src/gromacs/mdrun/md.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * 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.
  10  *
  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.
  15  *
  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.
  20  *
  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.
  25  *
  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.
  33  *
  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.
  36  */
  37 /*! \internal \file
  38  *
  39  * \brief Implements the integrator for normal molecular dynamics simulations
  40  *
  41  * \author David van der Spoel <david.vanderspoel@icm.uu.se>
  42  * \ingroup module_mdrun
  43  */
  44 #include "gmxpre.h"
  45
  46 #include <cinttypes>
  47 #include <cmath>
  48 #include <cstdio>
  49 #include <cstdlib>
  50
  51 #include <algorithm>
  52 #include <memory>
  53 #include <numeric>
  54
  55 #include "gromacs/applied_forces/awh/awh.h"
  56 #include "gromacs/applied_forces/awh/read_params.h"
  57 #include "gromacs/commandline/filenm.h"
  58 #include "gromacs/domdec/collect.h"
  59 #include "gromacs/domdec/dlbtiming.h"
  60 #include "gromacs/domdec/domdec.h"
  61 #include "gromacs/domdec/domdec_network.h"
  62 #include "gromacs/domdec/domdec_struct.h"
  63 #include "gromacs/domdec/gpuhaloexchange.h"
  64 #include "gromacs/domdec/mdsetup.h"
  65 #include "gromacs/domdec/partition.h"
  66 #include "gromacs/essentialdynamics/edsam.h"
  67 #include "gromacs/ewald/pme_load_balancing.h"
  68 #include "gromacs/ewald/pme_pp.h"
  69 #include "gromacs/fileio/trxio.h"
  70 #include "gromacs/gmxlib/network.h"
  71 #include "gromacs/gmxlib/nrnb.h"
  72 #include "gromacs/gpu_utils/device_stream_manager.h"
  73 #include "gromacs/gpu_utils/gpu_utils.h"
  74 #include "gromacs/imd/imd.h"
  75 #include "gromacs/listed_forces/listed_forces.h"
  76 #include "gromacs/math/functions.h"
  77 #include "gromacs/math/invertmatrix.h"
  78 #include "gromacs/math/vec.h"
  79 #include "gromacs/math/vectypes.h"
  80 #include "gromacs/mdlib/checkpointhandler.h"
  81 #include "gromacs/mdlib/compute_io.h"
  82 #include "gromacs/mdlib/constr.h"
  83 #include "gromacs/mdlib/coupling.h"
  84 #include "gromacs/mdlib/ebin.h"
  85 #include "gromacs/mdlib/enerdata_utils.h"
  86 #include "gromacs/mdlib/energyoutput.h"
  87 #include "gromacs/mdlib/expanded.h"
  88 #include "gromacs/mdlib/force.h"
  89 #include "gromacs/mdlib/force_flags.h"
  90 #include "gromacs/mdlib/forcerec.h"
  91 #include "gromacs/mdlib/freeenergyparameters.h"
  92 #include "gromacs/mdlib/md_support.h"
  93 #include "gromacs/mdlib/mdatoms.h"
  94 #include "gromacs/mdlib/mdoutf.h"
  95 #include "gromacs/mdlib/membed.h"
  96 #include "gromacs/mdlib/resethandler.h"
  97 #include "gromacs/mdlib/sighandler.h"
  98 #include "gromacs/mdlib/simulationsignal.h"
  99 #include "gromacs/mdlib/stat.h"
 100 #include "gromacs/mdlib/stophandler.h"
 101 #include "gromacs/mdlib/tgroup.h"
 102 #include "gromacs/mdlib/trajectory_writing.h"
 103 #include "gromacs/mdlib/update.h"
 104 #include "gromacs/mdlib/update_constrain_gpu.h"
 105 #include "gromacs/mdlib/update_vv.h"
 106 #include "gromacs/mdlib/vcm.h"
 107 #include "gromacs/mdlib/vsite.h"
 108 #include "gromacs/mdrunutility/handlerestart.h"
 109 #include "gromacs/mdrunutility/multisim.h"
 110 #include "gromacs/mdrunutility/printtime.h"
 111 #include "gromacs/mdtypes/awh_history.h"
 112 #include "gromacs/mdtypes/awh_params.h"
 113 #include "gromacs/mdtypes/commrec.h"
 114 #include "gromacs/mdtypes/df_history.h"
 115 #include "gromacs/mdtypes/energyhistory.h"
 116 #include "gromacs/mdtypes/fcdata.h"
 117 #include "gromacs/mdtypes/forcebuffers.h"
 118 #include "gromacs/mdtypes/forcerec.h"
 119 #include "gromacs/mdtypes/group.h"
 120 #include "gromacs/mdtypes/inputrec.h"
 121 #include "gromacs/mdtypes/interaction_const.h"
 122 #include "gromacs/mdtypes/md_enums.h"
 123 #include "gromacs/mdtypes/mdatom.h"
 124 #include "gromacs/mdtypes/mdrunoptions.h"
 125 #include "gromacs/mdtypes/multipletimestepping.h"
 126 #include "gromacs/mdtypes/observableshistory.h"
 127 #include "gromacs/mdtypes/pullhistory.h"
 128 #include "gromacs/mdtypes/simulation_workload.h"
 129 #include "gromacs/mdtypes/state.h"
 130 #include "gromacs/mdtypes/state_propagator_data_gpu.h"
 131 #include "gromacs/modularsimulator/energydata.h"
 132 #include "gromacs/nbnxm/gpu_data_mgmt.h"
 133 #include "gromacs/nbnxm/nbnxm.h"
 134 #include "gromacs/pbcutil/pbc.h"
 135 #include "gromacs/pulling/output.h"
 136 #include "gromacs/pulling/pull.h"
 137 #include "gromacs/swap/swapcoords.h"
 138 #include "gromacs/timing/wallcycle.h"
 139 #include "gromacs/timing/walltime_accounting.h"
 140 #include "gromacs/topology/atoms.h"
 141 #include "gromacs/topology/idef.h"
 142 #include "gromacs/topology/mtop_util.h"
 143 #include "gromacs/topology/topology.h"
 144 #include "gromacs/trajectory/trajectoryframe.h"
 145 #include "gromacs/utility/basedefinitions.h"
 146 #include "gromacs/utility/cstringutil.h"
 147 #include "gromacs/utility/fatalerror.h"
 148 #include "gromacs/utility/logger.h"
 149 #include "gromacs/utility/real.h"
 150 #include "gromacs/utility/smalloc.h"
 151
 152 #include "legacysimulator.h"
 153 #include "replicaexchange.h"
 154 #include "shellfc.h"
 155
 156 using gmx::SimulationSignaller;
 157
 158 void gmx::LegacySimulator::do_md()
 159 {
 160     // TODO Historically, the EM and MD "integrators" used different
 161     // names for the t_inputrec *parameter, but these must have the
 162     // same name, now that it's a member of a struct. We use this ir
 163     // alias to avoid a large ripple of nearly useless changes.
 164     // t_inputrec is being replaced by IMdpOptionsProvider, so this
 165     // will go away eventually.
 166     const t_inputrec* ir = inputrec;
 167
 168     int64_t      step, step_rel;
 169     double       t, t0 = ir->init_t;
 170     gmx_bool     bGStatEveryStep, bGStat, bCalcVir, bCalcEnerStep, bCalcEner;
 171     gmx_bool     bNS = FALSE, bNStList, bStopCM, bFirstStep, bInitStep, bLastStep = FALSE;
 172     gmx_bool     bDoDHDL = FALSE, bDoFEP = FALSE, bDoExpanded = FALSE;
 173     gmx_bool     do_ene, do_log, do_verbose;
 174     gmx_bool     bMasterState;
 175     unsigned int force_flags;
 176     tensor force_vir = { { 0 } }, shake_vir = { { 0 } }, total_vir = { { 0 } }, pres = { { 0 } };
 177     int    i, m;
 178     rvec   mu_tot;
 179     matrix pressureCouplingMu, M;
 180     gmx_repl_ex_t     repl_ex = nullptr;
 181     gmx_global_stat_t gstat;
 182     gmx_shellfc_t*    shellfc;
 183     gmx_bool          bSumEkinhOld, bDoReplEx, bExchanged, bNeedRepartition;
 184     gmx_bool          bTrotter;
 185     real              dvdl_constr;
 186     std::vector<RVec> cbuf;
 187     matrix            lastbox;
 188     int               lamnew = 0;
 189     /* for FEP */
 190     int       nstfep = 0;
 191     double    cycles;
 192     real      saved_conserved_quantity = 0;
 193     real      last_ekin                = 0;
 194     t_extmass MassQ;
 195     char      sbuf[STEPSTRSIZE], sbuf2[STEPSTRSIZE];
 196
 197     /* PME load balancing data for GPU kernels */
 198     gmx_bool bPMETune         = FALSE;
 199     gmx_bool bPMETunePrinting = FALSE;
 200
 201     bool bInteractiveMDstep = false;
 202
 203     SimulationSignals signals;
 204     // Most global communnication stages don't propagate mdrun
 205     // signals, and will use this object to achieve that.
 206     SimulationSignaller nullSignaller(nullptr, nullptr, nullptr, false, false);
 207
 208     if (!mdrunOptions.writeConfout)
 209     {
 210         // This is on by default, and the main known use case for
 211         // turning it off is for convenience in benchmarking, which is
 212         // something that should not show up in the general user
 213         // interface.
