src/gromacs/mdlib/force.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 by the GROMACS development team.
   7  * Copyright (c) 2018,2019,2020,2021, by the GROMACS development team, led by
   8  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   9  * and including many others, as listed in the AUTHORS file in the
  10  * top-level source directory and at http://www.gromacs.org.
  11  *
  12  * GROMACS is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public License
  14  * as published by the Free Software Foundation; either version 2.1
  15  * of the License, or (at your option) any later version.
  16  *
  17  * GROMACS is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with GROMACS; if not, see
  24  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  25  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  26  *
  27  * If you want to redistribute modifications to GROMACS, please
  28  * consider that scientific software is very special. Version
  29  * control is crucial - bugs must be traceable. We will be happy to
  30  * consider code for inclusion in the official distribution, but
  31  * derived work must not be called official GROMACS. Details are found
  32  * in the README & COPYING files - if they are missing, get the
  33  * official version at http://www.gromacs.org.
  34  *
  35  * To help us fund GROMACS development, we humbly ask that you cite
  36  * the research papers on the package. Check out http://www.gromacs.org.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "force.h"
  41
  42 #include <cassert>
  43 #include <cmath>
  44 #include <cstring>
  45
  46 #include "gromacs/domdec/dlbtiming.h"
  47 #include "gromacs/domdec/domdec.h"
  48 #include "gromacs/domdec/domdec_struct.h"
  49 #include "gromacs/ewald/ewald.h"
  50 #include "gromacs/ewald/long_range_correction.h"
  51 #include "gromacs/ewald/pme.h"
  52 #include "gromacs/gmxlib/network.h"
  53 #include "gromacs/gmxlib/nrnb.h"
  54 #include "gromacs/math/vec.h"
  55 #include "gromacs/math/vecdump.h"
  56 #include "gromacs/mdlib/gmx_omp_nthreads.h"
  57 #include "gromacs/mdtypes/commrec.h"
  58 #include "gromacs/mdtypes/enerdata.h"
  59 #include "gromacs/mdtypes/forceoutput.h"
  60 #include "gromacs/mdtypes/forcerec.h"
  61 #include "gromacs/mdtypes/inputrec.h"
  62 #include "gromacs/mdtypes/interaction_const.h"
  63 #include "gromacs/mdtypes/md_enums.h"
  64 #include "gromacs/mdtypes/mdatom.h"
  65 #include "gromacs/mdtypes/simulation_workload.h"
  66 #include "gromacs/pbcutil/pbc.h"
  67 #include "gromacs/timing/wallcycle.h"
  68 #include "gromacs/utility/exceptions.h"
  69 #include "gromacs/utility/fatalerror.h"
  70
  71 using gmx::ArrayRef;
  72 using gmx::RVec;
  73
  74 static void clearEwaldThreadOutput(ewald_corr_thread_t* ewc_t)
  75 {
  76     ewc_t->Vcorr_q                                        = 0;
  77     ewc_t->Vcorr_lj                                       = 0;
  78     ewc_t->dvdl[FreeEnergyPerturbationCouplingType::Coul] = 0;
  79     ewc_t->dvdl[FreeEnergyPerturbationCouplingType::Vdw]  = 0;
  80     clear_mat(ewc_t->vir_q);
  81     clear_mat(ewc_t->vir_lj);
  82 }
  83
  84 static void reduceEwaldThreadOuput(int nthreads, gmx::ArrayRef<ewald_corr_thread_t> ewc_t)
  85 {
  86     ewald_corr_thread_t& dest = ewc_t[0];
  87
  88     for (int t = 1; t < nthreads; t++)
  89     {
  90         dest.Vcorr_q += ewc_t[t].Vcorr_q;
  91         dest.Vcorr_lj += ewc_t[t].Vcorr_lj;
  92         dest.dvdl[FreeEnergyPerturbationCouplingType::Coul] +=
  93                 ewc_t[t].dvdl[FreeEnergyPerturbationCouplingType::Coul];
  94         dest.dvdl[FreeEnergyPerturbationCouplingType::Vdw] +=
  95                 ewc_t[t].dvdl[FreeEnergyPerturbationCouplingType::Vdw];
  96         m_add(dest.vir_q, ewc_t[t].vir_q, dest.vir_q);
  97         m_add(dest.vir_lj, ewc_t[t].vir_lj, dest.vir_lj);
  98     }
  99 }
 100
 101 CpuPpLongRangeNonbondeds::CpuPpLongRangeNonbondeds(int                         numberOfTestPaticles,
 102                                                    real                        ewaldCoeffQ,
