src/gromacs/mdlib/force.c

   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, 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 "config.h"
  38
  39 #include <assert.h>
  40 #include <math.h>
  41 #include <string.h>
  42
  43 #include "gromacs/legacyheaders/typedefs.h"
  44 #include "gromacs/legacyheaders/macros.h"
  45 #include "gromacs/legacyheaders/force.h"
  46 #include "gromacs/legacyheaders/nonbonded.h"
  47 #include "gromacs/legacyheaders/names.h"
  48 #include "gromacs/legacyheaders/network.h"
  49 #include "gromacs/legacyheaders/ns.h"
  50 #include "gromacs/legacyheaders/nrnb.h"
  51 #include "gromacs/legacyheaders/bondf.h"
  52 #include "gromacs/legacyheaders/txtdump.h"
  53 #include "gromacs/legacyheaders/coulomb.h"
  54 #include "gromacs/legacyheaders/pme.h"
  55 #include "gromacs/legacyheaders/mdrun.h"
  56 #include "gromacs/legacyheaders/domdec.h"
  57 #include "gromacs/legacyheaders/qmmm.h"
  58 #include "gromacs/legacyheaders/gmx_omp_nthreads.h"
  59
  60 #include "gromacs/legacyheaders/types/commrec.h"
  61 #include "gromacs/math/vec.h"
  62 #include "gromacs/pbcutil/ishift.h"
  63 #include "gromacs/pbcutil/mshift.h"
  64 #include "gromacs/pbcutil/pbc.h"
  65 #include "gromacs/timing/wallcycle.h"
  66 #include "gromacs/utility/fatalerror.h"
  67 #include "gromacs/utility/smalloc.h"
  68
  69 void ns(FILE              *fp,
  70         t_forcerec        *fr,
  71         matrix             box,
  72         gmx_groups_t      *groups,
  73         gmx_localtop_t    *top,
  74         t_mdatoms         *md,
  75         t_commrec         *cr,
  76         t_nrnb            *nrnb,
  77         gmx_bool           bFillGrid,
  78         gmx_bool           bDoLongRangeNS)
  79 {
  80     char   *ptr;
  81     int     nsearch;
  82
  83
  84     if (!fr->ns.nblist_initialized)
  85     {
  86         init_neighbor_list(fp, fr, md->homenr);
  87     }
  88
  89     if (fr->bTwinRange)
  90     {
  91         fr->nlr = 0;
  92     }
  93
  94     nsearch = search_neighbours(fp, fr, box, top, groups, cr, nrnb, md,
  95                                 bFillGrid, bDoLongRangeNS);
  96     if (debug)
  97     {
  98         fprintf(debug, "nsearch = %d\n", nsearch);
  99     }
 100
 101     /* Check whether we have to do dynamic load balancing */
 102     /*if ((nsb->nstDlb > 0) && (mod(step,nsb->nstDlb) == 0))
 103        count_nb(cr,nsb,&(top->blocks[ebCGS]),nns,fr->nlr,
 104        &(top->idef),opts->ngener);
 105      */
 106     if (fr->ns.dump_nl > 0)
 107     {
 108         dump_nblist(fp, cr, fr, fr->ns.dump_nl);
 109     }
 110 }
 111
 112 static void reduce_thread_forces(int n, rvec *f,
 113                                  tensor vir_q, tensor vir_lj,
 114                                  real *Vcorr_q, real *Vcorr_lj,
 115                                  real *dvdl_q, real *dvdl_lj,
 116                                  int nthreads, f_thread_t *f_t)
 117 {
 118     int t, i;
 119
 120     /* This reduction can run over any number of threads */
 121 #pragma omp parallel for num_threads(gmx_omp_nthreads_get(emntBonded)) private(t) schedule(static)
 122     for (i = 0; i < n; i++)
 123     {
 124         for (t = 1; t < nthreads; t++)
 125         {
 126             rvec_inc(f[i], f_t[t].f[i]);
 127         }
 128     }
 129     for (t = 1; t < nthreads; t++)
 130     {
 131         *Vcorr_q  += f_t[t].Vcorr_q;
 132         *Vcorr_lj += f_t[t].Vcorr_lj;
 133         *dvdl_q   += f_t[t].dvdl[efptCOUL];
 134         *dvdl_lj  += f_t[t].dvdl[efptVDW];
 135         m_add(vir_q, f_t[t].vir_q, vir_q);
 136         m_add(vir_lj, f_t[t].vir_lj, vir_lj);
 137     }
 138 }
 139
 140 void do_force_lowlevel(t_forcerec *fr,      t_inputrec *ir,