 214         GMX_LOG(mdlog.info)
 215                 .asParagraph()
 216                 .appendText(
 217                         "The -noconfout functionality is deprecated, and may be removed in a "
 218                         "future version.");
 219     }
 220
 221     /* md-vv uses averaged full step velocities for T-control
 222        md-vv-avek uses averaged half step velocities for T-control (but full step ekin for P control)
 223        md uses averaged half step kinetic energies to determine temperature unless defined otherwise by GMX_EKIN_AVE_VEL; */
 224     bTrotter = (EI_VV(ir->eI)
 225                 && (inputrecNptTrotter(ir) || inputrecNphTrotter(ir) || inputrecNvtTrotter(ir)));
 226
 227     const bool bRerunMD = false;
 228
 229     int nstglobalcomm = computeGlobalCommunicationPeriod(mdlog, ir, cr);
 230     bGStatEveryStep   = (nstglobalcomm == 1);
 231
 232     const SimulationGroups* groups = &top_global.groups;
 233
 234     std::unique_ptr<EssentialDynamics> ed = nullptr;
 235     if (opt2bSet("-ei", nfile, fnm))
 236     {
 237         /* Initialize essential dynamics sampling */
 238         ed = init_edsam(mdlog,
 239                         opt2fn_null("-ei", nfile, fnm),
 240                         opt2fn("-eo", nfile, fnm),
 241                         top_global,
 242                         *ir,
 243                         cr,
 244                         constr,
 245                         state_global,
 246                         observablesHistory,
 247                         oenv,
 248                         startingBehavior);
 249     }
 250     else if (observablesHistory->edsamHistory)
 251     {
 252         gmx_fatal(FARGS,
 253                   "The checkpoint is from a run with essential dynamics sampling, "
 254                   "but the current run did not specify the -ei option. "
 255                   "Either specify the -ei option to mdrun, or do not use this checkpoint file.");
 256     }
 257
 258     int*                fep_state = MASTER(cr) ? &state_global->fep_state : nullptr;
 259     gmx::ArrayRef<real> lambda    = MASTER(cr) ? state_global->lambda : gmx::ArrayRef<real>();
 260     initialize_lambdas(fplog,
 261                        ir->efep,
 262                        ir->bSimTemp,
 263                        *ir->fepvals,
 264                        ir->simtempvals->temperatures,
 265                        gmx::arrayRefFromArray(ir->opts.ref_t, ir->opts.ngtc),
 266                        MASTER(cr),
 267                        fep_state,
 268                        lambda);
 269     Update upd(*ir, deform);
 270     bool   doSimulatedAnnealing = false;
 271     {
 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);
 276     }
 277     const bool useReplicaExchange = (replExParams.exchangeInterval > 0);
 278
 279     const t_fcdata& fcdata = *fr->fcdata;
 280
 281     bool simulationsShareState = false;
 282     int  nstSignalComm         = nstglobalcomm;
 283     {
 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 = (fcdata.disres->nsystems > 1) || ((ms != nullptr) && fcdata.orires);
 287         bool awhUsesMultiSim = (ir->bDoAwh && ir->awhParams->shareBiasMultisim() && (ms != nullptr));
 288
 289         // Replica exchange, ensemble restraints and AWH need all
 290         // simulations to remain synchronized, so they need
 291         // checkpoints and stop conditions to act on the same step, so
 292         // the propagation of such signals must take place between
 293         // simulations, not just within simulations.
 294         // TODO: Make algorithm initializers set these flags.
 295         simulationsShareState = useReplicaExchange || usingEnsembleRestraints || awhUsesMultiSim;
 296
 297         if (simulationsShareState)
 298         {
 299             // Inter-simulation signal communication does not need to happen
 300             // often, so we use a minimum of 200 steps to reduce overhead.
 301             const int c_minimumInterSimulationSignallingInterval = 200;
 302             nstSignalComm = ((c_minimumInterSimulationSignallingInterval + nstglobalcomm - 1) / nstglobalcomm)
 303                             * nstglobalcomm;
 304         }
 305     }
 306
 307     if (startingBehavior != StartingBehavior::RestartWithAppending)
 308     {
 309         pleaseCiteCouplingAlgorithms(fplog, *ir);
 310     }
 311     gmx_mdoutf*       outf = init_mdoutf(fplog,
 312                                    nfile,
 313                                    fnm,
 314                                    mdrunOptions,
 315                                    cr,
 316                                    outputProvider,
 317                                    mdModulesNotifiers,
 318                                    ir,
 319                                    top_global,
 320                                    oenv,
 321                                    wcycle,
 322                                    startingBehavior,
 323                                    simulationsShareState,
 324                                    ms);
 325     gmx::EnergyOutput energyOutput(mdoutf_get_fp_ene(outf),
 326                                    top_global,
 327                                    *ir,
 328                                    pull_work,
 329                                    mdoutf_get_fp_dhdl(outf),
 330                                    false,
 331                                    startingBehavior,
 332                                    simulationsShareState,
 333                                    mdModulesNotifiers);
 334
 335     gstat = global_stat_init(ir);
 336
 337     const auto& simulationWork     = runScheduleWork->simulationWork;
 338     const bool  useGpuForPme       = simulationWork.useGpuPme;
 339     const bool  useGpuForNonbonded = simulationWork.useGpuNonbonded;
 340     const bool  useGpuForBufferOps = simulationWork.useGpuBufferOps;
 341     const bool  useGpuForUpdate    = simulationWork.useGpuUpdate;
 342
 343     /* Check for polarizable models and flexible constraints */
 344     shellfc = init_shell_flexcon(fplog,
 345                                  top_global,
 346                                  constr ? constr->numFlexibleConstraints() : 0,
 347                                  ir->nstcalcenergy,
 348                                  DOMAINDECOMP(cr),
 349                                  useGpuForPme);
 350
 351     {
 352         double io = compute_io(ir, top_global.natoms, *groups, energyOutput.numEnergyTerms(), 1);
 353         if ((io > 2000) && MASTER(cr))
 354         {
 355             fprintf(stderr, "\nWARNING: This run will generate roughly %.0f Mb of data\n\n", io);
 356         }
 357     }
 358
 359     // Local state only becomes valid now.
 360     std::unique_ptr<t_state> stateInstance;
 361     t_state*                 state;
 362
 363     gmx_localtop_t top(top_global.ffparams);
 364
 365     ForceBuffers     f(fr->useMts,
 366                    ((useGpuForNonbonded && useGpuForBufferOps) || useGpuForUpdate)
 367                                ? PinningPolicy::PinnedIfSupported
 368                                : PinningPolicy::CannotBePinned);
 369     const t_mdatoms* md = mdAtoms->mdatoms();
 370     if (DOMAINDECOMP(cr))
 371     {
 372         stateInstance = std::make_unique<t_state>();
 373         state         = stateInstance.get();
 374         dd_init_local_state(*cr->dd, state_global, state);
 375
 376         /* Distribute the charge groups over the nodes from the master node */
 377         dd_partition_system(fplog,
 378                             mdlog,
 379                             ir->init_step,
 380                             cr,
 381                             TRUE,
 382                             1,
 383                             state_global,
 384                             top_global,
 385                             *ir,
 386                             imdSession,
 387                             pull_work,
 388                             state,
 389                             &f,
 390                             mdAtoms,
 391                             &top,
 392                             fr,
 393                             vsite,
 394                             constr,
 395                             nrnb,
 396                             nullptr,
 397                             FALSE);
 398         upd.updateAfterPartition(state->natoms,
 399                                  md->cFREEZE ? gmx::arrayRefFromArray(md->cFREEZE, md->nr)
 400                                              : gmx::ArrayRef<const unsigned short>(),
 401                                  md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
 402                                          : gmx::ArrayRef<const unsigned short>());
 403     }
 404     else
 405     {
 406         state_change_natoms(state_global, state_global->natoms);
 407         /* Copy the pointer to the global state */
 408         state = state_global;
 409
 410         /* Generate and initialize new topology */
 411         mdAlgorithmsSetupAtomData(cr, *ir, top_global, &top, fr, &f, mdAtoms, constr, vsite, shellfc);
 412
 413         upd.updateAfterPartition(state->natoms,
 414                                  md->cFREEZE ? gmx::arrayRefFromArray(md->cFREEZE, md->nr)
 415                                              : gmx::ArrayRef<const unsigned short>(),
 416                                  md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
 417                                          : gmx::ArrayRef<const unsigned short>());
 418     }
 419
 420     std::unique_ptr<UpdateConstrainGpu> integrator;
 421
 422     StatePropagatorDataGpu* stateGpu = fr->stateGpu;
 423
 424     // TODO: the assertions below should be handled by UpdateConstraintsBuilder.
 425     if (useGpuForUpdate)
 426     {
 427         GMX_RELEASE_ASSERT(!DOMAINDECOMP(cr) || ddUsesUpdateGroups(*cr->dd) || constr == nullptr
 428                                    || constr->numConstraintsTotal() == 0,
 429                            "Constraints in domain decomposition are only supported with update "
 430                            "groups if using GPU update.\n");
 431         GMX_RELEASE_ASSERT(ir->eConstrAlg != ConstraintAlgorithm::Shake || constr == nullptr
 432                                    || constr->numConstraintsTotal() == 0,
 433                            "SHAKE is not supported with GPU update.");
 434         GMX_RELEASE_ASSERT(useGpuForPme || (useGpuForNonbonded && simulationWork.useGpuBufferOps),
 435                            "Either PME or short-ranged non-bonded interaction tasks must run on "
 436                            "the GPU to use GPU update.\n");
 437         GMX_RELEASE_ASSERT(ir->eI == IntegrationAlgorithm::MD,
 438                            "Only the md integrator is supported with the GPU update.\n");
 439         GMX_RELEASE_ASSERT(
 440                 ir->etc != TemperatureCoupling::NoseHoover,
 441                 "Nose-Hoover temperature coupling is not supported with the GPU update.\n");
 442         GMX_RELEASE_ASSERT(
 443                 ir->epc == PressureCoupling::No || ir->epc == PressureCoupling::ParrinelloRahman
 444                         || ir->epc == PressureCoupling::Berendsen || ir->epc == PressureCoupling::CRescale,
 445                 "Only Parrinello-Rahman, Berendsen, and C-rescale pressure coupling are supported "
 446                 "with the GPU update.\n");
 447         GMX_RELEASE_ASSERT(!md->haveVsites,
 448                            "Virtual sites are not supported with the GPU update.\n");
 449         GMX_RELEASE_ASSERT(ed == nullptr,
 450                            "Essential dynamics is not supported with the GPU update.\n");
 451         GMX_RELEASE_ASSERT(!ir->bPull || !pull_have_constraint(*ir->pull),
 452                            "Constraints pulling is not supported with the GPU update.\n");
 453         GMX_RELEASE_ASSERT(fcdata.orires == nullptr,
 454                            "Orientation restraints are not supported with the GPU update.\n");
 455         GMX_RELEASE_ASSERT(
 456                 ir->efep == FreeEnergyPerturbationType::No
 457                         || (!haveFepPerturbedMasses(top_global) && !havePerturbedConstraints(top_global)),
 458                 "Free energy perturbation of masses and constraints are not supported with the GPU "
 459                 "update.");
 460
 461         if (constr != nullptr && constr->numConstraintsTotal() > 0)
 462         {
 463             GMX_LOG(mdlog.info)
 464                     .asParagraph()
 465                     .appendText("Updating coordinates and applying constraints on the GPU.");
 466         }
 467         else
 468         {
 469             GMX_LOG(mdlog.info).asParagraph().appendText("Updating coordinates on the GPU.");
 470         }
 471         GMX_RELEASE_ASSERT(fr->deviceStreamManager != nullptr,
 472                            "Device stream manager should be initialized in order to use GPU "
 473                            "update-constraints.");
 474         GMX_RELEASE_ASSERT(
 475                 fr->deviceStreamManager->streamIsValid(gmx::DeviceStreamType::UpdateAndConstraints),
 476                 "Update stream should be initialized in order to use GPU "
 477                 "update-constraints.");
 478         integrator = std::make_unique<UpdateConstrainGpu>(
 479                 *ir,
 480                 top_global,
 481                 ekind->ngtc,
 482                 fr->deviceStreamManager->context(),
 483                 fr->deviceStreamManager->stream(gmx::DeviceStreamType::UpdateAndConstraints),
 484                 stateGpu->xUpdatedOnDevice(),
 485                 wcycle);
 486
 487         integrator->setPbc(PbcType::Xyz, state->box);
 488     }
 489
 490     if (useGpuForPme || (useGpuForNonbonded && useGpuForBufferOps) || useGpuForUpdate)
 491     {
 492         changePinningPolicy(&state->x, PinningPolicy::PinnedIfSupported);
 493     }
 494     if (useGpuForUpdate)
 495     {
 496         changePinningPolicy(&state->v, PinningPolicy::PinnedIfSupported);
 497     }
 498
 499     // NOTE: The global state is no longer used at this point.