 103                                                    real                        epsilonR,
 104                                                    gmx::ArrayRef<const double> chargeC6Sum,
 105                                                    CoulombInteractionType      eeltype,
 106                                                    VanDerWaalsType             vdwtype,
 107                                                    const t_inputrec&           inputrec,
 108                                                    t_nrnb*                     nrnb,
 109                                                    gmx_wallcycle*              wcycle,
 110                                                    FILE*                       fplog) :
 111     numTpiAtoms_(numberOfTestPaticles),
 112     ewaldCoeffQ_(ewaldCoeffQ),
 113     epsilonR_(epsilonR),
 114     chargeC6Sum_(chargeC6Sum),
 115     coulombInteractionType_(eeltype),
 116     vanDerWaalsType_(vdwtype),
 117     ewaldGeometry_(inputrec.ewald_geometry),
 118     epsilonSurface_(inputrec.epsilon_surface),
 119     haveEwaldSurfaceTerm_(haveEwaldSurfaceContribution(inputrec)),
 120     wallEwaldZfac_(inputrec.wall_ewald_zfac),
 121     havePbcXY2Walls_(inputrecPbcXY2Walls(&inputrec)),
 122     freeEnergyPerturbationType_(inputrec.efep),
 123     nrnb_(nrnb),
 124     wcycle_(wcycle)
 125 {
 126     GMX_ASSERT(epsilonR == inputrec.epsilon_r,
 127                "Forcerec and inputrec relative dielectrics don't match");
 128
 129     // Use the thread count for the bonded module since reducing CPU-side
 130     // non-bonded contributions does not currently have its own thread count.
 131     outputPerThread_.resize(gmx_omp_nthreads_get(ModuleMultiThread::Bonded));
 132
 133     if (inputrec.coulombtype == CoulombInteractionType::Ewald)
 134     {
 135         ewaldTable_ = std::make_unique<gmx_ewald_tab_t>(inputrec, fplog);
 136     }
 137 }
 138
 139 CpuPpLongRangeNonbondeds::~CpuPpLongRangeNonbondeds() = default;
 140
 141 void CpuPpLongRangeNonbondeds::updateAfterPartition(const t_mdatoms& md)
 142 {
 143     homenr_        = md.homenr;
 144     havePerturbed_ = md.nChargePerturbed != 0;
 145     chargeA_ = md.chargeA ? gmx::constArrayRefFromArray(md.chargeA, md.nr) : gmx::ArrayRef<const real>{};
 146     chargeB_ = md.chargeB ? gmx::constArrayRefFromArray(md.chargeB, md.nr) : gmx::ArrayRef<const real>{};
 147     sqrt_c6A_ = md.sqrt_c6A ? gmx::constArrayRefFromArray(md.sqrt_c6A, md.nr)
 148                             : gmx::ArrayRef<const real>{};
 149     sqrt_c6B_ = md.sqrt_c6B ? gmx::constArrayRefFromArray(md.sqrt_c6B, md.nr)
 150                             : gmx::ArrayRef<const real>{};
 151     sigmaA_ = md.sigmaA ? gmx::constArrayRefFromArray(md.sigmaA, md.nr) : gmx::ArrayRef<const real>{};
 152     sigmaB_ = md.sigmaB ? gmx::constArrayRefFromArray(md.sigmaB, md.nr) : gmx::ArrayRef<const real>{};
 153 }
 154
 155 void CpuPpLongRangeNonbondeds::calculate(gmx_pme_t*                     pmedata,
 156                                          const t_commrec*               commrec,
 157                                          gmx::ArrayRef<const RVec>      coordinates,
 158                                          gmx::ForceWithVirial*          forceWithVirial,
 159                                          gmx_enerdata_t*                enerd,
 160                                          const matrix                   box,
 161                                          gmx::ArrayRef<const real>      lambda,
 162                                          gmx::ArrayRef<const gmx::RVec> mu_tot,
 163                                          const gmx::StepWorkload&       stepWork,
 164                                          const DDBalanceRegionHandler&  ddBalanceRegionHandler)
 165 {
 166     const bool computePmeOnCpu = (EEL_PME(coulombInteractionType_) || EVDW_PME(vanDerWaalsType_))
 167                                  && thisRankHasDuty(commrec, DUTY_PME)
 168                                  && (pme_run_mode(pmedata) == PmeRunMode::CPU);
 169
 170     /* Do long-range electrostatics and/or LJ-PME
 171      * and compute PME surface terms when necessary.
