src/gromacs/mdlib/sim_util.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) 2013,2014,2015,2016,2017,2018,2019, 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 #include "gmxpre.h"
  38
  39 #include "sim_util.h"
  40
  41 #include "config.h"
  42
  43 #include <cmath>
  44 #include <cstdint>
  45 #include <cstdio>
  46 #include <cstring>
  47
  48 #include <array>
  49
  50 #include "gromacs/awh/awh.h"
  51 #include "gromacs/domdec/dlbtiming.h"
  52 #include "gromacs/domdec/domdec.h"
  53 #include "gromacs/domdec/domdec_struct.h"
  54 #include "gromacs/domdec/partition.h"
  55 #include "gromacs/essentialdynamics/edsam.h"
  56 #include "gromacs/ewald/pme.h"
  57 #include "gromacs/gmxlib/chargegroup.h"
  58 #include "gromacs/gmxlib/network.h"
  59 #include "gromacs/gmxlib/nrnb.h"
  60 #include "gromacs/gmxlib/nonbonded/nb_free_energy.h"
  61 #include "gromacs/gmxlib/nonbonded/nb_kernel.h"
  62 #include "gromacs/gmxlib/nonbonded/nonbonded.h"
  63 #include "gromacs/gpu_utils/gpu_utils.h"
  64 #include "gromacs/imd/imd.h"
  65 #include "gromacs/listed_forces/bonded.h"
  66 #include "gromacs/listed_forces/disre.h"
  67 #include "gromacs/listed_forces/gpubonded.h"
  68 #include "gromacs/listed_forces/listed_forces.h"
  69 #include "gromacs/listed_forces/manage_threading.h"
  70 #include "gromacs/listed_forces/orires.h"
  71 #include "gromacs/math/arrayrefwithpadding.h"
  72 #include "gromacs/math/functions.h"
  73 #include "gromacs/math/units.h"
  74 #include "gromacs/math/vec.h"
  75 #include "gromacs/math/vecdump.h"
  76 #include "gromacs/mdlib/calcmu.h"
  77 #include "gromacs/mdlib/calcvir.h"
  78 #include "gromacs/mdlib/constr.h"
  79 #include "gromacs/mdlib/force.h"
  80 #include "gromacs/mdlib/forcerec.h"
  81 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  82 #include "gromacs/mdlib/ppforceworkload.h"
  83 #include "gromacs/mdlib/qmmm.h"
  84 #include "gromacs/mdlib/update.h"
  85 #include "gromacs/mdtypes/commrec.h"
  86 #include "gromacs/mdtypes/enerdata.h"
  87 #include "gromacs/mdtypes/forceoutput.h"
  88 #include "gromacs/mdtypes/iforceprovider.h"
  89 #include "gromacs/mdtypes/inputrec.h"
  90 #include "gromacs/mdtypes/md_enums.h"
  91 #include "gromacs/mdtypes/state.h"
  92 #include "gromacs/nbnxm/atomdata.h"
  93 #include "gromacs/nbnxm/gpu_data_mgmt.h"
  94 #include "gromacs/nbnxm/nbnxm.h"
  95 #include "gromacs/pbcutil/ishift.h"
  96 #include "gromacs/pbcutil/mshift.h"
  97 #include "gromacs/pbcutil/pbc.h"
  98 #include "gromacs/pulling/pull.h"
  99 #include "gromacs/pulling/pull_rotation.h"
 100 #include "gromacs/timing/cyclecounter.h"
 101 #include "gromacs/timing/gpu_timing.h"
 102 #include "gromacs/timing/wallcycle.h"
 103 #include "gromacs/timing/wallcyclereporting.h"
 104 #include "gromacs/timing/walltime_accounting.h"
 105 #include "gromacs/topology/topology.h"
 106 #include "gromacs/utility/arrayref.h"
 107 #include "gromacs/utility/basedefinitions.h"
 108 #include "gromacs/utility/cstringutil.h"
 109 #include "gromacs/utility/exceptions.h"
 110 #include "gromacs/utility/fatalerror.h"
 111 #include "gromacs/utility/gmxassert.h"
 112 #include "gromacs/utility/gmxmpi.h"
 113 #include "gromacs/utility/logger.h"
 114 #include "gromacs/utility/smalloc.h"
 115 #include "gromacs/utility/strconvert.h"
 116 #include "gromacs/utility/sysinfo.h"
 117
 118 // TODO: this environment variable allows us to verify before release
 119 // that on less common architectures the total cost of polling is not larger than
 120 // a blocking wait (so polling does not introduce overhead when the static
 121 // PME-first ordering would suffice).
 122 static const bool c_disableAlternatingWait = (getenv("GMX_DISABLE_ALTERNATING_GPU_WAIT") != nullptr);
 123
 124
 125 static void sum_forces(rvec f[], gmx::ArrayRef<const gmx::RVec> forceToAdd)
 126 {
 127     const int      end = forceToAdd.size();
 128
 129     int gmx_unused nt = gmx_omp_nthreads_get(emntDefault);
 130 #pragma omp parallel for num_threads(nt) schedule(static)
 131     for (int i = 0; i < end; i++)
 132     {
 133         rvec_inc(f[i], forceToAdd[i]);
 134     }
 135 }
 136
 137 static void calc_virial(int start, int homenr, const rvec x[], const rvec f[],
 138                         tensor vir_part, const t_graph *graph, const matrix box,
 139                         t_nrnb *nrnb, const t_forcerec *fr, int ePBC)
 140 {
 141     /* The short-range virial from surrounding boxes */
 142     calc_vir(SHIFTS, fr->shift_vec, fr->fshift, vir_part, ePBC == epbcSCREW, box);
 143     inc_nrnb(nrnb, eNR_VIRIAL, SHIFTS);
 144
 145     /* Calculate partial virial, for local atoms only, based on short range.
 146      * Total virial is computed in global_stat, called from do_md
 147      */
 148     f_calc_vir(start, start+homenr, x, f, vir_part, graph, box);
 149     inc_nrnb(nrnb, eNR_VIRIAL, homenr);
 150
 151     if (debug)
 152     {
 153         pr_rvecs(debug, 0, "vir_part", vir_part, DIM);
 154     }
 155 }
 156
 157 static void pull_potential_wrapper(const t_commrec *cr,
 158                                    const t_inputrec *ir,
 159                                    const matrix box, gmx::ArrayRef<const gmx::RVec> x,
 160                                    gmx::ForceWithVirial *force,
 161                                    const t_mdatoms *mdatoms,
 162                                    gmx_enerdata_t *enerd,
 163                                    const real *lambda,
 164                                    double t,
 165                                    gmx_wallcycle_t wcycle)
 166 {
 167     t_pbc  pbc;
 168     real   dvdl;
 169
 170     /* Calculate the center of mass forces, this requires communication,
 171      * which is why pull_potential is called close to other communication.
 172      */
 173     wallcycle_start(wcycle, ewcPULLPOT);
 174     set_pbc(&pbc, ir->ePBC, box);
 175     dvdl                     = 0;
 176     enerd->term[F_COM_PULL] +=
 177         pull_potential(ir->pull_work, mdatoms, &pbc,
 178                        cr, t, lambda[efptRESTRAINT], as_rvec_array(x.data()), force, &dvdl);
 179     enerd->dvdl_lin[efptRESTRAINT] += dvdl;
 180     wallcycle_stop(wcycle, ewcPULLPOT);
 181 }
 182
 183 static void pme_receive_force_ener(const t_commrec      *cr,
 184                                    gmx::ForceWithVirial *forceWithVirial,
 185                                    gmx_enerdata_t       *enerd,
 186                                    gmx_wallcycle_t       wcycle)
 187 {
 188     real   e_q, e_lj, dvdl_q, dvdl_lj;
 189     float  cycles_ppdpme, cycles_seppme;
 190
 191     cycles_ppdpme = wallcycle_stop(wcycle, ewcPPDURINGPME);
 192     dd_cycles_add(cr->dd, cycles_ppdpme, ddCyclPPduringPME);
 193
 194     /* In case of node-splitting, the PP nodes receive the long-range
 195      * forces, virial and energy from the PME nodes here.
 196      */
 197     wallcycle_start(wcycle, ewcPP_PMEWAITRECVF);
 198     dvdl_q  = 0;
 199     dvdl_lj = 0;
 200     gmx_pme_receive_f(cr, forceWithVirial, &e_q, &e_lj, &dvdl_q, &dvdl_lj,
 201                       &cycles_seppme);
 202     enerd->term[F_COUL_RECIP] += e_q;
 203     enerd->term[F_LJ_RECIP]   += e_lj;
 204     enerd->dvdl_lin[efptCOUL] += dvdl_q;
 205     enerd->dvdl_lin[efptVDW]  += dvdl_lj;
 206
 207     if (wcycle)
 208     {
 209         dd_cycles_add(cr->dd, cycles_seppme, ddCyclPME);
 210     }
 211     wallcycle_stop(wcycle, ewcPP_PMEWAITRECVF);
 212 }
 213
 214 static void print_large_forces(FILE            *fp,
 215                                const t_mdatoms *md,
 216                                const t_commrec *cr,
 217                                int64_t          step,
 218                                real             forceTolerance,
 219                                const rvec      *x,
 220                                const rvec      *f)
 221 {
 222     real           force2Tolerance = gmx::square(forceTolerance);
 223     gmx::index     numNonFinite    = 0;
 224     for (int i = 0; i < md->homenr; i++)
 225     {
 226         real force2    = norm2(f[i]);
 227         bool nonFinite = !std::isfinite(force2);
 228         if (force2 >= force2Tolerance || nonFinite)
 229         {
 230             fprintf(fp, "step %" PRId64 " atom %6d  x %8.3f %8.3f %8.3f  force %12.5e\n",
 231                     step,
 232                     ddglatnr(cr->dd, i), x[i][XX], x[i][YY], x[i][ZZ], std::sqrt(force2));
 233         }
 234         if (nonFinite)
 235         {
 236             numNonFinite++;
 237         }
 238     }
 239     if (numNonFinite > 0)
 240     {
 241         /* Note that with MPI this fatal call on one rank might interrupt
 242          * the printing on other ranks. But we can only avoid that with
 243          * an expensive MPI barrier that we would need at each step.
 244          */
 245         gmx_fatal(FARGS, "At step %" PRId64 " detected non-finite forces on %td atoms", step, numNonFinite);
 246     }
 247 }
 248
 249 static void post_process_forces(const t_commrec           *cr,
 250                                 int64_t                    step,
 251                                 t_nrnb                    *nrnb,
 252                                 gmx_wallcycle_t            wcycle,
 253                                 const gmx_localtop_t      *top,
 254                                 const matrix               box,
 255                                 const rvec                 x[],
 256                                 rvec                       f[],
 257                                 gmx::ForceWithVirial      *forceWithVirial,
 258                                 tensor                     vir_force,
 259                                 const t_mdatoms           *mdatoms,
 260                                 const t_graph             *graph,
 261                                 const t_forcerec          *fr,
 262                                 const gmx_vsite_t         *vsite,
 263                                 int                        flags)
 264 {
 265     if (fr->haveDirectVirialContributions)
 266     {
 267         rvec *fDirectVir = as_rvec_array(forceWithVirial->force_.data());
 268
 269         if (vsite)
 270         {
 271             /* Spread the mesh force on virtual sites to the other particles...