 141                        t_idef     *idef,    t_commrec  *cr,
 142                        t_nrnb     *nrnb,    gmx_wallcycle_t wcycle,
 143                        t_mdatoms  *md,
 144                        rvec       x[],      history_t  *hist,
 145                        rvec       f[],
 146                        rvec       f_longrange[],
 147                        gmx_enerdata_t *enerd,
 148                        t_fcdata   *fcd,
 149                        gmx_localtop_t *top,
 150                        gmx_genborn_t *born,
 151                        t_atomtypes *atype,
 152                        gmx_bool       bBornRadii,
 153                        matrix     box,
 154                        t_lambda   *fepvals,
 155                        real       *lambda,
 156                        t_graph    *graph,
 157                        t_blocka   *excl,
 158                        rvec       mu_tot[],
 159                        int        flags,
 160                        float      *cycles_pme)
 161 {
 162     int         i, j;
 163     int         donb_flags;
 164     gmx_bool    bDoEpot, bSB;
 165     int         pme_flags;
 166     matrix      boxs;
 167     rvec        box_size;
 168     t_pbc       pbc;
 169     char        buf[22];
 170     double      clam_i, vlam_i;
 171     real        dvdl_dum[efptNR], dvdl_nb[efptNR], lam_i[efptNR];
 172     real        dvdl_q, dvdl_lj;
 173
 174 #ifdef GMX_MPI
 175     double  t0 = 0.0, t1, t2, t3; /* time measurement for coarse load balancing */
 176 #endif
 177
 178     set_pbc(&pbc, fr->ePBC, box);
 179
 180     /* reset free energy components */
 181     for (i = 0; i < efptNR; i++)
 182     {
 183         dvdl_nb[i]  = 0;
 184         dvdl_dum[i] = 0;
 185     }
 186
 187     /* Reset box */
 188     for (i = 0; (i < DIM); i++)
 189     {
 190         box_size[i] = box[i][i];
 191     }
 192
 193     debug_gmx();
 194
 195     /* do QMMM first if requested */
 196     if (fr->bQMMM)
 197     {
 198         enerd->term[F_EQM] = calculate_QMMM(cr, x, f, fr);
 199     }
 200
 201     /* Call the short range functions all in one go. */
 202
 203 #ifdef GMX_MPI
 204     /*#define TAKETIME ((cr->npmenodes) && (fr->timesteps < 12))*/
 205 #define TAKETIME FALSE
 206     if (TAKETIME)
 207     {
 208         MPI_Barrier(cr->mpi_comm_mygroup);
 209         t0 = MPI_Wtime();
 210     }
 211 #endif
 212
 213     if (ir->nwall)
 214     {
 215         /* foreign lambda component for walls */
 216         real dvdl_walls = do_walls(ir, fr, box, md, x, f, lambda[efptVDW],
 217                                    enerd->grpp.ener[egLJSR], nrnb);
 218         enerd->dvdl_lin[efptVDW] += dvdl_walls;
 219     }
 220
 221     /* If doing GB, reset dvda and calculate the Born radii */
 222     if (ir->implicit_solvent)
 223     {
 224         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 225
 226         for (i = 0; i < born->nr; i++)
 227         {
 228             fr->dvda[i] = 0;
 229         }
 230
 231         if (bBornRadii)
 232         {
 233             calc_gb_rad(cr, fr, ir, top, x, &(fr->gblist), born, md, nrnb);
 234         }
 235
 236         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 237     }
 238
 239     where();
 240     /* We only do non-bonded calculation with group scheme here, the verlet
 241      * calls are done from do_force_cutsVERLET(). */
 242     if (fr->cutoff_scheme == ecutsGROUP && (flags & GMX_FORCE_NONBONDED))
 243     {
 244         donb_flags = 0;
 245         /* Add short-range interactions */
 246         donb_flags |= GMX_NONBONDED_DO_SR;
 247
 248         /* Currently all group scheme kernels always calculate (shift-)forces */