 500     // But state_global is still used as temporary storage space for writing
 501     // the global state to file and potentially for replica exchange.
 502     // (Global topology should persist.)
 503
 504     update_mdatoms(mdAtoms->mdatoms(), state->lambda[FreeEnergyPerturbationCouplingType::Mass]);
 505
 506     if (ir->bExpanded)
 507     {
 508         /* Check nstexpanded here, because the grompp check was broken */
 509         if (ir->expandedvals->nstexpanded % ir->nstcalcenergy != 0)
 510         {
 511             gmx_fatal(FARGS,
 512                       "With expanded ensemble, nstexpanded should be a multiple of nstcalcenergy");
 513         }
 514         init_expanded_ensemble(startingBehavior != StartingBehavior::NewSimulation, ir, state->dfhist);
 515     }
 516
 517     if (MASTER(cr))
 518     {
 519         EnergyData::initializeEnergyHistory(startingBehavior, observablesHistory, &energyOutput);
 520     }
 521
 522     preparePrevStepPullCom(ir,
 523                            pull_work,
 524                            gmx::arrayRefFromArray(md->massT, md->nr),
 525                            state,
 526                            state_global,
 527                            cr,
 528                            startingBehavior != StartingBehavior::NewSimulation);
 529
 530     // TODO: Remove this by converting AWH into a ForceProvider
 531     auto awh = prepareAwhModule(fplog,
 532                                 *ir,
 533                                 state_global,
 534                                 cr,
 535                                 ms,
 536                                 startingBehavior != StartingBehavior::NewSimulation,
 537                                 shellfc != nullptr,
 538                                 opt2fn("-awh", nfile, fnm),
 539                                 pull_work);
 540
 541     if (useReplicaExchange && MASTER(cr))
 542     {
 543         repl_ex = init_replica_exchange(fplog, ms, top_global.natoms, ir, replExParams);
 544     }
 545     /* PME tuning is only supported in the Verlet scheme, with PME for
 546      * Coulomb. It is not supported with only LJ PME. */
 547     bPMETune = (mdrunOptions.tunePme && EEL_PME(fr->ic->eeltype) && !mdrunOptions.reproducible
 548                 && ir->cutoff_scheme != CutoffScheme::Group);
 549
 550     pme_load_balancing_t* pme_loadbal = nullptr;
 551     if (bPMETune)
 552     {
 553         pme_loadbal_init(
 554                 &pme_loadbal, cr, mdlog, *ir, state->box, *fr->ic, *fr->nbv, fr->pmedata, fr->nbv->useGpu());
 555     }
 556
 557     if (!ir->bContinuation)
 558     {
 559         if (state->flags & enumValueToBitMask(StateEntry::V))
 560         {
 561             auto v = makeArrayRef(state->v);
 562             /* Set the velocities of vsites, shells and frozen atoms to zero */
 563             for (i = 0; i < md->homenr; i++)
 564             {
 565                 if (md->ptype[i] == ParticleType::Shell)
 566                 {
 567                     clear_rvec(v[i]);
 568                 }
 569                 else if (md->cFREEZE)
 570                 {
 571                     for (m = 0; m < DIM; m++)
 572                     {
 573                         if (ir->opts.nFreeze[md->cFREEZE[i]][m])
 574                         {
 575                             v[i][m] = 0;
 576                         }
 577                     }
 578                 }
 579             }
 580         }
 581
 582         if (constr)
 583         {
 584             /* Constrain the initial coordinates and velocities */
 585             do_constrain_first(fplog,
 586                                constr,
 587                                ir,
 588                                md->nr,
 589                                md->homenr,
 590                                state->x.arrayRefWithPadding(),
 591                                state->v.arrayRefWithPadding(),
 592                                state->box,
 593                                state->lambda[FreeEnergyPerturbationCouplingType::Bonded]);
 594         }
 595     }
 596
 597     if (ir->efep != FreeEnergyPerturbationType::No)
 598     {
 599         /* Set free energy calculation frequency as the greatest common
 600          * denominator of nstdhdl and repl_ex_nst. */
 601         nstfep = ir->fepvals->nstdhdl;
 602         if (ir->bExpanded)
 603         {
 604             nstfep = std::gcd(ir->expandedvals->nstexpanded, nstfep);
 605         }
 606         if (useReplicaExchange)
 607         {
 608             nstfep = std::gcd(replExParams.exchangeInterval, nstfep);
 609         }
 610         if (ir->bDoAwh)
 611         {
 612             nstfep = std::gcd(ir->awhParams->nstSampleCoord(), nstfep);
 613         }
 614     }
 615
 616     /* Be REALLY careful about what flags you set here. You CANNOT assume
 617      * this is the first step, since we might be restarting from a checkpoint,
 618      * and in that case we should not do any modifications to the state.
 619      */
 620     bStopCM = (ir->comm_mode != ComRemovalAlgorithm::No && !ir->bContinuation);
 621
 622     // When restarting from a checkpoint, it can be appropriate to
 623     // initialize ekind from quantities in the checkpoint. Otherwise,
 624     // compute_globals must initialize ekind before the simulation
 625     // starts/restarts. However, only the master rank knows what was
 626     // found in the checkpoint file, so we have to communicate in
 627     // order to coordinate the restart.
 628     //
 629     // TODO Consider removing this communication if/when checkpoint
 630     // reading directly follows .tpr reading, because all ranks can
 631     // agree on hasReadEkinState at that time.
 632     bool hasReadEkinState = MASTER(cr) ? state_global->ekinstate.hasReadEkinState : false;
 633     if (PAR(cr))
 634     {
 635         gmx_bcast(sizeof(hasReadEkinState), &hasReadEkinState, cr->mpi_comm_mygroup);
 636     }
 637     if (hasReadEkinState)
 638     {
 639         restore_ekinstate_from_state(cr, ekind, &state_global->ekinstate);
 640     }
 641
 642     unsigned int cglo_flags =
 643             (CGLO_TEMPERATURE | CGLO_GSTAT | (EI_VV(ir->eI) ? CGLO_PRESSURE : 0)
 644              | (EI_VV(ir->eI) ? CGLO_CONSTRAINT : 0) | (hasReadEkinState ? CGLO_READEKIN : 0));
 645
 646     bSumEkinhOld = FALSE;
 647
 648     t_vcm vcm(top_global.groups, *ir);
 649     reportComRemovalInfo(fplog, vcm);
 650
 651     /* To minimize communication, compute_globals computes the COM velocity
 652      * and the kinetic energy for the velocities without COM motion removed.
 653      * Thus to get the kinetic energy without the COM contribution, we need
 654      * to call compute_globals twice.
 655      */
 656     for (int cgloIteration = 0; cgloIteration < (bStopCM ? 2 : 1); cgloIteration++)
 657     {
 658         unsigned int cglo_flags_iteration = cglo_flags;
 659         if (bStopCM && cgloIteration == 0)
 660         {
 661             cglo_flags_iteration |= CGLO_STOPCM;
 662             cglo_flags_iteration &= ~CGLO_TEMPERATURE;
 663         }
 664         if (DOMAINDECOMP(cr) && shouldCheckNumberOfBondedInteractions(*cr->dd) && cgloIteration == 0)
 665         {
 666             cglo_flags_iteration |= CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS;
 667         }
 668         compute_globals(gstat,
 669                         cr,
 670                         ir,
 671                         fr,
 672                         ekind,
 673                         makeConstArrayRef(state->x),
 674                         makeConstArrayRef(state->v),
 675                         state->box,
 676                         md,
 677                         nrnb,
 678                         &vcm,
 679                         nullptr,
 680                         enerd,
 681                         force_vir,
 682                         shake_vir,
 683                         total_vir,
 684                         pres,
 685                         gmx::ArrayRef<real>{},
 686                         &nullSignaller,
 687                         state->box,
 688                         &bSumEkinhOld,
 689                         cglo_flags_iteration);
 690         if (cglo_flags_iteration & CGLO_STOPCM)
 691         {
 692             /* At initialization, do not pass x with acceleration-correction mode
 693              * to avoid (incorrect) correction of the initial coordinates.
 694              */
 695             auto x = (vcm.mode == ComRemovalAlgorithm::LinearAccelerationCorrection)
 696                              ? ArrayRef<RVec>()
 697                              : makeArrayRef(state->x);
 698             process_and_stopcm_grp(fplog, &vcm, *md, x, makeArrayRef(state->v));
 699             inc_nrnb(nrnb, eNR_STOPCM, md->homenr);
 700         }
 701     }
 702     if (DOMAINDECOMP(cr))
 703     {
 704         checkNumberOfBondedInteractions(
 705                 mdlog, cr, top_global, &top, makeConstArrayRef(state->x), state->box);
 706     }
 707     if (ir->eI == IntegrationAlgorithm::VVAK)
 708     {
 709         /* a second call to get the half step temperature initialized as well */
 710         /* we do the same call as above, but turn the pressure off -- internally to
 711            compute_globals, this is recognized as a velocity verlet half-step
 712            kinetic energy calculation.  This minimized excess variables, but
 713            perhaps loses some logic?*/
 714
 715         compute_globals(gstat,
 716                         cr,
 717                         ir,
 718                         fr,
 719                         ekind,
 720                         makeConstArrayRef(state->x),
 721                         makeConstArrayRef(state->v),
 722                         state->box,
 723                         md,
 724                         nrnb,
 725                         &vcm,
 726                         nullptr,
 727                         enerd,
 728                         force_vir,
 729                         shake_vir,
 730                         total_vir,
 731                         pres,
 732                         gmx::ArrayRef<real>{},
 733                         &nullSignaller,
 734                         state->box,
 735                         &bSumEkinhOld,
 736                         cglo_flags & ~CGLO_PRESSURE);
 737     }
 738
 739     /* Calculate the initial half step temperature, and save the ekinh_old */
 740     if (startingBehavior == StartingBehavior::NewSimulation)
 741     {
 742         for (i = 0; (i < ir->opts.ngtc); i++)
 743         {
 744             copy_mat(ekind->tcstat[i].ekinh, ekind->tcstat[i].ekinh_old);
 745         }
 746     }
 747
 748     /* need to make an initiation call to get the Trotter variables set, as well as other constants
 749        for non-trotter temperature control */
 750     auto trotter_seq = init_npt_vars(ir, state, &MassQ, bTrotter);
 751
 752     if (MASTER(cr))
 753     {
 754         if (!ir->bContinuation)
 755         {
 756             if (constr && ir->eConstrAlg == ConstraintAlgorithm::Lincs)
 757             {
 758                 fprintf(fplog,
 759                         "RMS relative constraint deviation after constraining: %.2e\n",
 760                         constr->rmsd());
 761             }
 762             if (EI_STATE_VELOCITY(ir->eI))
 763             {
 764                 real temp = enerd->term[F_TEMP];
 765                 if (ir->eI != IntegrationAlgorithm::VV)
 766                 {
 767                     /* Result of Ekin averaged over velocities of -half
 768                      * and +half step, while we only have -half step here.