 272              * This is parallellized. MPI communication is performed
 273              * if the constructing atoms aren't local.
 274              */
 275             matrix virial = { { 0 } };
 276             spread_vsite_f(vsite, x, fDirectVir, nullptr,
 277                            (flags & GMX_FORCE_VIRIAL) != 0, virial,
 278                            nrnb,
 279                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr, wcycle);
 280             forceWithVirial->addVirialContribution(virial);
 281         }
 282
 283         if (flags & GMX_FORCE_VIRIAL)
 284         {
 285             /* Now add the forces, this is local */
 286             sum_forces(f, forceWithVirial->force_);
 287
 288             /* Add the direct virial contributions */
 289             GMX_ASSERT(forceWithVirial->computeVirial_, "forceWithVirial should request virial computation when we request the virial");
 290             m_add(vir_force, forceWithVirial->getVirial(), vir_force);
 291
 292             if (debug)
 293             {
 294                 pr_rvecs(debug, 0, "vir_force", vir_force, DIM);
 295             }
 296         }
 297     }
 298
 299     if (fr->print_force >= 0)
 300     {
 301         print_large_forces(stderr, mdatoms, cr, step, fr->print_force, x, f);
 302     }
 303 }
 304
 305 static void do_nb_verlet(t_forcerec                       *fr,
 306                          const interaction_const_t        *ic,
 307                          gmx_enerdata_t                   *enerd,
 308                          const int                         flags,
 309                          const Nbnxm::InteractionLocality  ilocality,
 310                          const int                         clearF,
 311                          const int64_t                     step,
 312                          t_nrnb                           *nrnb,
 313                          gmx_wallcycle_t                   wcycle)
 314 {
 315     if (!(flags & GMX_FORCE_NONBONDED))
 316     {
 317         /* skip non-bonded calculation */
 318         return;
 319     }
 320
 321     nonbonded_verlet_t *nbv  = fr->nbv.get();
 322
 323     /* GPU kernel launch overhead is already timed separately */
 324     if (fr->cutoff_scheme != ecutsVERLET)
 325     {
 326         gmx_incons("Invalid cut-off scheme passed!");
 327     }
 328
 329     if (!nbv->useGpu())
 330     {
 331         /* When dynamic pair-list  pruning is requested, we need to prune
 332          * at nstlistPrune steps.
 333          */
 334         if (nbv->pairlistSets().isDynamicPruningStepCpu(step))
 335         {
 336             /* Prune the pair-list beyond fr->ic->rlistPrune using
 337              * the current coordinates of the atoms.
 338              */
 339             wallcycle_sub_start(wcycle, ewcsNONBONDED_PRUNING);
 340             nbv->dispatchPruneKernelCpu(ilocality, fr->shift_vec);
 341             wallcycle_sub_stop(wcycle, ewcsNONBONDED_PRUNING);
 342         }
 343
 344         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 345     }
 346
 347     nbv->dispatchNonbondedKernel(ilocality, *ic, flags, clearF, fr, enerd, nrnb);
 348
 349     if (!nbv->useGpu())
 350     {
 351         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 352     }
 353 }
 354
 355 gmx_bool use_GPU(const nonbonded_verlet_t *nbv)
 356 {
 357     return nbv != nullptr && nbv->useGpu();
 358 }
 359
 360 static inline void clear_rvecs_omp(int n, rvec v[])
 361 {
 362     int nth = gmx_omp_nthreads_get_simple_rvec_task(emntDefault, n);
 363
 364     /* Note that we would like to avoid this conditional by putting it
 365      * into the omp pragma instead, but then we still take the full
 366      * omp parallel for overhead (at least with gcc5).
 367      */
 368     if (nth == 1)
 369     {
 370         for (int i = 0; i < n; i++)
 371         {
 372             clear_rvec(v[i]);
 373         }
 374     }
 375     else
 376     {
 377 #pragma omp parallel for num_threads(nth) schedule(static)
 378         for (int i = 0; i < n; i++)
 379         {
 380             clear_rvec(v[i]);
 381         }
 382     }
 383 }
 384
 385 /*! \brief Return an estimate of the average kinetic energy or 0 when unreliable
 386  *
 387  * \param groupOptions  Group options, containing T-coupling options
 388  */
 389 static real averageKineticEnergyEstimate(const t_grpopts &groupOptions)
 390 {
 391     real nrdfCoupled   = 0;
 392     real nrdfUncoupled = 0;
 393     real kineticEnergy = 0;
 394     for (int g = 0; g < groupOptions.ngtc; g++)
 395     {
 396         if (groupOptions.tau_t[g] >= 0)
 397         {
 398             nrdfCoupled   += groupOptions.nrdf[g];
 399             kineticEnergy += groupOptions.nrdf[g]*0.5*groupOptions.ref_t[g]*BOLTZ;
 400         }
 401         else
 402         {
 403             nrdfUncoupled += groupOptions.nrdf[g];
 404         }
 405     }
 406
 407     /* This conditional with > also catches nrdf=0 */
 408     if (nrdfCoupled > nrdfUncoupled)
 409     {
 410         return kineticEnergy*(nrdfCoupled + nrdfUncoupled)/nrdfCoupled;
 411     }
 412     else
 413     {
 414         return 0;
 415     }
 416 }
 417
 418 /*! \brief This routine checks that the potential energy is finite.
 419  *
 420  * Always checks that the potential energy is finite. If step equals
 421  * inputrec.init_step also checks that the magnitude of the potential energy
 422  * is reasonable. Terminates with a fatal error when a check fails.
 423  * Note that passing this check does not guarantee finite forces,
 424  * since those use slightly different arithmetics. But in most cases
 425  * there is just a narrow coordinate range where forces are not finite
 426  * and energies are finite.
 427  *
 428  * \param[in] step      The step number, used for checking and printing
 429  * \param[in] enerd     The energy data; the non-bonded group energies need to be added to enerd.term[F_EPOT] before calling this routine
 430  * \param[in] inputrec  The input record
 431  */
 432 static void checkPotentialEnergyValidity(int64_t               step,
 433                                          const gmx_enerdata_t &enerd,
 434                                          const t_inputrec     &inputrec)
 435 {
 436     /* Threshold valid for comparing absolute potential energy against
 437      * the kinetic energy. Normally one should not consider absolute
 438      * potential energy values, but with a factor of one million
 439      * we should never get false positives.
 440      */
 441     constexpr real c_thresholdFactor = 1e6;
 442
 443     bool           energyIsNotFinite    = !std::isfinite(enerd.term[F_EPOT]);
 444     real           averageKineticEnergy = 0;
 445     /* We only check for large potential energy at the initial step,
 446      * because that is by far the most likely step for this too occur
 447      * and because computing the average kinetic energy is not free.
 448      * Note: nstcalcenergy >> 1 often does not allow to catch large energies
 449      * before they become NaN.
 450      */
 451     if (step == inputrec.init_step && EI_DYNAMICS(inputrec.eI))
 452     {
 453         averageKineticEnergy = averageKineticEnergyEstimate(inputrec.opts);
 454     }
 455
 456     if (energyIsNotFinite || (averageKineticEnergy > 0 &&
 457                               enerd.term[F_EPOT] > c_thresholdFactor*averageKineticEnergy))
 458     {
 459         gmx_fatal(FARGS, "Step %" PRId64 ": The total potential energy is %g, which is %s. The LJ and electrostatic contributions to the energy are %g and %g, respectively. A %s potential energy can be caused by overlapping interactions in bonded interactions or very large%s coordinate values. Usually this is caused by a badly- or non-equilibrated initial configuration, incorrect interactions or parameters in the topology.",
 460                   step,
 461                   enerd.term[F_EPOT],
 462                   energyIsNotFinite ? "not finite" : "extremely high",
 463                   enerd.term[F_LJ],
 464                   enerd.term[F_COUL_SR],
 465                   energyIsNotFinite ? "non-finite" : "very high",
 466                   energyIsNotFinite ? " or Nan" : "");
 467     }
 468 }
 469
 470 /*! \brief Return true if there are special forces computed this step.
 471  *
 472  * The conditionals exactly correspond to those in computeSpecialForces().
 473  */
 474 static bool
 475 haveSpecialForces(const t_inputrec              *inputrec,
 476                   ForceProviders                *forceProviders,
 477                   int                            forceFlags,
 478                   const gmx_edsam               *ed)
 479 {
 480     const bool computeForces = (forceFlags & GMX_FORCE_FORCES) != 0;
 481
 482     return
 483         ((computeForces && forceProviders->hasForceProvider()) ||         // forceProviders
 484          (inputrec->bPull && pull_have_potential(inputrec->pull_work)) || // pull
 485          inputrec->bRot ||                                                // enforced rotation
 486          (ed != nullptr) ||                                               // flooding
 487          (inputrec->bIMD && computeForces));                              // IMD
 488 }
 489
 490 /*! \brief Compute forces and/or energies for special algorithms
 491  *
 492  * The intention is to collect all calls to algorithms that compute
 493  * forces on local atoms only and that do not contribute to the local
 494  * virial sum (but add their virial contribution separately).
 495  * Eventually these should likely all become ForceProviders.
 496  * Within this function the intention is to have algorithms that do
 497  * global communication at the end, so global barriers within the MD loop
 498  * are as close together as possible.
 499  *
 500  * \param[in]     fplog            The log file
 501  * \param[in]     cr               The communication record
 502  * \param[in]     inputrec         The input record
 503  * \param[in]     awh              The Awh module (nullptr if none in use).
 504  * \param[in]     enforcedRotation Enforced rotation module.
 505  * \param[in]     step             The current MD step
 506  * \param[in]     t                The current time
 507  * \param[in,out] wcycle           Wallcycle accounting struct
 508  * \param[in,out] forceProviders   Pointer to a list of force providers
 509  * \param[in]     box              The unit cell
 510  * \param[in]     x                The coordinates
 511  * \param[in]     mdatoms          Per atom properties
 512  * \param[in]     lambda           Array of free-energy lambda values
 513  * \param[in]     forceFlags       Flags that tell whether we should compute forces/energies/virial
 514  * \param[in,out] forceWithVirial  Force and virial buffers
 515  * \param[in,out] enerd            Energy buffer
 516  * \param[in,out] ed               Essential dynamics pointer
 517  * \param[in]     bNS              Tells if we did neighbor searching this step, used for ED sampling
 518  *
 519  * \todo Remove bNS, which is used incorrectly.
 520  * \todo Convert all other algorithms called here to ForceProviders.