 249         if (flags & GMX_FORCE_FORCES)
 250         {
 251             donb_flags |= GMX_NONBONDED_DO_FORCE;
 252         }
 253         if (flags & GMX_FORCE_VIRIAL)
 254         {
 255             donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 256         }
 257         if (flags & GMX_FORCE_ENERGY)
 258         {
 259             donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 260         }
 261         if (flags & GMX_FORCE_DO_LR)
 262         {
 263             donb_flags |= GMX_NONBONDED_DO_LR;
 264         }
 265
 266         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 267         do_nonbonded(fr, x, f, f_longrange, md, excl,
 268                      &enerd->grpp, nrnb,
 269                      lambda, dvdl_nb, -1, -1, donb_flags);
 270
 271         /* If we do foreign lambda and we have soft-core interactions
 272          * we have to recalculate the (non-linear) energies contributions.
 273          */
 274         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 275         {
 276             for (i = 0; i < enerd->n_lambda; i++)
 277             {
 278                 for (j = 0; j < efptNR; j++)
 279                 {
 280                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 281                 }
 282                 reset_foreign_enerdata(enerd);
 283                 do_nonbonded(fr, x, f, f_longrange, md, excl,
 284                              &(enerd->foreign_grpp), nrnb,
 285                              lam_i, dvdl_dum, -1, -1,
 286                              (donb_flags & ~GMX_NONBONDED_DO_FORCE) | GMX_NONBONDED_DO_FOREIGNLAMBDA);
 287                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 288                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 289             }
 290         }
 291         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 292         where();
 293     }
 294
 295     /* If we are doing GB, calculate bonded forces and apply corrections
 296      * to the solvation forces */
 297     /* MRS: Eventually, many need to include free energy contribution here! */
 298     if (ir->implicit_solvent)
 299     {
 300         wallcycle_sub_start(wcycle, ewcsBONDED);
 301         calc_gb_forces(cr, md, born, top, x, f, fr, idef,
 302                        ir->gb_algorithm, ir->sa_algorithm, nrnb, &pbc, graph, enerd);
 303         wallcycle_sub_stop(wcycle, ewcsBONDED);
 304     }
 305
 306 #ifdef GMX_MPI
 307     if (TAKETIME)
 308     {
 309         t1          = MPI_Wtime();
 310         fr->t_fnbf += t1-t0;
 311     }
 312 #endif
 313
 314     if (fepvals->sc_alpha != 0)
 315     {
 316         enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
 317     }
 318     else
 319     {
 320         enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
 321     }
 322
 323     if (fepvals->sc_alpha != 0)
 324
 325     /* even though coulomb part is linear, we already added it, beacuse we
 326        need to go through the vdw calculation anyway */
 327     {
 328         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 329     }
 330     else
 331     {
 332         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 333     }
 334
 335     debug_gmx();
 336
 337
 338     if (debug)
 339     {
 340         pr_rvecs(debug, 0, "fshift after SR", fr->fshift, SHIFTS);
 341     }
 342
 343     /* Shift the coordinates. Must be done before bonded forces and PPPM,
 344      * but is also necessary for SHAKE and update, therefore it can NOT
 345      * go when no bonded forces have to be evaluated.
 346      */
 347
 348     /* Here sometimes we would not need to shift with NBFonly,
 349      * but we do so anyhow for consistency of the returned coordinates.