 769                      */
 770                     temp *= 2;
 771                 }
 772                 fprintf(fplog, "Initial temperature: %g K\n", temp);
 773             }
 774         }
 775
 776         char tbuf[20];
 777         fprintf(stderr, "starting mdrun '%s'\n", *(top_global.name));
 778         if (ir->nsteps >= 0)
 779         {
 780             sprintf(tbuf, "%8.1f", (ir->init_step + ir->nsteps) * ir->delta_t);
 781         }
 782         else
 783         {
 784             sprintf(tbuf, "%s", "infinite");
 785         }
 786         if (ir->init_step > 0)
 787         {
 788             fprintf(stderr,
 789                     "%s steps, %s ps (continuing from step %s, %8.1f ps).\n",
 790                     gmx_step_str(ir->init_step + ir->nsteps, sbuf),
 791                     tbuf,
 792                     gmx_step_str(ir->init_step, sbuf2),
 793                     ir->init_step * ir->delta_t);
 794         }
 795         else
 796         {
 797             fprintf(stderr, "%s steps, %s ps.\n", gmx_step_str(ir->nsteps, sbuf), tbuf);
 798         }
 799         fprintf(fplog, "\n");
 800     }
 801
 802     walltime_accounting_start_time(walltime_accounting);
 803     wallcycle_start(wcycle, WallCycleCounter::Run);
 804     print_start(fplog, cr, walltime_accounting, "mdrun");
 805
 806     /***********************************************************
 807      *
 808      *             Loop over MD steps
 809      *
 810      ************************************************************/
 811
 812     bFirstStep = TRUE;
 813     /* Skip the first Nose-Hoover integration when we get the state from tpx */
 814     bInitStep        = startingBehavior == StartingBehavior::NewSimulation || EI_VV(ir->eI);
 815     bSumEkinhOld     = FALSE;
 816     bExchanged       = FALSE;
 817     bNeedRepartition = FALSE;
 818
 819     step     = ir->init_step;
 820     step_rel = 0;
 821
 822     auto stopHandler = stopHandlerBuilder->getStopHandlerMD(
 823             compat::not_null<SimulationSignal*>(&signals[eglsSTOPCOND]),
 824             simulationsShareState,
 825             MASTER(cr),
 826             ir->nstlist,
 827             mdrunOptions.reproducible,
 828             nstSignalComm,
 829             mdrunOptions.maximumHoursToRun,
 830             ir->nstlist == 0,
 831             fplog,
 832             step,
 833             bNS,
 834             walltime_accounting);
 835
 836     auto checkpointHandler = std::make_unique<CheckpointHandler>(
 837             compat::make_not_null<SimulationSignal*>(&signals[eglsCHKPT]),
 838             simulationsShareState,
 839             ir->nstlist == 0,
 840             MASTER(cr),
 841             mdrunOptions.writeConfout,
 842             mdrunOptions.checkpointOptions.period);
 843
 844     const bool resetCountersIsLocal = true;
 845     auto       resetHandler         = std::make_unique<ResetHandler>(
 846             compat::make_not_null<SimulationSignal*>(&signals[eglsRESETCOUNTERS]),
 847             !resetCountersIsLocal,
 848             ir->nsteps,
 849             MASTER(cr),
 850             mdrunOptions.timingOptions.resetHalfway,
 851             mdrunOptions.maximumHoursToRun,
 852             mdlog,
 853             wcycle,
 854             walltime_accounting);
 855
 856     const DDBalanceRegionHandler ddBalanceRegionHandler(cr);
 857
 858     if (MASTER(cr) && isMultiSim(ms) && !useReplicaExchange)
 859     {
 860         logInitialMultisimStatus(ms, cr, mdlog, simulationsShareState, ir->nsteps, ir->init_step);
 861     }
 862
 863     /* and stop now if we should */
 864     bLastStep = (bLastStep || (ir->nsteps >= 0 && step_rel > ir->nsteps));
 865     while (!bLastStep)
 866     {
 867
 868         /* Determine if this is a neighbor search step */
 869         bNStList = (ir->nstlist > 0 && step % ir->nstlist == 0);
 870
 871         if (bPMETune && bNStList)
 872         {
 873             // This has to be here because PME load balancing is called so early.
 874             // TODO: Move to after all booleans are defined.
 875             if (useGpuForUpdate && !bFirstStep)
 876             {
 877                 stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
 878                 stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
 879             }
 880             /* PME grid + cut-off optimization with GPUs or PME nodes */
 881             pme_loadbal_do(pme_loadbal,
 882                            cr,
 883                            (mdrunOptions.verbose && MASTER(cr)) ? stderr : nullptr,
 884                            fplog,
 885                            mdlog,
 886                            *ir,
 887                            fr,
 888                            state->box,
 889                            state->x,
 890                            wcycle,
 891                            step,
 892                            step_rel,
 893                            &bPMETunePrinting,
 894                            simulationWork.useGpuPmePpCommunication);
 895         }
 896
 897         wallcycle_start(wcycle, WallCycleCounter::Step);
 898
 899         bLastStep = (step_rel == ir->nsteps);
 900         t         = t0 + step * ir->delta_t;
 901
 902         // TODO Refactor this, so that nstfep does not need a default value of zero
 903         if (ir->efep != FreeEnergyPerturbationType::No || ir->bSimTemp)
 904         {
 905             /* find and set the current lambdas */
 906             state->lambda = currentLambdas(step, *(ir->fepvals), state->fep_state);
 907
 908             bDoDHDL = do_per_step(step, ir->fepvals->nstdhdl);
 909             bDoFEP  = ((ir->efep != FreeEnergyPerturbationType::No) && do_per_step(step, nstfep));
 910             bDoExpanded = (do_per_step(step, ir->expandedvals->nstexpanded) && (ir->bExpanded)
 911                            && (!bFirstStep));
 912         }
 913
 914         bDoReplEx = (useReplicaExchange && (step > 0) && !bLastStep
 915                      && do_per_step(step, replExParams.exchangeInterval));
 916
 917         if (doSimulatedAnnealing)
 918         {
 919             // TODO: Avoid changing inputrec (#3854)
 920             // Simulated annealing updates the reference temperature.
 921             auto* nonConstInputrec = const_cast<t_inputrec*>(inputrec);
 922             update_annealing_target_temp(nonConstInputrec, t, &upd);
 923         }
 924
 925         /* Stop Center of Mass motion */
 926         bStopCM = (ir->comm_mode != ComRemovalAlgorithm::No && do_per_step(step, ir->nstcomm));
 927
 928         /* Determine whether or not to do Neighbour Searching */
 929         bNS = (bFirstStep || bNStList || bExchanged || bNeedRepartition);
 930
 931         /* Note that the stopHandler will cause termination at nstglobalcomm
 932          * steps. Since this concides with nstcalcenergy, nsttcouple and/or
 933          * nstpcouple steps, we have computed the half-step kinetic energy
 934          * of the previous step and can always output energies at the last step.
 935          */
 936         bLastStep = bLastStep || stopHandler->stoppingAfterCurrentStep(bNS);
 937
 938         /* do_log triggers energy and virial calculation. Because this leads
 939          * to different code paths, forces can be different. Thus for exact
 940          * continuation we should avoid extra log output.
 941          * Note that the || bLastStep can result in non-exact continuation
 942          * beyond the last step. But we don't consider that to be an issue.
 943          */
 944         do_log     = (do_per_step(step, ir->nstlog)
 945                   || (bFirstStep && startingBehavior == StartingBehavior::NewSimulation) || bLastStep);
 946         do_verbose = mdrunOptions.verbose
 947                      && (step % mdrunOptions.verboseStepPrintInterval == 0 || bFirstStep || bLastStep);
 948
 949         if (useGpuForUpdate && !bFirstStep && bNS)
 950         {
 951             // Copy velocities from the GPU on search steps to keep a copy on host (device buffers are reinitialized).
 952             stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
 953             stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
 954             // Copy coordinate from the GPU when needed at the search step.
 955             // NOTE: The cases when coordinates needed on CPU for force evaluation are handled in sim_utils.
 956             // NOTE: If the coordinates are to be written into output file they are also copied separately before the output.