 521  */
 522 static void
 523 computeSpecialForces(FILE                          *fplog,
 524                      const t_commrec               *cr,
 525                      const t_inputrec              *inputrec,
 526                      gmx::Awh                      *awh,
 527                      gmx_enfrot                    *enforcedRotation,
 528                      int64_t                        step,
 529                      double                         t,
 530                      gmx_wallcycle_t                wcycle,
 531                      ForceProviders                *forceProviders,
 532                      matrix                         box,
 533                      gmx::ArrayRef<const gmx::RVec> x,
 534                      const t_mdatoms               *mdatoms,
 535                      real                          *lambda,
 536                      int                            forceFlags,
 537                      gmx::ForceWithVirial          *forceWithVirial,
 538                      gmx_enerdata_t                *enerd,
 539                      gmx_edsam                     *ed,
 540                      gmx_bool                       bNS)
 541 {
 542     const bool computeForces = (forceFlags & GMX_FORCE_FORCES) != 0;
 543
 544     /* NOTE: Currently all ForceProviders only provide forces.
 545      *       When they also provide energies, remove this conditional.
 546      */
 547     if (computeForces)
 548     {
 549         gmx::ForceProviderInput  forceProviderInput(x, *mdatoms, t, box, *cr);
 550         gmx::ForceProviderOutput forceProviderOutput(forceWithVirial, enerd);
 551
 552         /* Collect forces from modules */
 553         forceProviders->calculateForces(forceProviderInput, &forceProviderOutput);
 554     }
 555
 556     if (inputrec->bPull && pull_have_potential(inputrec->pull_work))
 557     {
 558         pull_potential_wrapper(cr, inputrec, box, x,
 559                                forceWithVirial,
 560                                mdatoms, enerd, lambda, t,
 561                                wcycle);
 562
 563         if (awh)
 564         {
 565             enerd->term[F_COM_PULL] +=
 566                 awh->applyBiasForcesAndUpdateBias(inputrec->ePBC, *mdatoms, box,
 567                                                   forceWithVirial,
 568                                                   t, step, wcycle, fplog);
 569         }
 570     }
 571
 572     rvec *f = as_rvec_array(forceWithVirial->force_.data());
 573
 574     /* Add the forces from enforced rotation potentials (if any) */
 575     if (inputrec->bRot)
 576     {
 577         wallcycle_start(wcycle, ewcROTadd);
 578         enerd->term[F_COM_PULL] += add_rot_forces(enforcedRotation, f, cr, step, t);
 579         wallcycle_stop(wcycle, ewcROTadd);
 580     }
 581
 582     if (ed)
 583     {
 584         /* Note that since init_edsam() is called after the initialization
 585          * of forcerec, edsam doesn't request the noVirSum force buffer.
 586          * Thus if no other algorithm (e.g. PME) requires it, the forces
 587          * here will contribute to the virial.
 588          */
 589         do_flood(cr, inputrec, as_rvec_array(x.data()), f, ed, box, step, bNS);
 590     }
 591
 592     /* Add forces from interactive molecular dynamics (IMD), if bIMD == TRUE. */
 593     if (inputrec->bIMD && computeForces)
 594     {
 595         IMD_apply_forces(inputrec->bIMD, inputrec->imd, cr, f, wcycle);
 596     }
 597 }
 598
 599 /*! \brief Launch the prepare_step and spread stages of PME GPU.
 600  *
 601  * \param[in]  pmedata       The PME structure
 602  * \param[in]  box           The box matrix
 603  * \param[in]  x             Coordinate array
 604  * \param[in]  flags         Force flags
 605  * \param[in]  pmeFlags      PME flags
 606  * \param[in]  wcycle        The wallcycle structure
 607  */
 608 static inline void launchPmeGpuSpread(gmx_pme_t      *pmedata,
 609                                       matrix          box,
 610                                       rvec            x[],
 611                                       int             flags,
 612                                       int             pmeFlags,
 613                                       gmx_wallcycle_t wcycle)
 614 {
 615     pme_gpu_prepare_computation(pmedata, (flags & GMX_FORCE_DYNAMICBOX) != 0, box, wcycle, pmeFlags);
 616     pme_gpu_launch_spread(pmedata, x, wcycle);
 617 }
 618
 619 /*! \brief Launch the FFT and gather stages of PME GPU
 620  *
 621  * This function only implements setting the output forces (no accumulation).
 622  *
 623  * \param[in]  pmedata        The PME structure
 624  * \param[in]  wcycle         The wallcycle structure
 625  */
 626 static void launchPmeGpuFftAndGather(gmx_pme_t        *pmedata,
 627                                      gmx_wallcycle_t   wcycle)
 628 {
 629     pme_gpu_launch_complex_transforms(pmedata, wcycle);
 630     pme_gpu_launch_gather(pmedata, wcycle, PmeForceOutputHandling::Set);
 631 }
 632
 633 /*! \brief
 634  *  Polling wait for either of the PME or nonbonded GPU tasks.
 635  *
 636  * Instead of a static order in waiting for GPU tasks, this function
 637  * polls checking which of the two tasks completes first, and does the
 638  * associated force buffer reduction overlapped with the other task.
 639  * By doing that, unlike static scheduling order, it can always overlap
 640  * one of the reductions, regardless of the GPU task completion order.
 641  *
 642  * \param[in]     nbv              Nonbonded verlet structure
 643  * \param[in,out] pmedata          PME module data
 644  * \param[in,out] force            Force array to reduce task outputs into.
 645  * \param[in,out] forceWithVirial  Force and virial buffers
 646  * \param[in,out] fshift           Shift force output vector results are reduced into
 647  * \param[in,out] enerd            Energy data structure results are reduced into
 648  * \param[in]     flags            Force flags
 649  * \param[in]     pmeFlags         PME flags
 650  * \param[in]     haveOtherWork    Tells whether there is other work than non-bonded in the stream(s)
 651  * \param[in]     wcycle           The wallcycle structure
 652  */
 653 static void alternatePmeNbGpuWaitReduce(nonbonded_verlet_t                  *nbv,
 654                                         gmx_pme_t                           *pmedata,
 655                                         gmx::ArrayRefWithPadding<gmx::RVec> *force,
 656                                         gmx::ForceWithVirial                *forceWithVirial,
 657                                         rvec                                 fshift[],
 658                                         gmx_enerdata_t                      *enerd,
 659                                         int                                  flags,
 660                                         int                                  pmeFlags,
 661                                         bool                                 haveOtherWork,
 662                                         gmx_wallcycle_t                      wcycle)
 663 {
 664     bool isPmeGpuDone = false;
 665     bool isNbGpuDone  = false;
 666
 667
 668     gmx::ArrayRef<const gmx::RVec> pmeGpuForces;
 669
 670     while (!isPmeGpuDone || !isNbGpuDone)
 671     {
 672         if (!isPmeGpuDone)
 673         {
 674             GpuTaskCompletion completionType = (isNbGpuDone) ? GpuTaskCompletion::Wait : GpuTaskCompletion::Check;
 675             isPmeGpuDone = pme_gpu_try_finish_task(pmedata, pmeFlags, wcycle, forceWithVirial, enerd, completionType);
 676         }
 677
 678         if (!isNbGpuDone)
 679         {
 680             GpuTaskCompletion completionType = (isPmeGpuDone) ? GpuTaskCompletion::Wait : GpuTaskCompletion::Check;
 681             wallcycle_start_nocount(wcycle, ewcWAIT_GPU_NB_L);
 682             isNbGpuDone = Nbnxm::gpu_try_finish_task(nbv->gpu_nbv,
 683                                                      flags,
 684                                                      Nbnxm::AtomLocality::Local,
 685                                                      haveOtherWork,
 686                                                      enerd->grpp.ener[egLJSR], enerd->grpp.ener[egCOULSR],
 687                                                      fshift, completionType);
 688             wallcycle_stop(wcycle, ewcWAIT_GPU_NB_L);
 689             // To get the call count right, when the task finished we
 690             // issue a start/stop.
 691             // TODO: move the ewcWAIT_GPU_NB_L cycle counting into nbnxn_gpu_try_finish_task()
 692             // and ewcNB_XF_BUF_OPS counting into nbnxn_atomdata_add_nbat_f_to_f().
 693             if (isNbGpuDone)
 694             {
 695                 wallcycle_start(wcycle, ewcWAIT_GPU_NB_L);
 696                 wallcycle_stop(wcycle, ewcWAIT_GPU_NB_L);
 697
 698                 nbv->atomdata_add_nbat_f_to_f(Nbnxm::AtomLocality::Local,
 699                                               as_rvec_array(force->unpaddedArrayRef().data()), wcycle);
 700             }
 701         }
 702     }
 703 }
 704
 705 /*! \brief Hack structure with force ouput buffers for do_force */
 706 struct ForceOutputs
 707 {
 708     //! Constructor
 709     ForceOutputs(rvec *f, gmx::ForceWithVirial const forceWithVirial) :
 710         f(f),
 711         forceWithVirial(forceWithVirial) {}
 712
 713     //! Force output buffer used by legacy modules
 714     rvec                 *const f;
 715     //! Force with direct virial contribution (if there are any)
 716     gmx::ForceWithVirial        forceWithVirial;
 717 };
 718
 719 /*! \brief Set up the different force buffers; also does clearing.
 720  *
 721  * \param[in] fr        force record pointer
 722  * \param[in] inputrec  input record
 723  * \param[in] force     force array
 724  * \param[in] bDoForces True if force are computed this step
 725  * \param[in] doVirial  True if virial is computed this step
 726  * \param[out] wcycle   wallcycle recording structure
 727  *
 728  * \returns             Cleared force output structure
 729  */
 730 static ForceOutputs
 731 setupForceOutputs(const t_forcerec                    *fr,
 732                   const t_inputrec                    &inputrec,
 733                   gmx::ArrayRefWithPadding<gmx::RVec>  force,
 734                   const bool                           bDoForces,
 735                   const bool                           doVirial,
 736                   gmx_wallcycle_t                      wcycle)
 737 {
 738     wallcycle_sub_start(wcycle, ewcsCLEAR_FORCE_BUFFER);
 739
 740     /* Temporary solution until all routines take PaddedRVecVector */
 741     rvec *const f = as_rvec_array(force.unpaddedArrayRef().data());
 742     if (bDoForces)
 743     {
 744         /* Clear the short- and long-range forces */
 745         clear_rvecs_omp(fr->natoms_force_constr, f);
 746     }
 747
 748     /* If we need to compute the virial, we might need a separate
 749      * force buffer for algorithms for which the virial is calculated
 750      * directly, such as PME. Otherwise, forceWithVirial uses the
 751      * the same force (f in legacy calls) buffer as other algorithms.
 752      */
 753     const bool useSeparateForceWithVirialBuffer = (bDoForces && (doVirial && fr->haveDirectVirialContributions));
 754
 755
 756     /* forceWithVirial uses the local atom range only */
 757     gmx::ForceWithVirial forceWithVirial (useSeparateForceWithVirialBuffer ?
