, nrnb, upd.deform(), !useGpuForUpdate);
1677 const bool doBerendsenPressureCoupling =
1678 (inputrec->epc == epcBERENDSEN && do_per_step(step, inputrec->nstpcouple));
1679 const bool doCRescalePressureCoupling =
1680 (inputrec->epc == epcCRESCALE && do_per_step(step, inputrec->nstpcouple));
1682 && (doBerendsenPressureCoupling || doCRescalePressureCoupling || doParrinelloRahman))
1684 integrator->scaleCoordinates(pressureCouplingMu);
1685 if (doCRescalePressureCoupling)
1687 matrix pressureCouplingInvMu;
1688 gmx::invertBoxMatrix(pressureCouplingMu, pressureCouplingInvMu);
1689 integrator->scaleVelocities(pressureCouplingInvMu);
1691 integrator->setPbc(PbcType::Xyz, state->box);
1694 /* ################# END UPDATE STEP 2 ################# */
1695 /* #### We now have r(t+dt) and v(t+dt/2) ############# */
1697 /* The coordinates (x) were unshifted in update */
1700 /* We will not sum ekinh_old,
1701 * so signal that we still have to do it.
1703 bSumEkinhOld = TRUE;
1708 /* ######### BEGIN PREPARING EDR OUTPUT ########### */
1710 /* use the directly determined last velocity, not actually the averaged half steps */
1711 if (bTrotter && ir->eI == eiVV)
1713 enerd->term[F_EKIN] = last_ekin;
1715 enerd->term[F_ETOT] = enerd->term[F_EPOT] + enerd->term[F_EKIN];
1717 if (integratorHasConservedEnergyQuantity(ir))
1721 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + saved_conserved_quantity;
1725 enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + NPT_energy(ir, state, &MassQ);
1728 /* ######### END PREPARING EDR OUTPUT ########### */
1734 if (fplog && do_log && bDoExpanded)
1736 /* only needed if doing expanded ensemble */
1737 PrintFreeEnergyInfoToFile(fplog,
1740 ir->bSimTemp ? ir->simtempvals : nullptr,
1741 state_global->dfhist,
1748 energyOutput.addDataAtEnergyStep(bDoDHDL,
1756 PTCouplingArrays{ state->boxv,
1757 state->nosehoover_xi,
1758 state->nosehoover_vxi,
1760 state->nhpres_vxi },
1772 energyOutput.recordNonEnergyStep();
1775 gmx_bool do_dr = do_per_step(step, ir->nstdisreout);
1776 gmx_bool do_or = do_per_step(step, ir->nstorireout);
1778 if (doSimulatedAnnealing)
1780 gmx::EnergyOutput::printAnnealingTemperatures(
1781 do_log ? fplog : nullptr, groups, &(ir->opts));
1783 if (do_log || do_ene || do_dr || do_or)
1785 energyOutput.printStepToEnergyFile(mdoutf_get_fp_ene(outf),
1789 do_log ? fplog : nullptr,
1795 if (do_log && ir->bDoAwh && awh->hasFepLambdaDimension())
1797 const bool isInitialOutput = false;
1798 printLambdaStateToLog(fplog, state->lambda, isInitialOutput);
1803 pull_print_output(pull_work, step, t);
1806 if (do_per_step(step, ir->nstlog))
1808 if (fflush(fplog) != 0)
1810 gmx_fatal(FARGS, "Cannot flush logfile - maybe you are out of disk space?");
1816 /* Have to do this part _after_ outputting the logfile and the edr file */
1817 /* Gets written into the state at the beginning of next loop*/
1818 state->fep_state = lamnew;
1820 else if (ir->bDoAwh && awh->needForeignEnergyDifferences(step))
1822 state->fep_state = awh->fepLambdaState();
1824 /* Print the remaining wall clock time for the run */
1825 if (isMasterSimMasterRank(ms, MASTER(cr)) && (do_verbose || gmx_got_usr_signal()) && !bPMETunePrinting)
1829 fprintf(stderr, "\n");
1831 print_time(stderr, walltime_accounting, step, ir, cr);
1834 /* Ion/water position swapping.