 957             stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
 958             stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
 959         }
 960
 961         // We only need to calculate virtual velocities if we are writing them in the current step
 962         const bool needVirtualVelocitiesThisStep =
 963                 (vsite != nullptr)
 964                 && (do_per_step(step, ir->nstvout) || checkpointHandler->isCheckpointingStep());
 965
 966         if (vsite != nullptr)
 967         {
 968             // Virtual sites need to be updated before domain decomposition and forces are calculated
 969             wallcycle_start(wcycle, WallCycleCounter::VsiteConstr);
 970             // md-vv calculates virtual velocities once it has full-step real velocities
 971             vsite->construct(state->x,
 972                              state->v,
 973                              state->box,
 974                              (!EI_VV(inputrec->eI) && needVirtualVelocitiesThisStep)
 975                                      ? VSiteOperation::PositionsAndVelocities
 976                                      : VSiteOperation::Positions);
 977             wallcycle_stop(wcycle, WallCycleCounter::VsiteConstr);
 978         }
 979
 980         if (bNS && !(bFirstStep && ir->bContinuation))
 981         {
 982             bMasterState = FALSE;
 983             /* Correct the new box if it is too skewed */
 984             if (inputrecDynamicBox(ir))
 985             {
 986                 if (correct_box(fplog, step, state->box))
 987                 {
 988                     bMasterState = TRUE;
 989                     // If update is offloaded, it should be informed about the box size change
 990                     if (useGpuForUpdate)
 991                     {
 992                         integrator->setPbc(PbcType::Xyz, state->box);
 993                     }
 994                 }
 995             }
 996             if (DOMAINDECOMP(cr) && bMasterState)
 997             {
 998                 dd_collect_state(cr->dd, state, state_global);
 999             }
1000
1001             if (DOMAINDECOMP(cr))
1002             {
1003                 /* Repartition the domain decomposition */
1004                 dd_partition_system(fplog,
1005                                     mdlog,
1006                                     step,
1007                                     cr,
1008                                     bMasterState,
1009                                     nstglobalcomm,
1010                                     state_global,
1011                                     top_global,
1012                                     *ir,
1013                                     imdSession,
1014                                     pull_work,
1015                                     state,
1016                                     &f,
1017                                     mdAtoms,
1018                                     &top,
1019                                     fr,
1020                                     vsite,
1021                                     constr,
1022                                     nrnb,
1023                                     wcycle,
1024                                     do_verbose && !bPMETunePrinting);
1025                 upd.updateAfterPartition(state->natoms,
1026                                          md->cFREEZE ? gmx::arrayRefFromArray(md->cFREEZE, md->nr)
1027                                                      : gmx::ArrayRef<const unsigned short>(),
1028                                          md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1029                                                  : gmx::ArrayRef<const unsigned short>());
1030             }
1031         }
1032
1033         // Allocate or re-size GPU halo exchange object, if necessary
1034         if (bNS && havePPDomainDecomposition(cr) && simulationWork.useGpuHaloExchange)
1035         {
1036             GMX_RELEASE_ASSERT(fr->deviceStreamManager != nullptr,
1037                                "GPU device manager has to be initialized to use GPU "
1038                                "version of halo exchange.");
1039             constructGpuHaloExchange(mdlog, *cr, *fr->deviceStreamManager, wcycle);
1040         }
1041
1042         if (MASTER(cr) && do_log)
1043         {
1044             gmx::EnergyOutput::printHeader(
1045                     fplog, step, t); /* can we improve the information printed here? */
1046         }
1047
1048         if (ir->efep != FreeEnergyPerturbationType::No)
1049         {
1050             update_mdatoms(mdAtoms->mdatoms(), state->lambda[FreeEnergyPerturbationCouplingType::Mass]);
1051         }
1052
1053         if (bExchanged)
1054         {
1055             /* We need the kinetic energy at minus the half step for determining
1056              * the full step kinetic energy and possibly for T-coupling.*/
1057             /* This may not be quite working correctly yet . . . . */
1058             int cglo_flags = CGLO_GSTAT | CGLO_TEMPERATURE;
1059             if (DOMAINDECOMP(cr) && shouldCheckNumberOfBondedInteractions(*cr->dd))
1060             {
1061                 cglo_flags |= CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS;
1062             }
1063             compute_globals(gstat,
1064                             cr,
1065                             ir,
1066                             fr,
1067                             ekind,
1068                             makeConstArrayRef(state->x),
1069                             makeConstArrayRef(state->v),
1070                             state->box,
1071                             md,
1072                             nrnb,
1073                             &vcm,
1074                             wcycle,
1075                             enerd,
1076                             nullptr,
1077                             nullptr,
1078                             nullptr,
1079                             nullptr,
1080                             gmx::ArrayRef<real>{},
1081                             &nullSignaller,
1082                             state->box,
1083                             &bSumEkinhOld,
1084                             cglo_flags);
1085             if (DOMAINDECOMP(cr))
1086             {
1087                 checkNumberOfBondedInteractions(
1088                         mdlog, cr, top_global, &top, makeConstArrayRef(state->x), state->box);
1089             }
1090         }
1091         clear_mat(force_vir);
1092
1093         checkpointHandler->decideIfCheckpointingThisStep(bNS, bFirstStep, bLastStep);
1094
1095         /* Determine the energy and pressure:
1096          * at nstcalcenergy steps and at energy output steps (set below).
1097          */
1098         if (EI_VV(ir->eI) && (!bInitStep))
1099         {
1100             bCalcEnerStep = do_per_step(step, ir->nstcalcenergy);
1101             bCalcVir      = bCalcEnerStep
1102                        || (ir->epc != PressureCoupling::No
1103                            && (do_per_step(step, ir->nstpcouple) || do_per_step(step - 1, ir->nstpcouple)));
1104         }
1105         else
1106         {
1107             bCalcEnerStep = do_per_step(step, ir->nstcalcenergy);
1108             bCalcVir      = bCalcEnerStep
1109                        || (ir->epc != PressureCoupling::No && do_per_step(step, ir->nstpcouple));
1110         }
1111         bCalcEner = bCalcEnerStep;
1112
1113         do_ene = (do_per_step(step, ir->nstenergy) || bLastStep);
1114
1115         if (do_ene || do_log || bDoReplEx)
1116         {
1117             bCalcVir  = TRUE;
1118             bCalcEner = TRUE;
1119         }
1120
1121         /* Do we need global communication ? */
1122         bGStat = (bCalcVir || bCalcEner || bStopCM || do_per_step(step, nstglobalcomm)
1123                   || (EI_VV(ir->eI) && inputrecNvtTrotter(ir) && do_per_step(step - 1, nstglobalcomm)));
1124
1125         force_flags = (GMX_FORCE_STATECHANGED | ((inputrecDynamicBox(ir)) ? GMX_FORCE_DYNAMICBOX : 0)
1126                        | GMX_FORCE_ALLFORCES | (bCalcVir ? GMX_FORCE_VIRIAL : 0)
1127                        | (bCalcEner ? GMX_FORCE_ENERGY : 0) | (bDoFEP ? GMX_FORCE_DHDL : 0));
1128         if (fr->useMts && !do_per_step(step, ir->nstfout))
1129         {
1130             // TODO: merge this with stepWork.useOnlyMtsCombinedForceBuffer
1131             force_flags |= GMX_FORCE_DO_NOT_NEED_NORMAL_FORCE;
1132         }
1133
1134         if (shellfc)
1135         {
1136             /* Now is the time to relax the shells */
1137             relax_shell_flexcon(fplog,
1138                                 cr,
1139                                 ms,
1140                                 mdrunOptions.verbose,
1141                                 enforcedRotation,
1142                                 step,
1143                                 ir,
1144                                 imdSession,
1145                                 pull_work,
1146                                 bNS,
1147                                 force_flags,
1148                                 &top,
1149                                 constr,
1150                                 enerd,
1151                                 state->natoms,
1152                                 state->x.arrayRefWithPadding(),
1153                                 state->v.arrayRefWithPadding(),
1154                                 state->box,
1155                                 state->lambda,
1156                                 &state->hist,
1157                                 &f.view(),
1158                                 force_vir,
1159                                 *md,
1160                                 nrnb,
1161                                 wcycle,
1162                                 shellfc,
1163                                 fr,
1164                                 runScheduleWork,
1165                                 t,
1166                                 mu_tot,
1167                                 vsite,
1168                                 ddBalanceRegionHandler);
1169         }
1170         else
1171         {
1172             /* The AWH history need to be saved _before_ doing force calculations where the AWH bias
1173                is updated (or the AWH update will be performed twice for one step when continuing).
1174                It would be best to call this update function from do_md_trajectory_writing but that
1175                would occur after do_force. One would have to divide the update_awh function into one
1176                function applying the AWH force and one doing the AWH bias update. The update AWH
1177                bias function could then be called after do_md_trajectory_writing (then containing
1178                update_awh_history). The checkpointing will in the future probably moved to the start
1179                of the md loop which will rid of this issue. */
1180             if (awh && checkpointHandler->isCheckpointingStep() && MASTER(cr))
1181             {
1182                 awh->updateHistory(state_global->awhHistory.get());
1183             }
1184
1185             /* The coordinates (x) are shifted (to get whole molecules)
1186              * in do_force.
1187              * This is parallellized as well, and does communication too.
1188              * Check comments in sim_util.c
1189              */
1190             do_force(fplog,
1191                      cr,
1192                      ms,
1193                      *ir,
1194                      awh.get(),
1195                      enforcedRotation,
1196                      imdSession,
1197                      pull_work,
1198                      step,
1199                      nrnb,
1200                      wcycle,
1201                      &top,
1202                      state->box,
1203                      state->x.arrayRefWithPadding(),
1204                      &state->hist,
1205                      &f.view(),
1206                      force_vir,
1207                      md,
1208                      enerd,
1209                      state->lambda,
1210                      fr,
1211                      runScheduleWork,
1212                      vsite,
1213                      mu_tot,
1214                      t,
1215                      ed ? ed->getLegacyED() : nullptr,
1216                      (bNS ? GMX_FORCE_NS : 0) | force_flags,
1217                      ddBalanceRegionHandler);
1218         }
1219
1220         // VV integrators do not need the following velocity half step
1221         // if it is the first step after starting from a checkpoint.
1222         // That is, the half step is needed on all other steps, and
1223         // also the first step when starting from a .tpr file.
1224         if (EI_VV(ir->eI))
1225         {
1226             integrateVVFirstStep(step,
1227                                  bFirstStep,
1228                                  bInitStep,
1229                                  startingBehavior,
1230                                  nstglobalcomm,
1231                                  ir,
1232                                  fr,
1233                                  cr,
1234                                  state,
1235                                  mdAtoms->mdatoms(),
1236                                  fcdata,
1237                                  &MassQ,
1238                                  &vcm,
1239                                  top_global,
1240                                  top,
1241                                  enerd,
1242                                  ekind,
1243                                  gstat,
1244                                  &last_ekin,
1245                                  bCalcVir,
1246                                  total_vir,
1247                                  shake_vir,
1248                                  force_vir,
1249                                  pres,
1250                                  M,
1251                                  do_log,
1252                                  do_ene,
1253                                  bCalcEner,
1254                                  bGStat,
1255                                  bStopCM,
1256                                  bTrotter,
1257                                  bExchanged,
1258                                  &bSumEkinhOld,
1259                                  &saved_conserved_quantity,
1260                                  &f,
1261                                  &upd,
1262                                  constr,
1263                                  &nullSignaller,
1264                                  trotter_seq,
1265                                  nrnb,
1266                                  mdlog,
1267                                  fplog,
1268                                  wcycle);
1269             if (vsite != nullptr && needVirtualVelocitiesThisStep)
1270             {
1271                 // Positions were calculated earlier
1272                 wallcycle_start(wcycle, WallCycleCounter::VsiteConstr);
1273                 vsite->construct(state->x, state->v, state->box, VSiteOperation::Velocities);
1274                 wallcycle_stop(wcycle, WallCycleCounter::VsiteConstr);
1275             }
1276         }
1277
1278         /* ########  END FIRST UPDATE STEP  ############## */
1279         /* ########  If doing VV, we now have v(dt) ###### */
1280         if (bDoExpanded)
1281         {
1282             /* perform extended ensemble sampling in lambda - we don't
1283                actually move to the new state before outputting
1284                statistics, but if performing simulated tempering, we
1285                do update the velocities and the tau_t. */
1286             // TODO: Avoid changing inputrec (#3854)
1287             // Simulated tempering updates the reference temperature.
1288             // Expanded ensemble without simulated tempering does not change the inputrec.