 758                                           *fr->forceBufferForDirectVirialContributions : force.unpaddedArrayRef(),
 759                                           doVirial);
 760
 761     if (useSeparateForceWithVirialBuffer)
 762     {
 763         /* TODO: update comment
 764          * We only compute forces on local atoms. Note that vsites can
 765          * spread to non-local atoms, but that part of the buffer is
 766          * cleared separately in the vsite spreading code.
 767          */
 768         clear_rvecs_omp(forceWithVirial.force_.size(), as_rvec_array(forceWithVirial.force_.data()));
 769     }
 770
 771     if (inputrec.bPull && pull_have_constraint(inputrec.pull_work))
 772     {
 773         clear_pull_forces(inputrec.pull_work);
 774     }
 775
 776     wallcycle_sub_stop(wcycle, ewcsCLEAR_FORCE_BUFFER);
 777
 778     return ForceOutputs(f, forceWithVirial);
 779 }
 780
 781
 782 /*! \brief Set up flags that indicate what type of work is there to compute.
 783  *
 784  * Currently we only update it at search steps,
 785  * but some properties may change more frequently (e.g. virial/non-virial step),
 786  * so when including those either the frequency of update (per-step) or the scope
 787  * of a flag will change (i.e. a set of flags for nstlist steps).
 788  *
 789  */
 790 static void
 791 setupForceWorkload(gmx::PpForceWorkload *forceWork,
 792                    const t_inputrec     *inputrec,
 793                    const t_forcerec     *fr,
 794                    const gmx_edsam      *ed,
 795                    const t_idef         &idef,
 796                    const t_fcdata       *fcd,
 797                    const int             forceFlags
 798                    )
 799 {
 800     forceWork->haveSpecialForces      = haveSpecialForces(inputrec, fr->forceProviders, forceFlags, ed);
 801     forceWork->haveCpuBondedWork      = haveCpuBondeds(*fr);
 802     forceWork->haveGpuBondedWork      = ((fr->gpuBonded != nullptr) && fr->gpuBonded->haveInteractions());
 803     forceWork->haveRestraintsWork     = havePositionRestraints(idef, *fcd);
 804     forceWork->haveCpuListedForceWork = haveCpuListedForces(*fr, idef, *fcd);
 805 }
 806
 807 static void do_force_cutsVERLET(FILE *fplog,
 808                                 const t_commrec *cr,
 809                                 const gmx_multisim_t *ms,
 810                                 const t_inputrec *inputrec,
 811                                 gmx::Awh *awh,
 812                                 gmx_enfrot *enforcedRotation,
 813                                 int64_t step,
 814                                 t_nrnb *nrnb,
 815                                 gmx_wallcycle_t wcycle,
 816                                 const gmx_localtop_t *top,
 817                                 const gmx_groups_t * /* groups */,
 818                                 matrix box, gmx::ArrayRefWithPadding<gmx::RVec> x,
 819                                 history_t *hist,
 820                                 gmx::ArrayRefWithPadding<gmx::RVec> force,
 821                                 tensor vir_force,
 822                                 const t_mdatoms *mdatoms,
 823                                 gmx_enerdata_t *enerd, t_fcdata *fcd,
 824                                 real *lambda,
 825                                 t_graph *graph,
 826                                 t_forcerec *fr,
 827                                 gmx::PpForceWorkload *ppForceWorkload,
 828                                 interaction_const_t *ic,
 829                                 const gmx_vsite_t *vsite,
 830                                 rvec mu_tot,
 831                                 double t,
 832                                 gmx_edsam *ed,
 833                                 const int flags,
 834                                 const DDBalanceRegionHandler &ddBalanceRegionHandler)
 835 {
 836     int                 cg1, i, j;
 837     double              mu[2*DIM];
 838     gmx_bool            bStateChanged, bNS, bFillGrid, bCalcCGCM;
 839     gmx_bool            bDoForces, bUseGPU, bUseOrEmulGPU;
 840     rvec                vzero, box_diag;
 841     float               cycles_pme, cycles_wait_gpu;
 842     nonbonded_verlet_t *nbv = fr->nbv.get();
 843
 844     bStateChanged = ((flags & GMX_FORCE_STATECHANGED) != 0);
 845     bNS           = ((flags & GMX_FORCE_NS) != 0);
 846     bFillGrid     = (bNS && bStateChanged);
 847     bCalcCGCM     = (bFillGrid && !DOMAINDECOMP(cr));
 848     bDoForces     = ((flags & GMX_FORCE_FORCES) != 0);
 849     bUseGPU       = fr->nbv->useGpu();
 850     bUseOrEmulGPU = bUseGPU || fr->nbv->emulateGpu();
 851
 852     const auto pmeRunMode = fr->pmedata ? pme_run_mode(fr->pmedata) : PmeRunMode::CPU;
 853     // TODO slim this conditional down - inputrec and duty checks should mean the same in proper code!
 854     const bool useGpuPme  = EEL_PME(fr->ic->eeltype) && thisRankHasDuty(cr, DUTY_PME) &&
 855         ((pmeRunMode == PmeRunMode::GPU) || (pmeRunMode == PmeRunMode::Mixed));
 856     const int  pmeFlags = GMX_PME_SPREAD | GMX_PME_SOLVE |
 857         ((flags & GMX_FORCE_VIRIAL) ? GMX_PME_CALC_ENER_VIR : 0) |
 858         ((flags & GMX_FORCE_ENERGY) ? GMX_PME_CALC_ENER_VIR : 0) |
 859         ((flags & GMX_FORCE_FORCES) ? GMX_PME_CALC_F : 0);
 860
 861     /* At a search step we need to start the first balancing region
 862      * somewhere early inside the step after communication during domain
 863      * decomposition (and not during the previous step as usual).
 864      */
 865     if (bNS)
 866     {
 867         ddBalanceRegionHandler.openBeforeForceComputationCpu(DdAllowBalanceRegionReopen::yes);
 868     }
 869
 870     cycles_wait_gpu = 0;
 871
 872     const int start  = 0;
 873     const int homenr = mdatoms->homenr;
 874
 875     clear_mat(vir_force);
 876
 877     if (DOMAINDECOMP(cr))
 878     {
 879         cg1 = cr->dd->globalAtomGroupIndices.size();
 880     }
 881     else
 882     {
 883         cg1 = top->cgs.nr;
 884     }
 885     if (fr->n_tpi > 0)
 886     {
 887         cg1--;
 888     }
 889
 890     if (bStateChanged)
 891     {
 892         update_forcerec(fr, box);
 893
 894         if (inputrecNeedMutot(inputrec))
 895         {
 896             /* Calculate total (local) dipole moment in a temporary common array.
 897              * This makes it possible to sum them over nodes faster.
 898              */
 899             calc_mu(start, homenr,
 900                     x.unpaddedArrayRef(), mdatoms->chargeA, mdatoms->chargeB, mdatoms->nChargePerturbed,
 901                     mu, mu+DIM);
 902         }
 903     }
 904
 905     if (fr->ePBC != epbcNONE)
 906     {
 907         /* Compute shift vectors every step,
 908          * because of pressure coupling or box deformation!
 909          */
 910         if ((flags & GMX_FORCE_DYNAMICBOX) && bStateChanged)
 911         {
 912             calc_shifts(box, fr->shift_vec);
 913         }
 914
 915         if (bCalcCGCM)
 916         {
 917             put_atoms_in_box_omp(fr->ePBC, box, x.unpaddedArrayRef().subArray(0, homenr));
 918             inc_nrnb(nrnb, eNR_SHIFTX, homenr);
 919         }
 920         else if (EI_ENERGY_MINIMIZATION(inputrec->eI) && graph)
 921         {
 922             unshift_self(graph, box, as_rvec_array(x.unpaddedArrayRef().data()));
 923         }
 924     }
 925
 926     nbnxn_atomdata_copy_shiftvec((flags & GMX_FORCE_DYNAMICBOX) != 0,
 927                                  fr->shift_vec, nbv->nbat.get());
 928
 929 #if GMX_MPI
 930     if (!thisRankHasDuty(cr, DUTY_PME))
 931     {
 932         /* Send particle coordinates to the pme nodes.
 933          * Since this is only implemented for domain decomposition
 934          * and domain decomposition does not use the graph,
 935          * we do not need to worry about shifting.
 936          */
 937         gmx_pme_send_coordinates(cr, box, as_rvec_array(x.unpaddedArrayRef().data()),
 938                                  lambda[efptCOUL], lambda[efptVDW],
 939                                  (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY)) != 0,
 940                                  step, wcycle);
 941     }
 942 #endif /* GMX_MPI */
 943
 944     if (useGpuPme)
 945     {
 946         launchPmeGpuSpread(fr->pmedata, box, as_rvec_array(x.unpaddedArrayRef().data()), flags, pmeFlags, wcycle);
 947     }
 948
 949     /* do gridding for pair search */
 950     if (bNS)
 951     {
 952         if (graph && bStateChanged)
 953         {
 954             /* Calculate intramolecular shift vectors to make molecules whole */
 955             mk_mshift(fplog, graph, fr->ePBC, box, as_rvec_array(x.unpaddedArrayRef().data()));
 956         }
 957
 958         clear_rvec(vzero);
 959         box_diag[XX] = box[XX][XX];
 960         box_diag[YY] = box[YY][YY];
 961         box_diag[ZZ] = box[ZZ][ZZ];
 962
 963         wallcycle_start(wcycle, ewcNS);
 964         if (!DOMAINDECOMP(cr))
 965         {
 966             wallcycle_sub_start(wcycle, ewcsNBS_GRID_LOCAL);
 967             nbnxn_put_on_grid(nbv, box,
 968                               0, vzero, box_diag,
 969                               nullptr, 0, mdatoms->homenr, -1,
 970                               fr->cginfo, x.unpaddedArrayRef(),
 971                               0, nullptr);
 972             wallcycle_sub_stop(wcycle, ewcsNBS_GRID_LOCAL);
 973         }
 974         else
 975         {
 976             wallcycle_sub_start(wcycle, ewcsNBS_GRID_NONLOCAL);
 977             nbnxn_put_on_grid_nonlocal(nbv, domdec_zones(cr->dd),
 978                                        fr->cginfo, x.unpaddedArrayRef());
 979             wallcycle_sub_stop(wcycle, ewcsNBS_GRID_NONLOCAL);
 980         }
 981
 982         nbv->setAtomProperties(*mdatoms, *fr->cginfo);
 983
 984         wallcycle_stop(wcycle, ewcNS);
 985
 986         /* initialize the GPU nbnxm atom data and bonded data structures */
 987         if (bUseGPU)
 988         {
 989             wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU);
 990
 991             wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_NONBONDED);
 992             Nbnxm::gpu_init_atomdata(nbv->gpu_nbv, nbv->nbat.get());
 993             wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
 994
 995             if (fr->gpuBonded)
 996             {
 997                 /* Now we put all atoms on the grid, we can assign bonded
 998                  * interactions to the GPU, where the grid order is
 999                  * needed. Also the xq, f and fshift device buffers have
1000                  * been reallocated if needed, so the bonded code can
1001                  * learn about them. */
1002                 // TODO the xq, f, and fshift buffers are now shared
1003                 // resources, so they should be maintained by a
1004                 // higher-level object than the nb module.