 172      */
 173     if ((computePmeOnCpu || coulombInteractionType_ == CoulombInteractionType::Ewald || haveEwaldSurfaceTerm_)
 174         && stepWork.computeNonbondedForces)
 175     {
 176         real Vlr_q = 0, Vlr_lj = 0;
 177         /* We reduce all virial, dV/dlambda and energy contributions, except
 178          * for the reciprocal energies (Vlr_q, Vlr_lj) into the same struct.
 179          */
 180         ewald_corr_thread_t& ewaldOutput = outputPerThread_[0];
 181         clearEwaldThreadOutput(&ewaldOutput);
 182
 183         if (EEL_PME_EWALD(coulombInteractionType_) || EVDW_PME(vanDerWaalsType_))
 184         {
 185             /* Calculate the Ewald surface force and energy contributions, when necessary */
 186             if (haveEwaldSurfaceTerm_)
 187             {
 188                 wallcycle_sub_start(wcycle_, WallCycleSubCounter::EwaldCorrection);
 189
 190                 int nthreads = gmx::ssize(outputPerThread_);
 191 #pragma omp parallel for num_threads(nthreads) schedule(static)
 192                 for (int t = 0; t < nthreads; t++)
 193                 {
 194                     try
 195                     {
 196                         ewald_corr_thread_t& ewc_t = outputPerThread_[t];
 197                         if (t > 0)
 198                         {
 199                             clearEwaldThreadOutput(&ewc_t);
 200                         }
 201
 202                         /* Threading is only supported with the Verlet cut-off
 203                          * scheme and then only single particle forces (no
 204                          * exclusion forces) are calculated, so we can store
 205                          * the forces in the normal, single forceWithVirial->force_ array.
 206                          */
 207                         ewald_LRcorrection(
 208                                 homenr_,
 209                                 commrec,
 210                                 nthreads,
 211                                 t,
 212                                 epsilonR_,
 213                                 chargeC6Sum_,
 214                                 ewaldGeometry_,
 215                                 epsilonSurface_,
 216                                 havePbcXY2Walls_,
 217                                 wallEwaldZfac_,
 218                                 chargeA_,
 219                                 chargeB_,
 220                                 havePerturbed_,
 221                                 coordinates,
 222                                 box,
 223                                 mu_tot,
 224                                 forceWithVirial->force_,
 225                                 &ewc_t.Vcorr_q,
 226                                 lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Coul)],
 227                                 &ewc_t.dvdl[FreeEnergyPerturbationCouplingType::Coul]);
 228                     }
 229                     GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
 230                 }
 231                 if (nthreads > 1)
 232                 {
 233                     reduceEwaldThreadOuput(nthreads, outputPerThread_);
 234                 }
 235                 wallcycle_sub_stop(wcycle_, WallCycleSubCounter::EwaldCorrection);
 236             }
 237
 238             if (EEL_PME_EWALD(coulombInteractionType_) && numTpiAtoms_ == 0)
 239             {
 240                 /* This is not in a subcounter because it takes a
 241                    negligible and constant-sized amount of time */
 242                 ewaldOutput.Vcorr_q += ewald_charge_correction(
 243                         commrec,
 244                         epsilonR_,
 245                         ewaldCoeffQ_,
 246                         chargeC6Sum_,
 247                         lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Coul)],
 248                         box,
 249                         &ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Coul],
 250                         ewaldOutput.vir_q);
 251             }
 252
 253             if (computePmeOnCpu)
 254             {
 255                 /* Do reciprocal PME for Coulomb and/or LJ. */
 256                 assert(numTpiAtoms_ >= 0);
 257                 if (numTpiAtoms_ == 0 || stepWork.stateChanged)
 258                 {
 259                     /* With domain decomposition we close the CPU side load
 260                      * balancing region here, because PME does global
 261                      * communication that acts as a global barrier.