1289             auto* nonConstInputrec = const_cast<t_inputrec*>(inputrec);
1290             lamnew                 = ExpandedEnsembleDynamics(fplog,
1291                                               nonConstInputrec,
1292                                               enerd,
1293                                               state,
1294                                               &MassQ,
1295                                               state->fep_state,
1296                                               state->dfhist,
1297                                               step,
1298                                               state->v.rvec_array(),
1299                                               md->homenr,
1300                                               md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1301                                                                       : gmx::ArrayRef<const unsigned short>());
1302             /* history is maintained in state->dfhist, but state_global is what is sent to trajectory and log output */
1303             if (MASTER(cr))
1304             {
1305                 copy_df_history(state_global->dfhist, state->dfhist);
1306             }
1307         }
1308
1309         // Copy coordinate from the GPU for the output/checkpointing if the update is offloaded and
1310         // coordinates have not already been copied for i) search or ii) CPU force tasks.
1311         if (useGpuForUpdate && !bNS && !runScheduleWork->domainWork.haveCpuLocalForceWork
1312             && (do_per_step(step, ir->nstxout) || do_per_step(step, ir->nstxout_compressed)
1313                 || checkpointHandler->isCheckpointingStep()))
1314         {
1315             stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
1316             stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
1317         }
1318         // Copy velocities if needed for the output/checkpointing.
1319         // NOTE: Copy on the search steps is done at the beginning of the step.
1320         if (useGpuForUpdate && !bNS
1321             && (do_per_step(step, ir->nstvout) || checkpointHandler->isCheckpointingStep()))
1322         {
1323             stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
1324             stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
1325         }
1326         // Copy forces for the output if the forces were reduced on the GPU (not the case on virial steps)
1327         // and update is offloaded hence forces are kept on the GPU for update and have not been
1328         // already transferred in do_force().
1329         // TODO: There should be an improved, explicit mechanism that ensures this copy is only executed
1330         //       when the forces are ready on the GPU -- the same synchronizer should be used as the one
1331         //       prior to GPU update.
1332         // TODO: When the output flags will be included in step workload, this copy can be combined with the
1333         //       copy call in do_force(...).
1334         // NOTE: The forces should not be copied here if the vsites are present, since they were modified
1335         //       on host after the D2H copy in do_force(...).
1336         if (runScheduleWork->stepWork.useGpuFBufferOps && (simulationWork.useGpuUpdate && !vsite)
1337             && do_per_step(step, ir->nstfout))
1338         {
1339             stateGpu->copyForcesFromGpu(f.view().force(), AtomLocality::Local);
1340             stateGpu->waitForcesReadyOnHost(AtomLocality::Local);
1341         }
1342         /* Now we have the energies and forces corresponding to the
1343          * coordinates at time t. We must output all of this before
1344          * the update.
1345          */
1346         do_md_trajectory_writing(fplog,
1347                                  cr,
1348                                  nfile,
1349                                  fnm,
1350                                  step,
1351                                  step_rel,
1352                                  t,
1353                                  ir,
1354                                  state,
1355                                  state_global,
1356                                  observablesHistory,
1357                                  top_global,
1358                                  fr,
1359                                  outf,
1360                                  energyOutput,
1361                                  ekind,
1362                                  f.view().force(),
1363                                  checkpointHandler->isCheckpointingStep(),
1364                                  bRerunMD,
1365                                  bLastStep,
1366                                  mdrunOptions.writeConfout,
1367                                  bSumEkinhOld);
1368         /* Check if IMD step and do IMD communication, if bIMD is TRUE. */
1369         bInteractiveMDstep = imdSession->run(step, bNS, state->box, state->x, t);
1370
1371         /* kludge -- virial is lost with restart for MTTK NPT control. Must reload (saved earlier). */
1372         if (startingBehavior != StartingBehavior::NewSimulation && bFirstStep
1373             && (inputrecNptTrotter(ir) || inputrecNphTrotter(ir)))
1374         {
1375             copy_mat(state->svir_prev, shake_vir);
1376             copy_mat(state->fvir_prev, force_vir);
1377         }
1378
1379         stopHandler->setSignal();
1380         resetHandler->setSignal(walltime_accounting);
1381
1382         if (bGStat || !PAR(cr))
1383         {
1384             /* In parallel we only have to check for checkpointing in steps
1385              * where we do global communication,
1386              *  otherwise the other nodes don't know.
1387              */
1388             checkpointHandler->setSignal(walltime_accounting);
1389         }
1390
1391         /* #########   START SECOND UPDATE STEP ################# */
1392
1393         /* at the start of step, randomize or scale the velocities ((if vv. Restriction of Andersen
1394            controlled in preprocessing */
1395
1396         if (ETC_ANDERSEN(ir->etc)) /* keep this outside of update_tcouple because of the extra info required to pass */
1397         {
1398             gmx_bool bIfRandomize;
1399             bIfRandomize = update_randomize_velocities(ir,
1400                                                        step,
1401                                                        cr,
1402                                                        md->homenr,
1403                                                        md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1404                                                                : gmx::ArrayRef<const unsigned short>(),
1405                                                        gmx::arrayRefFromArray(md->invmass, md->nr),
1406                                                        state->v,
1407                                                        &upd,
1408                                                        constr);
1409             /* if we have constraints, we have to remove the kinetic energy parallel to the bonds */
1410             if (constr && bIfRandomize)
1411             {
1412                 constrain_velocities(constr, do_log, do_ene, step, state, nullptr, false, nullptr);
1413             }
1414         }
1415         /* Box is changed in update() when we do pressure coupling,
1416          * but we should still use the old box for energy corrections and when
1417          * writing it to the energy file, so it matches the trajectory files for
1418          * the same timestep above. Make a copy in a separate array.
1419          */
1420         copy_mat(state->box, lastbox);
1421
1422         dvdl_constr = 0;
1423
1424         if (!useGpuForUpdate)
1425         {
1426             wallcycle_start(wcycle, WallCycleCounter::Update);
1427         }
1428         /* UPDATE PRESSURE VARIABLES IN TROTTER FORMULATION WITH CONSTRAINTS */
1429         if (bTrotter)
1430         {
1431             trotter_update(ir,
1432                            step,
1433                            ekind,
1434                            enerd,
1435                            state,
1436                            total_vir,
1437                            md->homenr,
1438                            md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1439                                    : gmx::ArrayRef<const unsigned short>(),
1440                            gmx::arrayRefFromArray(md->invmass, md->nr),
1441                            &MassQ,
1442                            trotter_seq,
1443                            ettTSEQ3);
1444             /* We can only do Berendsen coupling after we have summed
1445              * the kinetic energy or virial. Since the happens
1446              * in global_state after update, we should only do it at
1447              * step % nstlist = 1 with bGStatEveryStep=FALSE.
1448              */
1449         }
1450         else
1451         {
1452             update_tcouple(step,
1453                            ir,
1454                            state,
1455                            ekind,
1456                            &MassQ,
1457                            md->homenr,
1458                            md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1459                                    : gmx::ArrayRef<const unsigned short>());
1460             update_pcouple_before_coordinates(fplog, step, ir, state, pressureCouplingMu, M, bInitStep);
1461         }
1462
1463         /* With leap-frog type integrators we compute the kinetic energy
1464          * at a whole time step as the average of the half-time step kinetic
1465          * energies of two subsequent steps. Therefore we need to compute the
1466          * half step kinetic energy also if we need energies at the next step.
1467          */
1468         const bool needHalfStepKineticEnergy =
1469                 (!EI_VV(ir->eI) && (do_per_step(step + 1, nstglobalcomm) || step_rel + 1 == ir->nsteps));
1470
1471         // Parrinello-Rahman requires the pressure to be availible before the update to compute
1472         // the velocity scaling matrix. Hence, it runs one step after the nstpcouple step.
1473         const bool doParrinelloRahman = (ir->epc == PressureCoupling::ParrinelloRahman
1474                                          && do_per_step(step + ir->nstpcouple - 1, ir->nstpcouple));
1475
1476         if (EI_VV(ir->eI))
1477         {
1478             GMX_ASSERT(!useGpuForUpdate, "GPU update is not supported with VVAK integrator.");
1479
1480             integrateVVSecondStep(step,
1481                                   ir,
1482                                   fr,
1483                                   cr,
1484                                   state,
1485                                   mdAtoms->mdatoms(),
1486                                   fcdata,
1487                                   &MassQ,
1488                                   &vcm,
1489                                   pull_work,
1490                                   enerd,
1491                                   ekind,
1492                                   gstat,
1493                                   &dvdl_constr,
1494                                   bCalcVir,
1495                                   total_vir,
1496                                   shake_vir,
1497                                   force_vir,
1498                                   pres,
1499                                   M,
1500                                   lastbox,
1501                                   do_log,
1502                                   do_ene,
1503                                   bGStat,
1504                                   &bSumEkinhOld,
1505                                   &f,
1506                                   &cbuf,
1507                                   &upd,
1508                                   constr,
1509                                   &nullSignaller,
1510                                   trotter_seq,
1511                                   nrnb,
1512                                   wcycle);
1513         }
1514         else
1515         {
1516             if (useGpuForUpdate)
1517             {
1518                 if (bNS && (bFirstStep || DOMAINDECOMP(cr)))
1519                 {
1520                     integrator->set(stateGpu->getCoordinates(),
1521                                     stateGpu->getVelocities(),
1522                                     stateGpu->getForces(),
1523                                     top.idef,
1524                                     *md);
1525
1526                     // Copy data to the GPU after buffers might have being reinitialized
1527                     // coordinates have been copied already if PME or buffer ops has not needed it this step.
1528                     stateGpu->copyVelocitiesToGpu(state->v, AtomLocality::Local);
1529                     const bool useGpuPmeOnThisRank = runScheduleWork->simulationWork.useGpuPme
1530                                                      && thisRankHasDuty(cr, DUTY_PME)
1531                                                      && runScheduleWork->stepWork.computeSlowForces;
1532                     if (!useGpuPmeOnThisRank && !runScheduleWork->stepWork.useGpuXBufferOps)
1533                     {
1534                         stateGpu->copyCoordinatesToGpu(state->x, AtomLocality::Local);
1535                     }
1536                 }
1537
1538                 if (simulationWork.useGpuPme && !runScheduleWork->simulationWork.useGpuPmePpCommunication
1539                     && !thisRankHasDuty(cr, DUTY_PME))
1540                 {
1541                     // The PME forces were recieved to the host, so have to be copied
1542                     stateGpu->copyForcesToGpu(f.view().force(), AtomLocality::All);
1543                 }
1544                 else if (!runScheduleWork->stepWork.useGpuFBufferOps)
1545                 {
1546                     // The buffer ops were not offloaded this step, so the forces are on the
1547                     // host and have to be copied
1548                     stateGpu->copyForcesToGpu(f.view().force(), AtomLocality::Local);
1549                 }
1550
1551                 const bool doTemperatureScaling =
1552                         (ir->etc != TemperatureCoupling::No
1553                          && do_per_step(step + ir->nsttcouple - 1, ir->nsttcouple));
1554
1555                 // This applies Leap-Frog, LINCS and SETTLE in succession
1556                 integrator->integrate(
1557                         stateGpu->getForcesReadyOnDeviceEvent(
1558                                 AtomLocality::Local, runScheduleWork->stepWork.useGpuFBufferOps),
1559                         ir->delta_t,
1560                         true,
1561                         bCalcVir,
1562                         shake_vir,
1563                         doTemperatureScaling,
1564                         ekind->tcstat,
1565                         doParrinelloRahman,
1566                         ir->nstpcouple * ir->delta_t,
1567                         M);
1568
1569                 // Copy velocities D2H after update if:
1570                 // - Globals are computed this step (includes the energy output steps).