1005                 fr->gpuBonded->updateInteractionListsAndDeviceBuffers(nbv->getGridIndices(),
1006                                                                       top->idef,
1007                                                                       Nbnxm::gpu_get_xq(nbv->gpu_nbv),
1008                                                                       Nbnxm::gpu_get_f(nbv->gpu_nbv),
1009                                                                       Nbnxm::gpu_get_fshift(nbv->gpu_nbv));
1010             }
1011             wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1012         }
1013
1014         // Need to run after the GPU-offload bonded interaction lists
1015         // are set up to be able to determine whether there is bonded work.
1016         setupForceWorkload(ppForceWorkload,
1017                            inputrec,
1018                            fr,
1019                            ed,
1020                            top->idef,
1021                            fcd,
1022                            flags);
1023     }
1024
1025     /* do local pair search */
1026     if (bNS)
1027     {
1028         // TODO: fuse this branch with the above bNS block
1029         wallcycle_start_nocount(wcycle, ewcNS);
1030         wallcycle_sub_start(wcycle, ewcsNBS_SEARCH_LOCAL);
1031         /* Note that with a GPU the launch overhead of the list transfer is not timed separately */
1032         nbv->constructPairlist(Nbnxm::InteractionLocality::Local,
1033                                &top->excls, step, nrnb);
1034         wallcycle_sub_stop(wcycle, ewcsNBS_SEARCH_LOCAL);
1035         wallcycle_stop(wcycle, ewcNS);
1036     }
1037     else
1038     {
1039         nbv->setCoordinates(Nbnxm::AtomLocality::Local, false,
1040                             x.unpaddedArrayRef(), wcycle);
1041     }
1042
1043     if (bUseGPU)
1044     {
1045         ddBalanceRegionHandler.openBeforeForceComputationGpu();
1046
1047         wallcycle_start(wcycle, ewcLAUNCH_GPU);
1048
1049         wallcycle_sub_start(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1050         Nbnxm::gpu_upload_shiftvec(nbv->gpu_nbv, nbv->nbat.get());
1051         Nbnxm::gpu_copy_xq_to_gpu(nbv->gpu_nbv, nbv->nbat.get(),
1052                                   Nbnxm::AtomLocality::Local,
1053                                   ppForceWorkload->haveGpuBondedWork);
1054         wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1055
1056         // bonded work not split into separate local and non-local, so with DD
1057         // we can only launch the kernel after non-local coordinates have been received.
1058         if (ppForceWorkload->haveGpuBondedWork && !havePPDomainDecomposition(cr))
1059         {
1060             wallcycle_sub_start(wcycle, ewcsLAUNCH_GPU_BONDED);
1061             fr->gpuBonded->launchKernels(fr, flags, box);
1062             wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_BONDED);
1063         }
1064
1065         /* launch local nonbonded work on GPU */
1066         wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1067         do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::Local, enbvClearFNo,
1068                      step, nrnb, wcycle);
1069         wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1070         wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1071     }
1072
1073     if (useGpuPme)
1074     {
1075         // In PME GPU and mixed mode we launch FFT / gather after the
1076         // X copy/transform to allow overlap as well as after the GPU NB
1077         // launch to avoid FFT launch overhead hijacking the CPU and delaying
1078         // the nonbonded kernel.
1079         launchPmeGpuFftAndGather(fr->pmedata, wcycle);
1080     }
1081
1082     /* Communicate coordinates and sum dipole if necessary +
1083        do non-local pair search */
1084     if (havePPDomainDecomposition(cr))
1085     {
1086         if (bNS)
1087         {
1088             // TODO: fuse this branch with the above large bNS block
1089             wallcycle_start_nocount(wcycle, ewcNS);
1090             wallcycle_sub_start(wcycle, ewcsNBS_SEARCH_NONLOCAL);
1091             /* Note that with a GPU the launch overhead of the list transfer is not timed separately */
1092             nbv->constructPairlist(Nbnxm::InteractionLocality::NonLocal,
1093                                    &top->excls, step, nrnb);
1094             wallcycle_sub_stop(wcycle, ewcsNBS_SEARCH_NONLOCAL);
1095             wallcycle_stop(wcycle, ewcNS);
1096         }
1097         else
1098         {
1099             dd_move_x(cr->dd, box, x.unpaddedArrayRef(), wcycle);
1100
1101             nbv->setCoordinates(Nbnxm::AtomLocality::NonLocal, false,
1102                                 x.unpaddedArrayRef(), wcycle);
1103         }
1104
1105         if (bUseGPU)
1106         {
1107             wallcycle_start(wcycle, ewcLAUNCH_GPU);
1108
1109             /* launch non-local nonbonded tasks on GPU */
1110             wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1111             Nbnxm::gpu_copy_xq_to_gpu(nbv->gpu_nbv, nbv->nbat.get(),
1112                                       Nbnxm::AtomLocality::NonLocal,
1113                                       ppForceWorkload->haveGpuBondedWork);
1114             wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1115
1116             if (ppForceWorkload->haveGpuBondedWork)
1117             {
1118                 wallcycle_sub_start(wcycle, ewcsLAUNCH_GPU_BONDED);
1119                 fr->gpuBonded->launchKernels(fr, flags, box);
1120                 wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_BONDED);
1121             }
1122
1123             wallcycle_sub_start(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1124             do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::NonLocal, enbvClearFNo,
1125                          step, nrnb, wcycle);
1126             wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1127
1128             wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1129         }
1130     }
1131
1132     if (bUseGPU)
1133     {
1134         /* launch D2H copy-back F */
1135         wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU);
1136         wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1137         if (havePPDomainDecomposition(cr))
1138         {
1139             Nbnxm::gpu_launch_cpyback(nbv->gpu_nbv, nbv->nbat.get(),
1140                                       flags, Nbnxm::AtomLocality::NonLocal, ppForceWorkload->haveGpuBondedWork);
1141         }
1142         Nbnxm::gpu_launch_cpyback(nbv->gpu_nbv, nbv->nbat.get(),
1143                                   flags, Nbnxm::AtomLocality::Local, ppForceWorkload->haveGpuBondedWork);
1144         wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1145
1146         if (ppForceWorkload->haveGpuBondedWork && (flags & GMX_FORCE_ENERGY))
1147         {
1148             wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_BONDED);
1149             fr->gpuBonded->launchEnergyTransfer();
1150             wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_BONDED);
1151         }
1152         wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1153     }
1154
1155     if (bStateChanged && inputrecNeedMutot(inputrec))
1156     {
1157         if (PAR(cr))
1158         {
1159             gmx_sumd(2*DIM, mu, cr);
1160
1161             ddBalanceRegionHandler.reopenRegionCpu();
1162         }
1163
1164         for (i = 0; i < 2; i++)
1165         {
1166             for (j = 0; j < DIM; j++)
1167             {
1168                 fr->mu_tot[i][j] = mu[i*DIM + j];
1169             }
1170         }
1171     }
1172     if (fr->efep == efepNO)
1173     {
1174         copy_rvec(fr->mu_tot[0], mu_tot);
1175     }
1176     else
1177     {
1178         for (j = 0; j < DIM; j++)
1179         {
1180             mu_tot[j] =
1181                 (1.0 - lambda[efptCOUL])*fr->mu_tot[0][j] +
1182                 lambda[efptCOUL]*fr->mu_tot[1][j];
1183         }
1184     }
1185
1186     /* Reset energies */
1187     reset_enerdata(enerd);
1188     clear_rvecs(SHIFTS, fr->fshift);
1189
1190     if (DOMAINDECOMP(cr) && !thisRankHasDuty(cr, DUTY_PME))
1191     {
1192         wallcycle_start(wcycle, ewcPPDURINGPME);
1193         dd_force_flop_start(cr->dd, nrnb);
1194     }
1195
1196     if (inputrec->bRot)
1197     {
1198         wallcycle_start(wcycle, ewcROT);
1199         do_rotation(cr, enforcedRotation, box, as_rvec_array(x.unpaddedArrayRef().data()), t, step, bNS);
1200         wallcycle_stop(wcycle, ewcROT);
1201     }
1202
1203     /* Start the force cycle counter.
1204      * Note that a different counter is used for dynamic load balancing.
1205      */
1206     wallcycle_start(wcycle, ewcFORCE);
1207
1208     // set up and clear force outputs
1209     struct ForceOutputs forceOut = setupForceOutputs(fr, *inputrec, force, bDoForces, ((flags & GMX_FORCE_VIRIAL) != 0), wcycle);
1210
1211     /* We calculate the non-bonded forces, when done on the CPU, here.
1212      * We do this before calling do_force_lowlevel, because in that
1213      * function, the listed forces are calculated before PME, which
1214      * does communication.  With this order, non-bonded and listed
1215      * force calculation imbalance can be balanced out by the domain
1216      * decomposition load balancing.
1217      */
1218
1219     if (!bUseOrEmulGPU)
1220     {
1221         do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::Local, enbvClearFYes,
1222                      step, nrnb, wcycle);
1223     }
1224
1225     if (fr->efep != efepNO)
1226     {
1227         /* Calculate the local and non-local free energy interactions here.
1228          * Happens here on the CPU both with and without GPU.
1229          */
1230         wallcycle_sub_start(wcycle, ewcsNONBONDED);
1231         nbv->dispatchFreeEnergyKernel(Nbnxm::InteractionLocality::Local,
1232                                       fr, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, *mdatoms,
1233                                       inputrec->fepvals, lambda,
1234                                       enerd, flags, nrnb);
1235
1236         if (havePPDomainDecomposition(cr))
1237         {
1238             nbv->dispatchFreeEnergyKernel(Nbnxm::InteractionLocality::NonLocal,
1239                                           fr, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, *mdatoms,
1240                                           inputrec->fepvals, lambda,
1241                                           enerd, flags, nrnb);
1242         }
1243         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
1244     }
1245
1246     if (!bUseOrEmulGPU)
1247     {
1248         if (havePPDomainDecomposition(cr))
1249         {
1250             do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::NonLocal, enbvClearFNo,
1251                          step, nrnb, wcycle);
1252         }
1253
1254         /* Add all the non-bonded force to the normal force array.
1255          * This can be split into a local and a non-local part when overlapping
1256          * communication with calculation with domain decomposition.