 262                      */
 263                     ddBalanceRegionHandler.closeAfterForceComputationCpu();
 264
 265                     wallcycle_start(wcycle_, WallCycleCounter::PmeMesh);
 266                     int status = gmx_pme_do(
 267                             pmedata,
 268                             gmx::constArrayRefFromArray(coordinates.data(), homenr_ - numTpiAtoms_),
 269                             forceWithVirial->force_,
 270                             chargeA_,
 271                             chargeB_,
 272                             sqrt_c6A_,
 273                             sqrt_c6B_,
 274                             sigmaA_,
 275                             sigmaB_,
 276                             box,
 277                             commrec,
 278                             haveDDAtomOrdering(*commrec) ? dd_pme_maxshift_x(*commrec->dd) : 0,
 279                             haveDDAtomOrdering(*commrec) ? dd_pme_maxshift_y(*commrec->dd) : 0,
 280                             nrnb_,
 281                             wcycle_,
 282                             ewaldOutput.vir_q,
 283                             ewaldOutput.vir_lj,
 284                             &Vlr_q,
 285                             &Vlr_lj,
 286                             lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Coul)],
 287                             lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Vdw)],
 288                             &ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Coul],
 289                             &ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Vdw],
 290                             stepWork);
 291                     wallcycle_stop(wcycle_, WallCycleCounter::PmeMesh);
 292                     if (status != 0)
 293                     {
 294                         gmx_fatal(FARGS, "Error %d in reciprocal PME routine", status);
 295                     }
 296
 297                     /* We should try to do as little computation after
 298                      * this as possible, because parallel PME synchronizes
 299                      * the nodes, so we want all load imbalance of the
 300                      * rest of the force calculation to be before the PME
 301                      * call.  DD load balancing is done on the whole time
 302                      * of the force call (without PME).
 303                      */
 304                 }
 305                 if (numTpiAtoms_ > 0)
 306                 {
 307                     /* Determine the PME grid energy of the test molecule
 308                      * with the PME grid potential of the other charges.
 309                      */
 310                     Vlr_q = gmx_pme_calc_energy(
 311                             pmedata,
 312                             coordinates.subArray(homenr_ - numTpiAtoms_, numTpiAtoms_),
 313                             chargeA_.subArray(homenr_ - numTpiAtoms_, numTpiAtoms_));
 314                 }
 315             }
 316         }
 317
 318         if (coulombInteractionType_ == CoulombInteractionType::Ewald)
 319         {
 320             Vlr_q = do_ewald(havePbcXY2Walls_,
 321                              wallEwaldZfac_,
 322                              epsilonR_,
 323                              freeEnergyPerturbationType_,
 324                              coordinates,
 325                              forceWithVirial->force_,
 326                              chargeA_,
 327                              chargeB_,
 328                              box,
 329                              commrec,
 330                              homenr_,
 331                              ewaldOutput.vir_q,
 332                              ewaldCoeffQ_,
 333                              lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Coul)],
 334                              &ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Coul],
 335                              ewaldTable_.get());
 336         }
 337
 338         /* Note that with separate PME nodes we get the real energies later */
 339         // TODO it would be simpler if we just accumulated a single
 340         // long-range virial contribution.
 341         forceWithVirial->addVirialContribution(ewaldOutput.vir_q);
 342         forceWithVirial->addVirialContribution(ewaldOutput.vir_lj);
 343         enerd->dvdl_lin[FreeEnergyPerturbationCouplingType::Coul] +=
 344                 ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Coul];
 345         enerd->dvdl_lin[FreeEnergyPerturbationCouplingType::Vdw] +=
 346                 ewaldOutput.dvdl[FreeEnergyPerturbationCouplingType::Vdw];
 347         enerd->term[F_COUL_RECIP] = Vlr_q + ewaldOutput.Vcorr_q;
 348         enerd->term[F_LJ_RECIP]   = Vlr_lj + ewaldOutput.Vcorr_lj;
 349
 350         if (debug)
 351         {
 352             fprintf(debug,
 353                     "Vlr_q = %g, Vcorr_q = %g, Vlr_corr_q = %g\n",
 354                     Vlr_q,
 355                     ewaldOutput.Vcorr_q,
 356                     enerd->term[F_COUL_RECIP]);
 357             pr_rvecs(debug, 0, "vir_el_recip after corr", ewaldOutput.vir_q, DIM);
 358             fprintf(debug,
 359                     "Vlr_lj: %g, Vcorr_lj = %g, Vlr_corr_lj = %g\n",
 360                     Vlr_lj,
 361                     ewaldOutput.Vcorr_lj,
 362                     enerd->term[F_LJ_RECIP]);
 363             pr_rvecs(debug, 0, "vir_lj_recip after corr", ewaldOutput.vir_lj, DIM);
 364         }
 365     }
 366
 367     if (debug)
 368     {
 369         print_nrnb(debug, nrnb_);
 370     }
 371 }