1571                 // - Temperature is needed for the next step.
1572                 if (bGStat || needHalfStepKineticEnergy)
1573                 {
1574                     stateGpu->copyVelocitiesFromGpu(state->v, AtomLocality::Local);
1575                     stateGpu->waitVelocitiesReadyOnHost(AtomLocality::Local);
1576                 }
1577             }
1578             else
1579             {
1580                 /* With multiple time stepping we need to do an additional normal
1581                  * update step to obtain the virial, as the actual MTS integration
1582                  * using an acceleration where the slow forces are multiplied by mtsFactor.
1583                  * Using that acceleration would result in a virial with the slow
1584                  * force contribution would be a factor mtsFactor too large.
1585                  */
1586                 if (fr->useMts && bCalcVir && constr != nullptr)
1587                 {
1588                     upd.update_for_constraint_virial(*ir,
1589                                                      md->homenr,
1590                                                      md->havePartiallyFrozenAtoms,
1591                                                      gmx::arrayRefFromArray(md->invmass, md->nr),
1592                                                      gmx::arrayRefFromArray(md->invMassPerDim, md->nr),
1593                                                      *state,
1594                                                      f.view().forceWithPadding(),
1595                                                      *ekind);
1596
1597                     constrain_coordinates(constr,
1598                                           do_log,
1599                                           do_ene,
1600                                           step,
1601                                           state,
1602                                           upd.xp()->arrayRefWithPadding(),
1603                                           &dvdl_constr,
1604                                           bCalcVir,
1605                                           shake_vir);
1606                 }
1607
1608                 ArrayRefWithPadding<const RVec> forceCombined =
1609                         (fr->useMts && step % ir->mtsLevels[1].stepFactor == 0)
1610                                 ? f.view().forceMtsCombinedWithPadding()
1611                                 : f.view().forceWithPadding();
1612                 upd.update_coords(*ir,
1613                                   step,
1614                                   md->homenr,
1615                                   md->havePartiallyFrozenAtoms,
1616                                   gmx::arrayRefFromArray(md->ptype, md->nr),
1617                                   gmx::arrayRefFromArray(md->invmass, md->nr),
1618                                   gmx::arrayRefFromArray(md->invMassPerDim, md->nr),
1619                                   state,
1620                                   forceCombined,
1621                                   fcdata,
1622                                   ekind,
1623                                   M,
1624                                   etrtPOSITION,
1625                                   cr,
1626                                   constr != nullptr);
1627
1628                 wallcycle_stop(wcycle, WallCycleCounter::Update);
1629
1630                 constrain_coordinates(constr,
1631                                       do_log,
1632                                       do_ene,
1633                                       step,
1634                                       state,
1635                                       upd.xp()->arrayRefWithPadding(),
1636                                       &dvdl_constr,
1637                                       bCalcVir && !fr->useMts,
1638                                       shake_vir);
1639
1640                 upd.update_sd_second_half(*ir,
1641                                           step,
1642                                           &dvdl_constr,
1643                                           md->homenr,
1644                                           gmx::arrayRefFromArray(md->ptype, md->nr),
1645                                           gmx::arrayRefFromArray(md->invmass, md->nr),
1646                                           state,
1647                                           cr,
1648                                           nrnb,
1649                                           wcycle,
1650                                           constr,
1651                                           do_log,
1652                                           do_ene);
1653                 upd.finish_update(
1654                         *ir, md->havePartiallyFrozenAtoms, md->homenr, state, wcycle, constr != nullptr);
1655             }
1656
1657             if (ir->bPull && ir->pull->bSetPbcRefToPrevStepCOM)
1658             {
1659                 updatePrevStepPullCom(pull_work, state);
1660             }
1661
1662             enerd->term[F_DVDL_CONSTR] += dvdl_constr;
1663         }
1664
1665         /* ############## IF NOT VV, Calculate globals HERE  ############ */
1666         /* With Leap-Frog we can skip compute_globals at
1667          * non-communication steps, but we need to calculate
1668          * the kinetic energy one step before communication.
1669          */
1670         {
1671             // Organize to do inter-simulation signalling on steps if
1672             // and when algorithms require it.
1673             const bool doInterSimSignal = (simulationsShareState && do_per_step(step, nstSignalComm));
1674
1675             if (bGStat || needHalfStepKineticEnergy || doInterSimSignal)
1676             {
1677                 // Copy coordinates when needed to stop the CM motion.
1678                 if (useGpuForUpdate && !EI_VV(ir->eI) && bStopCM)
1679                 {
1680                     stateGpu->copyCoordinatesFromGpu(state->x, AtomLocality::Local);
1681                     stateGpu->waitCoordinatesReadyOnHost(AtomLocality::Local);
1682                 }
1683                 // Since we're already communicating at this step, we
1684                 // can propagate intra-simulation signals. Note that
1685                 // check_nstglobalcomm has the responsibility for
1686                 // choosing the value of nstglobalcomm that is one way
1687                 // bGStat becomes true, so we can't get into a
1688                 // situation where e.g. checkpointing can't be
1689                 // signalled.
1690                 bool                doIntraSimSignal = true;
1691                 SimulationSignaller signaller(&signals, cr, ms, doInterSimSignal, doIntraSimSignal);
1692
1693                 compute_globals(
1694                         gstat,
1695                         cr,
1696                         ir,
1697                         fr,
1698                         ekind,
1699                         makeConstArrayRef(state->x),
1700                         makeConstArrayRef(state->v),
1701                         state->box,
1702                         md,
1703                         nrnb,
1704                         &vcm,
1705                         wcycle,
1706                         enerd,
1707                         force_vir,
1708                         shake_vir,
1709                         total_vir,
1710                         pres,
1711                         (!EI_VV(ir->eI) && bCalcEner && constr != nullptr) ? constr->rmsdData()
1712                                                                            : gmx::ArrayRef<real>{},
1713                         &signaller,
1714                         lastbox,
1715                         &bSumEkinhOld,
1716                         (bGStat ? CGLO_GSTAT : 0) | (!EI_VV(ir->eI) && bCalcEner ? CGLO_ENERGY : 0)
1717                                 | (!EI_VV(ir->eI) && bStopCM ? CGLO_STOPCM : 0)
1718                                 | (!EI_VV(ir->eI) ? CGLO_TEMPERATURE : 0)
1719                                 | (!EI_VV(ir->eI) ? CGLO_PRESSURE : 0) | CGLO_CONSTRAINT
1720                                 | (DOMAINDECOMP(cr) && shouldCheckNumberOfBondedInteractions(*cr->dd)
1721                                            ? CGLO_CHECK_NUMBER_OF_BONDED_INTERACTIONS
1722                                            : 0));
1723                 if (DOMAINDECOMP(cr))
1724                 {
1725                     checkNumberOfBondedInteractions(
1726                             mdlog, cr, top_global, &top, makeConstArrayRef(state->x), state->box);
1727                 }
1728                 if (!EI_VV(ir->eI) && bStopCM)
1729                 {
1730                     process_and_stopcm_grp(
1731                             fplog, &vcm, *md, makeArrayRef(state->x), makeArrayRef(state->v));
1732                     inc_nrnb(nrnb, eNR_STOPCM, md->homenr);
1733
1734                     // TODO: The special case of removing CM motion should be dealt more gracefully
1735                     if (useGpuForUpdate)
1736                     {
1737                         stateGpu->copyCoordinatesToGpu(state->x, AtomLocality::Local);
1738                         // Here we block until the H2D copy completes because event sync with the
1739                         // force kernels that use the coordinates on the next steps is not implemented
1740                         // (not because of a race on state->x being modified on the CPU while H2D is in progress).
1741                         stateGpu->waitCoordinatesCopiedToDevice(AtomLocality::Local);
1742                         // If the COM removal changed the velocities on the CPU, this has to be accounted for.
1743                         if (vcm.mode != ComRemovalAlgorithm::No)
1744                         {
1745                             stateGpu->copyVelocitiesToGpu(state->v, AtomLocality::Local);
1746                         }
1747                     }
1748                 }
1749             }
1750         }
1751
1752         /* #############  END CALC EKIN AND PRESSURE ################# */
1753
1754         /* Note: this is OK, but there are some numerical precision issues with using the convergence of
1755            the virial that should probably be addressed eventually. state->veta has better properies,
1756            but what we actually need entering the new cycle is the new shake_vir value. Ideally, we could
1757            generate the new shake_vir, but test the veta value for convergence.  This will take some thought. */
1758
1759         if (ir->efep != FreeEnergyPerturbationType::No && !EI_VV(ir->eI))
1760         {
1761             /* Sum up the foreign energy and dK/dl terms for md and sd.
1762                Currently done every step so that dH/dl is correct in the .edr */
1763             accumulateKineticLambdaComponents(enerd, state->lambda, *ir->fepvals);
1764         }
1765
1766         update_pcouple_after_coordinates(fplog,
1767                                          step,
1768                                          ir,
1769                                          md->homenr,
1770                                          md->cFREEZE ? gmx::arrayRefFromArray(md->cFREEZE, md->nr)
1771                                                      : gmx::ArrayRef<const unsigned short>(),
1772                                          pres,
1773                                          force_vir,
1774                                          shake_vir,
1775                                          pressureCouplingMu,
1776                                          state,
1777                                          nrnb,
1778                                          upd.deform(),
1779                                          !useGpuForUpdate);
1780
1781         const bool doBerendsenPressureCoupling = (inputrec->epc == PressureCoupling::Berendsen
1782                                                   && do_per_step(step, inputrec->nstpcouple));
1783         const bool doCRescalePressureCoupling  = (inputrec->epc == PressureCoupling::CRescale
1784                                                  && do_per_step(step, inputrec->nstpcouple));
1785         if (useGpuForUpdate
1786             && (doBerendsenPressureCoupling || doCRescalePressureCoupling || doParrinelloRahman))
1787         {
1788             integrator->scaleCoordinates(pressureCouplingMu);
1789             if (doCRescalePressureCoupling)
1790             {
1791                 matrix pressureCouplingInvMu;
1792                 gmx::invertBoxMatrix(pressureCouplingMu, pressureCouplingInvMu);
1793                 integrator->scaleVelocities(pressureCouplingInvMu);
1794             }
1795             integrator->setPbc(PbcType::Xyz, state->box);
1796         }
1797
1798         /* ################# END UPDATE STEP 2 ################# */
1799         /* #### We now have r(t+dt) and v(t+dt/2)  ############# */
1800
1801         /* The coordinates (x) were unshifted in update */
1802         if (!bGStat)
1803         {
1804             /* We will not sum ekinh_old,
1805              * so signal that we still have to do it.