1257          */
1258         wallcycle_stop(wcycle, ewcFORCE);
1259
1260         nbv->atomdata_add_nbat_f_to_f(Nbnxm::AtomLocality::All, forceOut.f, wcycle);
1261
1262         wallcycle_start_nocount(wcycle, ewcFORCE);
1263
1264         /* If there are multiple fshift output buffers we need to reduce them */
1265         if (flags & GMX_FORCE_VIRIAL)
1266         {
1267             /* This is not in a subcounter because it takes a
1268                negligible and constant-sized amount of time */
1269             nbnxn_atomdata_add_nbat_fshift_to_fshift(nbv->nbat.get(),
1270                                                      fr->fshift);
1271         }
1272     }
1273
1274     /* update QMMMrec, if necessary */
1275     if (fr->bQMMM)
1276     {
1277         update_QMMMrec(cr, fr, as_rvec_array(x.unpaddedArrayRef().data()), mdatoms, box);
1278     }
1279
1280     /* Compute the bonded and non-bonded energies and optionally forces */
1281     do_force_lowlevel(fr, inputrec, &(top->idef),
1282                       cr, ms, nrnb, wcycle, mdatoms,
1283                       as_rvec_array(x.unpaddedArrayRef().data()), hist, forceOut.f, &forceOut.forceWithVirial, enerd, fcd,
1284                       box, inputrec->fepvals, lambda, graph, &(top->excls), fr->mu_tot,
1285                       flags,
1286                       &cycles_pme, ddBalanceRegionHandler);
1287
1288     wallcycle_stop(wcycle, ewcFORCE);
1289
1290     computeSpecialForces(fplog, cr, inputrec, awh, enforcedRotation,
1291                          step, t, wcycle,
1292                          fr->forceProviders, box, x.unpaddedArrayRef(), mdatoms, lambda,
1293                          flags, &forceOut.forceWithVirial, enerd,
1294                          ed, bNS);
1295
1296     if (bUseOrEmulGPU)
1297     {
1298         /* wait for non-local forces (or calculate in emulation mode) */
1299         if (havePPDomainDecomposition(cr))
1300         {
1301             if (bUseGPU)
1302             {
1303                 wallcycle_start(wcycle, ewcWAIT_GPU_NB_NL);
1304                 Nbnxm::gpu_wait_finish_task(nbv->gpu_nbv,
1305                                             flags, Nbnxm::AtomLocality::NonLocal,
1306                                             ppForceWorkload->haveGpuBondedWork,
1307                                             enerd->grpp.ener[egLJSR], enerd->grpp.ener[egCOULSR],
1308                                             fr->fshift);
1309                 cycles_wait_gpu += wallcycle_stop(wcycle, ewcWAIT_GPU_NB_NL);
1310             }
1311             else
1312             {
1313                 wallcycle_start_nocount(wcycle, ewcFORCE);
1314                 do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::NonLocal, enbvClearFYes,
1315                              step, nrnb, wcycle);
1316                 wallcycle_stop(wcycle, ewcFORCE);
1317             }
1318
1319             nbv->atomdata_add_nbat_f_to_f(Nbnxm::AtomLocality::NonLocal,
1320                                           forceOut.f, wcycle);
1321         }
1322     }
1323
1324     if (havePPDomainDecomposition(cr))
1325     {
1326         /* We are done with the CPU compute.
1327          * We will now communicate the non-local forces.
1328          * If we use a GPU this will overlap with GPU work, so in that case
1329          * we do not close the DD force balancing region here.
1330          */
1331         ddBalanceRegionHandler.closeAfterForceComputationCpu();
1332
1333         if (bDoForces)
1334         {
1335             dd_move_f(cr->dd, force.unpaddedArrayRef(), fr->fshift, wcycle);
1336         }
1337     }
1338
1339     // With both nonbonded and PME offloaded a GPU on the same rank, we use
1340     // an alternating wait/reduction scheme.
1341     bool alternateGpuWait = (!c_disableAlternatingWait && useGpuPme && bUseGPU && !DOMAINDECOMP(cr));
1342     if (alternateGpuWait)
1343     {
1344         alternatePmeNbGpuWaitReduce(fr->nbv.get(), fr->pmedata, &force, &forceOut.forceWithVirial, fr->fshift, enerd,
1345                                     flags, pmeFlags, ppForceWorkload->haveGpuBondedWork, wcycle);
1346     }
1347
1348     if (!alternateGpuWait && useGpuPme)
1349     {
1350         pme_gpu_wait_and_reduce(fr->pmedata, pmeFlags, wcycle, &forceOut.forceWithVirial, enerd);
1351     }
1352
1353     /* Wait for local GPU NB outputs on the non-alternating wait path */
1354     if (!alternateGpuWait && bUseGPU)
1355     {
1356         /* Measured overhead on CUDA and OpenCL with(out) GPU sharing
1357          * is between 0.5 and 1.5 Mcycles. So 2 MCycles is an overestimate,
1358          * but even with a step of 0.1 ms the difference is less than 1%
1359          * of the step time.
1360          */
1361         const float gpuWaitApiOverheadMargin = 2e6f; /* cycles */
1362
1363         wallcycle_start(wcycle, ewcWAIT_GPU_NB_L);
1364         Nbnxm::gpu_wait_finish_task(nbv->gpu_nbv,
1365                                     flags, Nbnxm::AtomLocality::Local, ppForceWorkload->haveGpuBondedWork,
1366                                     enerd->grpp.ener[egLJSR], enerd->grpp.ener[egCOULSR],
1367                                     fr->fshift);
1368         float cycles_tmp = wallcycle_stop(wcycle, ewcWAIT_GPU_NB_L);
1369
1370         if (ddBalanceRegionHandler.useBalancingRegion())
1371         {
1372             DdBalanceRegionWaitedForGpu waitedForGpu = DdBalanceRegionWaitedForGpu::yes;
1373             if (bDoForces && cycles_tmp <= gpuWaitApiOverheadMargin)
1374             {
1375                 /* We measured few cycles, it could be that the kernel
1376                  * and transfer finished earlier and there was no actual
1377                  * wait time, only API call overhead.
1378                  * Then the actual time could be anywhere between 0 and
1379                  * cycles_wait_est. We will use half of cycles_wait_est.
1380                  */
1381                 waitedForGpu = DdBalanceRegionWaitedForGpu::no;
1382             }
1383             ddBalanceRegionHandler.closeAfterForceComputationGpu(cycles_wait_gpu, waitedForGpu);
1384         }
1385     }
1386
1387     if (fr->nbv->emulateGpu())
1388     {
1389         // NOTE: emulation kernel is not included in the balancing region,
1390         // but emulation mode does not target performance anyway
1391         wallcycle_start_nocount(wcycle, ewcFORCE);
1392         do_nb_verlet(fr, ic, enerd, flags, Nbnxm::InteractionLocality::Local,
1393                      DOMAINDECOMP(cr) ? enbvClearFNo : enbvClearFYes,
1394                      step, nrnb, wcycle);
1395         wallcycle_stop(wcycle, ewcFORCE);
1396     }
1397
1398     if (useGpuPme)
1399     {
1400         pme_gpu_reinit_computation(fr->pmedata, wcycle);
1401     }
1402
1403     if (bUseGPU)
1404     {
1405         /* now clear the GPU outputs while we finish the step on the CPU */
1406         wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU);
1407         wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1408         Nbnxm::gpu_clear_outputs(nbv->gpu_nbv, flags);
1409
1410         if (nbv->pairlistSets().isDynamicPruningStepGpu(step))
1411         {
1412             nbv->dispatchPruneKernelGpu(step);
1413         }
1414         wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_NONBONDED);
1415         wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1416     }
1417
1418     if (ppForceWorkload->haveGpuBondedWork && (flags & GMX_FORCE_ENERGY))
1419     {
1420         wallcycle_start(wcycle, ewcWAIT_GPU_BONDED);
1421         // in principle this should be included in the DD balancing region,
1422         // but generally it is infrequent so we'll omit it for the sake of
1423         // simpler code
1424         fr->gpuBonded->accumulateEnergyTerms(enerd);
1425         wallcycle_stop(wcycle, ewcWAIT_GPU_BONDED);
1426
1427         wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU);
1428         wallcycle_sub_start_nocount(wcycle, ewcsLAUNCH_GPU_BONDED);
1429         fr->gpuBonded->clearEnergies();
1430         wallcycle_sub_stop(wcycle, ewcsLAUNCH_GPU_BONDED);
1431         wallcycle_stop(wcycle, ewcLAUNCH_GPU);
1432     }
1433
1434     /* Do the nonbonded GPU (or emulation) force buffer reduction
1435      * on the non-alternating path. */
1436     if (bUseOrEmulGPU && !alternateGpuWait)
1437     {
1438         nbv->atomdata_add_nbat_f_to_f(Nbnxm::AtomLocality::Local,
1439                                       forceOut.f, wcycle);
1440     }
1441     if (DOMAINDECOMP(cr))
1442     {
1443         dd_force_flop_stop(cr->dd, nrnb);
1444     }
1445
1446     if (bDoForces)
1447     {
1448         /* If we have NoVirSum forces, but we do not calculate the virial,
1449          * we sum fr->f_novirsum=forceOut.f later.
1450          */
1451         if (vsite && !(fr->haveDirectVirialContributions && !(flags & GMX_FORCE_VIRIAL)))
1452         {
1453             spread_vsite_f(vsite, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, fr->fshift, FALSE, nullptr, nrnb,
1454                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr, wcycle);
1455         }
1456
1457         if (flags & GMX_FORCE_VIRIAL)
1458         {
1459             /* Calculation of the virial must be done after vsites! */
1460             calc_virial(0, mdatoms->homenr, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f,
1461                         vir_force, graph, box, nrnb, fr, inputrec->ePBC);
1462         }
1463     }
1464
1465     if (PAR(cr) && !thisRankHasDuty(cr, DUTY_PME))
1466     {
1467         /* In case of node-splitting, the PP nodes receive the long-range
1468          * forces, virial and energy from the PME nodes here.