1835 * Not done in last step since trajectory writing happens before this call
1836 * in the MD loop and exchanges would be lost anyway. */
1837 bNeedRepartition = FALSE;
1838 if ((ir->eSwapCoords != eswapNO) && (step > 0) && !bLastStep && do_per_step(step, ir->swap->nstswap))
1840 bNeedRepartition = do_swapcoords(cr,
1846 as_rvec_array(state->x.data()),
1848 MASTER(cr) && mdrunOptions.verbose,
1851 if (bNeedRepartition && DOMAINDECOMP(cr))
1853 dd_collect_state(cr->dd, state, state_global);
1857 /* Replica exchange */
1861 bExchanged = replica_exchange(fplog, cr, ms, repl_ex, state_global, enerd, state, step, t);
1864 if ((bExchanged || bNeedRepartition) && DOMAINDECOMP(cr))
1866 dd_partition_system(fplog,
1887 shouldCheckNumberOfBondedInteractions = true;
1888 upd.setNumAtoms(state->natoms);
1894 /* ####### SET VARIABLES FOR NEXT ITERATION IF THEY STILL NEED IT ###### */
1895 /* With all integrators, except VV, we need to retain the pressure
1896 * at the current step for coupling at the next step.
1898 if ((state->flags & (1U << estPRES_PREV))
1899 && (bGStatEveryStep || (ir->nstpcouple > 0 && step % ir->nstpcouple == 0)))
1901 /* Store the pressure in t_state for pressure coupling
1902 * at the next MD step.
1904 copy_mat(pres, state->pres_prev);
1907 /* ####### END SET VARIABLES FOR NEXT ITERATION ###### */
1909 if ((membed != nullptr) && (!bLastStep))
1911 rescale_membed(step_rel, membed, as_rvec_array(state_global->x.data()));
1914 cycles = wallcycle_stop(wcycle, ewcSTEP);
1915 if (DOMAINDECOMP(cr) && wcycle)
1917 dd_cycles_add(cr->dd, cycles, ddCyclStep);
1920 /* increase the MD step number */
1927 fcReportProgress(ir->nsteps + ir->init_step, step);
1931 resetHandler->resetCounters(
1932 step, step_rel, mdlog, fplog, cr, fr->nbv.get(), nrnb, fr->pmedata, pme_loadbal, wcycle, walltime_accounting);
1934 /* If bIMD is TRUE, the master updates the IMD energy record and sends positions to VMD client */
1935 imdSession->updateEnergyRecordAndSendPositionsAndEnergies(bInteractiveMDstep, step, bCalcEner);
1937 /* End of main MD loop */
1939 /* Closing TNG files can include compressing data. Therefore it is good to do that
1940 * before stopping the time measurements. */
1941 mdoutf_tng_close(outf);
1943 /* Stop measuring walltime */
1944 walltime_accounting_end_time(walltime_accounting);
1946 if (!thisRankHasDuty(cr, DUTY_PME))
1948 /* Tell the PME only node to finish */
1949 gmx_pme_send_finish(cr);
1954 if (ir->nstcalcenergy > 0)
1956 energyOutput.printEnergyConservation(fplog, ir->simulation_part, EI_MD(ir->eI));
1958 gmx::EnergyOutput::printAnnealingTemperatures(fplog, groups, &(ir->opts));
1959 energyOutput.printAverages(fplog, groups);
1966 pme_loadbal_done(pme_loadbal, fplog, mdlog, fr->nbv->useGpu());
1969 done_shellfc(fplog, shellfc, step_rel);
1971 if (useReplicaExchange && MASTER(cr))
1973 print_replica_exchange_statistics(fplog, repl_ex);
1976 walltime_accounting_set_nsteps_done(walltime_accounting, step_rel);
1978 global_stat_destroy(gstat);