1806              */
1807             bSumEkinhOld = TRUE;
1808         }
1809
1810         if (bCalcEner)
1811         {
1812             /* #########  BEGIN PREPARING EDR OUTPUT  ###########  */
1813
1814             /* use the directly determined last velocity, not actually the averaged half steps */
1815             if (bTrotter && ir->eI == IntegrationAlgorithm::VV)
1816             {
1817                 enerd->term[F_EKIN] = last_ekin;
1818             }
1819             enerd->term[F_ETOT] = enerd->term[F_EPOT] + enerd->term[F_EKIN];
1820
1821             if (integratorHasConservedEnergyQuantity(ir))
1822             {
1823                 if (EI_VV(ir->eI))
1824                 {
1825                     enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + saved_conserved_quantity;
1826                 }
1827                 else
1828                 {
1829                     enerd->term[F_ECONSERVED] = enerd->term[F_ETOT] + NPT_energy(ir, state, &MassQ);
1830                 }
1831             }
1832             /* #########  END PREPARING EDR OUTPUT  ###########  */
1833         }
1834
1835         /* Output stuff */
1836         if (MASTER(cr))
1837         {
1838             if (fplog && do_log && bDoExpanded)
1839             {
1840                 /* only needed if doing expanded ensemble */
1841                 PrintFreeEnergyInfoToFile(fplog,
1842                                           ir->fepvals.get(),
1843                                           ir->expandedvals.get(),
1844                                           ir->bSimTemp ? ir->simtempvals.get() : nullptr,
1845                                           state_global->dfhist,
1846                                           state->fep_state,
1847                                           ir->nstlog,
1848                                           step);
1849             }
1850             if (bCalcEner)
1851             {
1852                 energyOutput.addDataAtEnergyStep(bDoDHDL,
1853                                                  bCalcEnerStep,
1854                                                  t,
1855                                                  md->tmass,
1856                                                  enerd,
1857                                                  ir->fepvals.get(),
1858                                                  ir->expandedvals.get(),
1859                                                  lastbox,
1860                                                  PTCouplingArrays{ state->boxv,
1861                                                                    state->nosehoover_xi,
1862                                                                    state->nosehoover_vxi,
1863                                                                    state->nhpres_xi,
1864                                                                    state->nhpres_vxi },
1865                                                  state->fep_state,
1866                                                  shake_vir,
1867                                                  force_vir,
1868                                                  total_vir,
1869                                                  pres,
1870                                                  ekind,
1871                                                  mu_tot,
1872                                                  constr);
1873             }
1874             else
1875             {
1876                 energyOutput.recordNonEnergyStep();
1877             }
1878
1879             gmx_bool do_dr = do_per_step(step, ir->nstdisreout);
1880             gmx_bool do_or = do_per_step(step, ir->nstorireout);
1881
1882             if (doSimulatedAnnealing)
1883             {
1884                 gmx::EnergyOutput::printAnnealingTemperatures(
1885                         do_log ? fplog : nullptr, groups, &(ir->opts));
1886             }
1887             if (do_log || do_ene || do_dr || do_or)
1888             {
1889                 energyOutput.printStepToEnergyFile(mdoutf_get_fp_ene(outf),
1890                                                    do_ene,
1891                                                    do_dr,
1892                                                    do_or,
1893                                                    do_log ? fplog : nullptr,
1894                                                    step,
1895                                                    t,
1896                                                    fr->fcdata.get(),
1897                                                    awh.get());
1898             }
1899             if (do_log && ir->bDoAwh && awh->hasFepLambdaDimension())
1900             {
1901                 const bool isInitialOutput = false;
1902                 printLambdaStateToLog(fplog, state->lambda, isInitialOutput);
1903             }
1904
1905             if (ir->bPull)
1906             {
1907                 pull_print_output(pull_work, step, t);
1908             }
1909
1910             if (do_per_step(step, ir->nstlog))
1911             {
1912                 if (fflush(fplog) != 0)
1913                 {
1914                     gmx_fatal(FARGS, "Cannot flush logfile - maybe you are out of disk space?");
1915                 }
1916             }
1917         }
1918         if (bDoExpanded)
1919         {
1920             /* Have to do this part _after_ outputting the logfile and the edr file */
1921             /* Gets written into the state at the beginning of next loop*/
1922             state->fep_state = lamnew;
1923         }
1924         else if (ir->bDoAwh && awh->needForeignEnergyDifferences(step))
1925         {
1926             state->fep_state = awh->fepLambdaState();
1927         }
1928         /* Print the remaining wall clock time for the run */
1929         if (isMasterSimMasterRank(ms, MASTER(cr)) && (do_verbose || gmx_got_usr_signal()) && !bPMETunePrinting)
1930         {
1931             if (shellfc)
1932             {
1933                 fprintf(stderr, "\n");
1934             }
1935             print_time(stderr, walltime_accounting, step, ir, cr);
1936         }
1937
1938         /* Ion/water position swapping.
1939          * Not done in last step since trajectory writing happens before this call
1940          * in the MD loop and exchanges would be lost anyway. */
1941         bNeedRepartition = FALSE;
1942         if ((ir->eSwapCoords != SwapType::No) && (step > 0) && !bLastStep
1943             && do_per_step(step, ir->swap->nstswap))
1944         {
1945             bNeedRepartition = do_swapcoords(cr,
1946                                              step,
1947                                              t,
1948                                              ir,
1949                                              swap,
1950                                              wcycle,
1951                                              as_rvec_array(state->x.data()),
1952                                              state->box,
1953                                              MASTER(cr) && mdrunOptions.verbose,
1954                                              bRerunMD);
1955
1956             if (bNeedRepartition && DOMAINDECOMP(cr))
1957             {
1958                 dd_collect_state(cr->dd, state, state_global);
1959             }
1960         }
1961
1962         /* Replica exchange */
1963         bExchanged = FALSE;
1964         if (bDoReplEx)
1965         {
1966             bExchanged = replica_exchange(fplog, cr, ms, repl_ex, state_global, enerd, state, step, t);
1967         }
1968
1969         if ((bExchanged || bNeedRepartition) && DOMAINDECOMP(cr))
1970         {
1971             dd_partition_system(fplog,
1972                                 mdlog,
1973                                 step,
1974                                 cr,
1975                                 TRUE,
1976                                 1,
1977                                 state_global,
1978                                 top_global,
1979                                 *ir,
1980                                 imdSession,
1981                                 pull_work,
1982                                 state,
1983                                 &f,
1984                                 mdAtoms,
1985                                 &top,
1986                                 fr,
1987                                 vsite,
1988                                 constr,
1989                                 nrnb,
1990                                 wcycle,
1991                                 FALSE);
1992             upd.updateAfterPartition(state->natoms,
1993                                      md->cFREEZE ? gmx::arrayRefFromArray(md->cFREEZE, md->nr)
1994                                                  : gmx::ArrayRef<const unsigned short>(),
1995                                      md->cTC ? gmx::arrayRefFromArray(md->cTC, md->nr)
1996                                              : gmx::ArrayRef<const unsigned short>());
1997         }
1998
1999         bFirstStep = FALSE;
2000         bInitStep  = FALSE;
2001
2002         /* #######  SET VARIABLES FOR NEXT ITERATION IF THEY STILL NEED IT ###### */
2003         /* With all integrators, except VV, we need to retain the pressure
2004          * at the current step for coupling at the next step.
2005          */
2006         if ((state->flags & enumValueToBitMask(StateEntry::PressurePrevious))
2007             && (bGStatEveryStep || (ir->nstpcouple > 0 && step % ir->nstpcouple == 0)))
2008         {
2009             /* Store the pressure in t_state for pressure coupling
2010              * at the next MD step.
2011              */
2012             copy_mat(pres, state->pres_prev);
2013         }
2014
2015         /* #######  END SET VARIABLES FOR NEXT ITERATION ###### */
2016
2017         if ((membed != nullptr) && (!bLastStep))
2018         {
2019             rescale_membed(step_rel, membed, as_rvec_array(state_global->x.data()));
2020         }
2021
2022         cycles = wallcycle_stop(wcycle, WallCycleCounter::Step);
2023         if (DOMAINDECOMP(cr) && wcycle)
2024         {
2025             dd_cycles_add(cr->dd, cycles, ddCyclStep);
2026         }
2027
2028         /* increase the MD step number */
2029         step++;
2030         step_rel++;
2031
2032 #if GMX_FAHCORE
2033         if (MASTER(cr))
2034         {
2035             fcReportProgress(ir->nsteps + ir->init_step, step);
2036         }
2037 #endif
2038
2039         resetHandler->resetCounters(
2040                 step, step_rel, mdlog, fplog, cr, fr->nbv.get(), nrnb, fr->pmedata, pme_loadbal, wcycle, walltime_accounting);
2041
2042         /* If bIMD is TRUE, the master updates the IMD energy record and sends positions to VMD client */
2043         imdSession->updateEnergyRecordAndSendPositionsAndEnergies(bInteractiveMDstep, step, bCalcEner);
2044     }
2045     /* End of main MD loop */
2046
2047     /* Closing TNG files can include compressing data. Therefore it is good to do that
2048      * before stopping the time measurements. */
2049     mdoutf_tng_close(outf);
2050
2051     /* Stop measuring walltime */
2052     walltime_accounting_end_time(walltime_accounting);
2053
2054     if (!thisRankHasDuty(cr, DUTY_PME))
2055     {
2056         /* Tell the PME only node to finish */
2057         gmx_pme_send_finish(cr);
2058     }
2059
2060     if (MASTER(cr))
2061     {
2062         if (ir->nstcalcenergy > 0)
2063         {
2064             energyOutput.printEnergyConservation(fplog, ir->simulation_part, EI_MD(ir->eI));
2065
2066             gmx::EnergyOutput::printAnnealingTemperatures(fplog, groups, &(ir->opts));
2067             energyOutput.printAverages(fplog, groups);
2068         }
2069     }
2070     done_mdoutf(outf);
2071
2072     if (bPMETune)
2073     {
2074         pme_loadbal_done(pme_loadbal, fplog, mdlog, fr->nbv->useGpu());
2075     }
2076
2077     done_shellfc(fplog, shellfc, step_rel);
2078
2079     if (useReplicaExchange && MASTER(cr))
2080     {
2081         print_replica_exchange_statistics(fplog, repl_ex);
2082     }
2083
2084     walltime_accounting_set_nsteps_done(walltime_accounting, step_rel);
2085
2086     global_stat_destroy(gstat);
2087 }