1469          */
1470         pme_receive_force_ener(cr, &forceOut.forceWithVirial, enerd, wcycle);
1471     }
1472
1473     if (bDoForces)
1474     {
1475         post_process_forces(cr, step, nrnb, wcycle,
1476                             top, box, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, &forceOut.forceWithVirial,
1477                             vir_force, mdatoms, graph, fr, vsite,
1478                             flags);
1479     }
1480
1481     if (flags & GMX_FORCE_ENERGY)
1482     {
1483         /* Sum the potential energy terms from group contributions */
1484         sum_epot(&(enerd->grpp), enerd->term);
1485
1486         if (!EI_TPI(inputrec->eI))
1487         {
1488             checkPotentialEnergyValidity(step, *enerd, *inputrec);
1489         }
1490     }
1491 }
1492
1493 static void do_force_cutsGROUP(FILE *fplog,
1494                                const t_commrec *cr,
1495                                const gmx_multisim_t *ms,
1496                                const t_inputrec *inputrec,
1497                                gmx::Awh *awh,
1498                                gmx_enfrot *enforcedRotation,
1499                                int64_t step,
1500                                t_nrnb *nrnb,
1501                                gmx_wallcycle_t wcycle,
1502                                gmx_localtop_t *top,
1503                                const gmx_groups_t *groups,
1504                                matrix box, gmx::ArrayRefWithPadding<gmx::RVec> x,
1505                                history_t *hist,
1506                                gmx::ArrayRefWithPadding<gmx::RVec> force,
1507                                tensor vir_force,
1508                                const t_mdatoms *mdatoms,
1509                                gmx_enerdata_t *enerd,
1510                                t_fcdata *fcd,
1511                                real *lambda,
1512                                t_graph *graph,
1513                                t_forcerec *fr,
1514                                const gmx_vsite_t *vsite,
1515                                rvec mu_tot,
1516                                double t,
1517                                gmx_edsam *ed,
1518                                int flags,
1519                                const DDBalanceRegionHandler &ddBalanceRegionHandler)
1520 {
1521     int        cg0, cg1, i, j;
1522     double     mu[2*DIM];
1523     gmx_bool   bStateChanged, bNS, bFillGrid, bCalcCGCM;
1524     gmx_bool   bDoForces;
1525     float      cycles_pme;
1526
1527     const int  start  = 0;
1528     const int  homenr = mdatoms->homenr;
1529
1530     clear_mat(vir_force);
1531
1532     cg0 = 0;
1533     if (DOMAINDECOMP(cr))
1534     {
1535         cg1 = cr->dd->globalAtomGroupIndices.size();
1536     }
1537     else
1538     {
1539         cg1 = top->cgs.nr;
1540     }
1541     if (fr->n_tpi > 0)
1542     {
1543         cg1--;
1544     }
1545
1546     bStateChanged  = ((flags & GMX_FORCE_STATECHANGED) != 0);
1547     bNS            = ((flags & GMX_FORCE_NS) != 0);
1548     /* Should we perform the long-range nonbonded evaluation inside the neighborsearching? */
1549     bFillGrid      = (bNS && bStateChanged);
1550     bCalcCGCM      = (bFillGrid && !DOMAINDECOMP(cr));
1551     bDoForces      = ((flags & GMX_FORCE_FORCES) != 0);
1552
1553     if (bStateChanged)
1554     {
1555         update_forcerec(fr, box);
1556
1557         if (inputrecNeedMutot(inputrec))
1558         {
1559             /* Calculate total (local) dipole moment in a temporary common array.
1560              * This makes it possible to sum them over nodes faster.
1561              */
1562             calc_mu(start, homenr,
1563                     x.unpaddedArrayRef(), mdatoms->chargeA, mdatoms->chargeB, mdatoms->nChargePerturbed,
1564                     mu, mu+DIM);
1565         }
1566     }
1567
1568     if (fr->ePBC != epbcNONE)
1569     {
1570         /* Compute shift vectors every step,
1571          * because of pressure coupling or box deformation!
1572          */
1573         if ((flags & GMX_FORCE_DYNAMICBOX) && bStateChanged)
1574         {
1575             calc_shifts(box, fr->shift_vec);
1576         }
1577
1578         if (bCalcCGCM)
1579         {
1580             put_charge_groups_in_box(fplog, cg0, cg1, fr->ePBC, box,
1581                                      &(top->cgs), as_rvec_array(x.unpaddedArrayRef().data()), fr->cg_cm);
1582             inc_nrnb(nrnb, eNR_CGCM, homenr);
1583             inc_nrnb(nrnb, eNR_RESETX, cg1-cg0);
1584         }
1585         else if (EI_ENERGY_MINIMIZATION(inputrec->eI) && graph)
1586         {
1587             unshift_self(graph, box, as_rvec_array(x.unpaddedArrayRef().data()));
1588         }
1589     }
1590     else if (bCalcCGCM)
1591     {
1592         calc_cgcm(fplog, cg0, cg1, &(top->cgs), as_rvec_array(x.unpaddedArrayRef().data()), fr->cg_cm);
1593         inc_nrnb(nrnb, eNR_CGCM, homenr);
1594     }
1595
1596     if (bCalcCGCM && gmx_debug_at)
1597     {
1598         pr_rvecs(debug, 0, "cgcm", fr->cg_cm, top->cgs.nr);
1599     }
1600
1601 #if GMX_MPI
1602     if (!thisRankHasDuty(cr, DUTY_PME))
1603     {
1604         /* Send particle coordinates to the pme nodes.
1605          * Since this is only implemented for domain decomposition
1606          * and domain decomposition does not use the graph,
1607          * we do not need to worry about shifting.
1608          */
1609         gmx_pme_send_coordinates(cr, box, as_rvec_array(x.unpaddedArrayRef().data()),
1610                                  lambda[efptCOUL], lambda[efptVDW],
1611                                  (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY)) != 0,
1612                                  step, wcycle);
1613     }
1614 #endif /* GMX_MPI */
1615
1616     /* Communicate coordinates and sum dipole if necessary */
1617     if (DOMAINDECOMP(cr))
1618     {
1619         dd_move_x(cr->dd, box, x.unpaddedArrayRef(), wcycle);
1620
1621         /* No GPU support, no move_x overlap, so reopen the balance region here */
1622         ddBalanceRegionHandler.reopenRegionCpu();
1623     }
1624
1625     if (inputrecNeedMutot(inputrec))
1626     {
1627         if (bStateChanged)
1628         {
1629             if (PAR(cr))
1630             {
1631                 gmx_sumd(2*DIM, mu, cr);
1632
1633                 ddBalanceRegionHandler.reopenRegionCpu();
1634             }
1635             for (i = 0; i < 2; i++)
1636             {
1637                 for (j = 0; j < DIM; j++)
1638                 {
1639                     fr->mu_tot[i][j] = mu[i*DIM + j];
1640                 }
1641             }
1642         }
1643         if (fr->efep == efepNO)
1644         {
1645             copy_rvec(fr->mu_tot[0], mu_tot);
1646         }
1647         else
1648         {
1649             for (j = 0; j < DIM; j++)
1650             {
1651                 mu_tot[j] =
1652                     (1.0 - lambda[efptCOUL])*fr->mu_tot[0][j] + lambda[efptCOUL]*fr->mu_tot[1][j];
1653             }
1654         }
1655     }
1656
1657     /* Reset energies */
1658     reset_enerdata(enerd);
1659     clear_rvecs(SHIFTS, fr->fshift);
1660
1661     if (bNS)
1662     {
1663         wallcycle_start(wcycle, ewcNS);
1664
1665         if (graph && bStateChanged)
1666         {
1667             /* Calculate intramolecular shift vectors to make molecules whole */
1668             mk_mshift(fplog, graph, fr->ePBC, box, as_rvec_array(x.unpaddedArrayRef().data()));
1669         }
1670
1671         /* Do the actual neighbour searching */
1672         ns(fplog, fr, box,
1673            groups, top, mdatoms,
1674            cr, nrnb, bFillGrid);
1675
1676         wallcycle_stop(wcycle, ewcNS);
1677     }
1678
1679     if (DOMAINDECOMP(cr) && !thisRankHasDuty(cr, DUTY_PME))
1680     {
1681         wallcycle_start(wcycle, ewcPPDURINGPME);
1682         dd_force_flop_start(cr->dd, nrnb);
1683     }
1684
1685     if (inputrec->bRot)
1686     {
1687         wallcycle_start(wcycle, ewcROT);
1688         do_rotation(cr, enforcedRotation, box, as_rvec_array(x.unpaddedArrayRef().data()), t, step, bNS);
1689         wallcycle_stop(wcycle, ewcROT);
1690     }
1691
1692
1693     /* Start the force cycle counter.
1694      * Note that a different counter is used for dynamic load balancing.
1695      */
1696     wallcycle_start(wcycle, ewcFORCE);
1697
1698     // set up and clear force outputs
1699     struct ForceOutputs forceOut = setupForceOutputs(fr, *inputrec, force, bDoForces, ((flags & GMX_FORCE_VIRIAL) != 0), wcycle);
1700
1701     if (inputrec->bPull && pull_have_constraint(inputrec->pull_work))
1702     {
1703         clear_pull_forces(inputrec->pull_work);
1704     }
1705
1706     /* update QMMMrec, if necessary */
1707     if (fr->bQMMM)
1708     {
1709         update_QMMMrec(cr, fr, as_rvec_array(x.unpaddedArrayRef().data()), mdatoms, box);
1710     }
1711
1712     /* Compute the bonded and non-bonded energies and optionally forces */
1713     do_force_lowlevel(fr, inputrec, &(top->idef),
1714                       cr, ms, nrnb, wcycle, mdatoms,
1715                       as_rvec_array(x.unpaddedArrayRef().data()), hist, forceOut.f, &forceOut.forceWithVirial, enerd, fcd,
1716                       box, inputrec->fepvals, lambda,
1717                       graph, &(top->excls), fr->mu_tot,
1718                       flags,
1719                       &cycles_pme, ddBalanceRegionHandler);
1720
1721     wallcycle_stop(wcycle, ewcFORCE);
1722
1723     if (DOMAINDECOMP(cr))
1724     {
1725         dd_force_flop_stop(cr->dd, nrnb);
1726
1727         ddBalanceRegionHandler.closeAfterForceComputationCpu();
1728     }
1729
1730     computeSpecialForces(fplog, cr, inputrec, awh, enforcedRotation,
1731                          step, t, wcycle,
1732                          fr->forceProviders, box, x.unpaddedArrayRef(), mdatoms, lambda,
1733                          flags, &forceOut.forceWithVirial, enerd,
1734                          ed, bNS);
1735
1736     if (bDoForces)
1737     {
1738         /* Communicate the forces */
1739         if (DOMAINDECOMP(cr))
1740         {
1741             dd_move_f(cr->dd, force.unpaddedArrayRef(), fr->fshift, wcycle);
1742             /* Do we need to communicate the separate force array
1743              * for terms that do not contribute to the single sum virial?
1744              * Position restraints and electric fields do not introduce
1745              * inter-cg forces, only full electrostatics methods do.
1746              * When we do not calculate the virial, fr->f_novirsum = forceOut.f,
1747              * so we have already communicated these forces.
1748              */
1749             if (EEL_FULL(fr->ic->eeltype) && cr->dd->n_intercg_excl &&
1750                 (flags & GMX_FORCE_VIRIAL))
1751             {
1752                 dd_move_f(cr->dd, forceOut.forceWithVirial.force_, nullptr, wcycle);
1753             }
1754         }
1755
1756         /* If we have NoVirSum forces, but we do not calculate the virial,
1757          * we sum fr->f_novirsum=forceOut.f later.
1758          */
1759         if (vsite && !(fr->haveDirectVirialContributions && !(flags & GMX_FORCE_VIRIAL)))
1760         {
1761             spread_vsite_f(vsite, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, fr->fshift, FALSE, nullptr, nrnb,
1762                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr, wcycle);
1763         }
1764
1765         if (flags & GMX_FORCE_VIRIAL)
1766         {
1767             /* Calculation of the virial must be done after vsites! */
1768             calc_virial(0, mdatoms->homenr, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f,
1769                         vir_force, graph, box, nrnb, fr, inputrec->ePBC);
1770         }
1771     }
1772
1773     if (PAR(cr) && !thisRankHasDuty(cr, DUTY_PME))
1774     {
1775         /* In case of node-splitting, the PP nodes receive the long-range
1776          * forces, virial and energy from the PME nodes here.
1777          */
1778         pme_receive_force_ener(cr, &forceOut.forceWithVirial, enerd, wcycle);
1779     }
1780
1781     if (bDoForces)
1782     {
1783         post_process_forces(cr, step, nrnb, wcycle,
1784                             top, box, as_rvec_array(x.unpaddedArrayRef().data()), forceOut.f, &forceOut.forceWithVirial,
1785                             vir_force, mdatoms, graph, fr, vsite,
1786                             flags);
1787     }
1788
1789     if (flags & GMX_FORCE_ENERGY)
1790     {
1791         /* Sum the potential energy terms from group contributions */
1792         sum_epot(&(enerd->grpp), enerd->term);
1793
1794         if (!EI_TPI(inputrec->eI))
1795         {
1796             checkPotentialEnergyValidity(step, *enerd, *inputrec);
1797         }
1798     }
1799
1800 }
1801
1802 void do_force(FILE                                     *fplog,
1803               const t_commrec                          *cr,
1804               const gmx_multisim_t                     *ms,
1805               const t_inputrec                         *inputrec,
1806               gmx::Awh                                 *awh,
1807               gmx_enfrot                               *enforcedRotation,
1808               int64_t                                   step,
1809               t_nrnb                                   *nrnb,
1810               gmx_wallcycle_t                           wcycle,
1811               gmx_localtop_t                           *top,
1812               const gmx_groups_t                       *groups,
1813               matrix                                    box,
1814               gmx::ArrayRefWithPadding<gmx::RVec>       x,     //NOLINT(performance-unnecessary-value-param)
1815               history_t                                *hist,
1816               gmx::ArrayRefWithPadding<gmx::RVec>       force, //NOLINT(performance-unnecessary-value-param)
1817               tensor                                    vir_force,
1818               const t_mdatoms                          *mdatoms,
1819               gmx_enerdata_t                           *enerd,
1820               t_fcdata                                 *fcd,
1821               gmx::ArrayRef<real>                       lambda,
1822               t_graph                                  *graph,
1823               t_forcerec                               *fr,
1824               gmx::PpForceWorkload                     *ppForceWorkload,
1825               const gmx_vsite_t                        *vsite,
1826               rvec                                      mu_tot,
1827               double                                    t,
1828               gmx_edsam                                *ed,
1829               int                                       flags,
1830               const DDBalanceRegionHandler             &ddBalanceRegionHandler)
1831 {
1832     /* modify force flag if not doing nonbonded */
1833     if (!fr->bNonbonded)
1834     {
1835         flags &= ~GMX_FORCE_NONBONDED;
1836     }
1837
1838     switch (inputrec->cutoff_scheme)
1839     {
1840         case ecutsVERLET:
1841             do_force_cutsVERLET(fplog, cr, ms, inputrec,
1842                                 awh, enforcedRotation, step, nrnb, wcycle,
1843                                 top,
1844                                 groups,
1845                                 box, x, hist,
1846                                 force, vir_force,
1847                                 mdatoms,
1848                                 enerd, fcd,
1849                                 lambda.data(), graph,
1850                                 fr,
1851                                 ppForceWorkload,
1852                                 fr->ic,
1853                                 vsite, mu_tot,
1854                                 t, ed,
1855                                 flags,
1856                                 ddBalanceRegionHandler);
1857             break;
1858         case ecutsGROUP:
1859             do_force_cutsGROUP(fplog, cr, ms, inputrec,
1860                                awh, enforcedRotation, step, nrnb, wcycle,
1861                                top,
1862                                groups,
1863                                box, x, hist,
1864                                force, vir_force,
1865                                mdatoms,
1866                                enerd, fcd,
1867                                lambda.data(), graph,
1868                                fr, vsite, mu_tot,
1869                                t, ed,
1870                                flags,
1871                                ddBalanceRegionHandler);
1872             break;
1873         default:
1874             gmx_incons("Invalid cut-off scheme passed!");
1875     }
1876
1877     /* In case we don't have constraints and are using GPUs, the next balancing
1878      * region starts here.
1879      * Some "special" work at the end of do_force_cuts?, such as vsite spread,
1880      * virial calculation and COM pulling, is not thus not included in
1881      * the balance timing, which is ok as most tasks do communication.
1882      */
1883     ddBalanceRegionHandler.openBeforeForceComputationCpu(DdAllowBalanceRegionReopen::no);
1884 }
1885
1886
1887 void do_constrain_first(FILE *fplog, gmx::Constraints *constr,
1888                         const t_inputrec *ir, const t_mdatoms *md,
1889                         t_state *state)
1890 {
1891     int             i, m, start, end;
1892     int64_t         step;
1893     real            dt = ir->delta_t;
1894     real            dvdl_dum;
1895     rvec           *savex;
1896
1897     /* We need to allocate one element extra, since we might use
1898      * (unaligned) 4-wide SIMD loads to access rvec entries.
1899      */
1900     snew(savex, state->natoms + 1);
1901
1902     start = 0;
1903     end   = md->homenr;
1904
1905     if (debug)
1906     {
1907         fprintf(debug, "vcm: start=%d, homenr=%d, end=%d\n",
1908                 start, md->homenr, end);
1909     }
1910     /* Do a first constrain to reset particles... */
1911     step = ir->init_step;
1912     if (fplog)
1913     {
1914         char buf[STEPSTRSIZE];
1915         fprintf(fplog, "\nConstraining the starting coordinates (step %s)\n",
1916                 gmx_step_str(step, buf));
1917     }
1918     dvdl_dum = 0;
1919
1920     /* constrain the current position */
1921     constr->apply(TRUE, FALSE,
1922                   step, 0, 1.0,
1923                   state->x.rvec_array(), state->x.rvec_array(), nullptr,
1924                   state->box,
1925                   state->lambda[efptBONDED], &dvdl_dum,
1926                   nullptr, nullptr, gmx::ConstraintVariable::Positions);
1927     if (EI_VV(ir->eI))
1928     {
1929         /* constrain the inital velocity, and save it */
1930         /* also may be useful if we need the ekin from the halfstep for velocity verlet */
1931         constr->apply(TRUE, FALSE,
1932                       step, 0, 1.0,
1933                       state->x.rvec_array(), state->v.rvec_array(), state->v.rvec_array(),
1934                       state->box,
1935                       state->lambda[efptBONDED], &dvdl_dum,
1936                       nullptr, nullptr, gmx::ConstraintVariable::Velocities);
1937     }
1938     /* constrain the inital velocities at t-dt/2 */
1939     if (EI_STATE_VELOCITY(ir->eI) && ir->eI != eiVV)
1940     {
1941         auto x = makeArrayRef(state->x).subArray(start, end);
1942         auto v = makeArrayRef(state->v).subArray(start, end);
1943         for (i = start; (i < end); i++)
1944         {
1945             for (m = 0; (m < DIM); m++)
1946             {
1947                 /* Reverse the velocity */
1948                 v[i][m] = -v[i][m];
1949                 /* Store the position at t-dt in buf */
1950                 savex[i][m] = x[i][m] + dt*v[i][m];
1951             }
1952         }
1953         /* Shake the positions at t=-dt with the positions at t=0
1954          * as reference coordinates.
1955          */
1956         if (fplog)
1957         {
1958             char buf[STEPSTRSIZE];
1959             fprintf(fplog, "\nConstraining the coordinates at t0-dt (step %s)\n",
1960                     gmx_step_str(step, buf));
1961         }
1962         dvdl_dum = 0;
1963         constr->apply(TRUE, FALSE,
1964                       step, -1, 1.0,
1965                       state->x.rvec_array(), savex, nullptr,
1966                       state->box,
1967                       state->lambda[efptBONDED], &dvdl_dum,
1968                       state->v.rvec_array(), nullptr, gmx::ConstraintVariable::Positions);
1969
1970         for (i = start; i < end; i++)
1971         {
1972             for (m = 0; m < DIM; m++)
1973             {
1974                 /* Re-reverse the velocities */
1975                 v[i][m] = -v[i][m];
1976             }
1977         }
1978     }
1979     sfree(savex);
1980 }
1981
1982 void put_atoms_in_box_omp(int ePBC, const matrix box, gmx::ArrayRef<gmx::RVec> x)
1983 {
1984     int t, nth;
1985     nth = gmx_omp_nthreads_get(emntDefault);
1986
1987 #pragma omp parallel for num_threads(nth) schedule(static)
1988     for (t = 0; t < nth; t++)
1989     {
1990         try
1991         {
1992             size_t natoms = x.size();
1993             size_t offset = (natoms*t    )/nth;
1994             size_t len    = (natoms*(t + 1))/nth - offset;
1995             put_atoms_in_box(ePBC, box, x.subArray(offset, len));
1996         }
1997         GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
1998     }
1999 }
2000
2001 void initialize_lambdas(FILE               *fplog,
2002                         const t_inputrec   &ir,
2003                         bool                isMaster,
2004                         int                *fep_state,
2005                         gmx::ArrayRef<real> lambda,
2006                         double             *lam0)
2007 {
2008     /* TODO: Clean up initialization of fep_state and lambda in
2009        t_state.  This function works, but could probably use a logic
2010        rewrite to keep all the different types of efep straight. */
2011
2012     if ((ir.efep == efepNO) && (!ir.bSimTemp))
2013     {
2014         return;
2015     }
2016
2017     const t_lambda *fep = ir.fepvals;
2018     if (isMaster)
2019     {
2020         *fep_state = fep->init_fep_state; /* this might overwrite the checkpoint
2021                                              if checkpoint is set -- a kludge is in for now
2022                                              to prevent this.*/
2023     }
2024
2025     for (int i = 0; i < efptNR; i++)
2026     {
2027         double thisLambda;
2028         /* overwrite lambda state with init_lambda for now for backwards compatibility */
2029         if (fep->init_lambda >= 0) /* if it's -1, it was never initialized */
2030         {
2031             thisLambda = fep->init_lambda;
2032         }
2033         else
2034         {
2035             thisLambda = fep->all_lambda[i][fep->init_fep_state];
2036         }
2037         if (isMaster)
2038         {
2039             lambda[i] = thisLambda;
2040         }
2041         if (lam0 != nullptr)
2042         {
2043             lam0[i] = thisLambda;
2044         }
2045     }
2046     if (ir.bSimTemp)
2047     {
2048         /* need to rescale control temperatures to match current state */
2049         for (int i = 0; i < ir.opts.ngtc; i++)
2050         {
2051             if (ir.opts.ref_t[i] > 0)
2052             {
2053                 ir.opts.ref_t[i] = ir.simtempvals->temperatures[fep->init_fep_state];
2054             }
2055         }
2056     }
2057
2058     /* Send to the log the information on the current lambdas */
2059     if (fplog != nullptr)
2060     {
2061         fprintf(fplog, "Initial vector of lambda components:[ ");
2062         for (const auto &l : lambda)
2063         {
2064             fprintf(fplog, "%10.4f ", l);
2065         }
2066         fprintf(fplog, "]\n");
2067     }
2068 }