 350      */
 351     if (graph)
 352     {
 353         shift_self(graph, box, x);
 354         if (TRICLINIC(box))
 355         {
 356             inc_nrnb(nrnb, eNR_SHIFTX, 2*graph->nnodes);
 357         }
 358         else
 359         {
 360             inc_nrnb(nrnb, eNR_SHIFTX, graph->nnodes);
 361         }
 362     }
 363     /* Check whether we need to do bondeds or correct for exclusions */
 364     if (fr->bMolPBC &&
 365         ((flags & GMX_FORCE_BONDED)
 366          || EEL_RF(fr->eeltype) || EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype)))
 367     {
 368         /* Since all atoms are in the rectangular or triclinic unit-cell,
 369          * only single box vector shifts (2 in x) are required.
 370          */
 371         set_pbc_dd(&pbc, fr->ePBC, cr->dd, TRUE, box);
 372     }
 373     debug_gmx();
 374
 375     if (flags & GMX_FORCE_BONDED)
 376     {
 377         wallcycle_sub_start(wcycle, ewcsBONDED);
 378         calc_bonds(cr->ms,
 379                    idef, x, hist, f, fr, &pbc, graph, enerd, nrnb, lambda, md, fcd,
 380                    DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL, atype, born,
 381                    flags);
 382
 383         /* Check if we have to determine energy differences
 384          * at foreign lambda's.
 385          */
 386         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) &&
 387             idef->ilsort != ilsortNO_FE)
 388         {
 389             if (idef->ilsort != ilsortFE_SORTED)
 390             {
 391                 gmx_incons("The bonded interactions are not sorted for free energy");
 392             }
 393             for (i = 0; i < enerd->n_lambda; i++)
 394             {
 395                 reset_foreign_enerdata(enerd);
 396                 for (j = 0; j < efptNR; j++)
 397                 {
 398                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 399                 }
 400                 calc_bonds_lambda(idef, x, fr, &pbc, graph, &(enerd->foreign_grpp), enerd->foreign_term, nrnb, lam_i, md,
 401                                   fcd, DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL);
 402                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 403                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 404             }
 405         }
 406         debug_gmx();
 407
 408         wallcycle_sub_stop(wcycle, ewcsBONDED);
 409     }
 410
 411     where();
 412
 413     *cycles_pme = 0;
 414     clear_mat(fr->vir_el_recip);
 415     clear_mat(fr->vir_lj_recip);
 416
 417     /* Do long-range electrostatics and/or LJ-PME, including related short-range
 418      * corrections.
 419      */
 420     if (EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype))
 421     {
 422         real Vlr               = 0, Vcorr = 0;
 423         real dvdl_long_range   = 0;
 424         int  status            = 0;
 425         real Vlr_q             = 0, Vlr_lj = 0, Vcorr_q = 0, Vcorr_lj = 0;
 426         real dvdl_long_range_q = 0, dvdl_long_range_lj = 0;
 427
 428         bSB = (ir->nwall == 2);
 429         if (bSB)
 430         {
 431             copy_mat(box, boxs);
 432             svmul(ir->wall_ewald_zfac, boxs[ZZ], boxs[ZZ]);
 433             box_size[ZZ] *= ir->wall_ewald_zfac;
 434         }
 435
 436         if (EEL_PME_EWALD(fr->eeltype) || EVDW_PME(fr->vdwtype))
 437         {
 438             real dvdl_long_range_correction_q   = 0;
 439             real dvdl_long_range_correction_lj  = 0;
 440             /* With the Verlet scheme exclusion forces are calculated
 441              * in the non-bonded kernel.
 442              */
 443             /* The TPI molecule does not have exclusions with the rest
 444              * of the system and no intra-molecular PME grid
 445              * contributions will be calculated in
 446              * gmx_pme_calc_energy.
 447              */
 448             if ((ir->cutoff_scheme == ecutsGROUP && fr->n_tpi == 0) ||
 449                 ir->ewald_geometry != eewg3D ||
 450                 ir->epsilon_surface != 0)
 451             {
 452                 int nthreads, t;
 453
 454                 wallcycle_sub_start(wcycle, ewcsEWALD_CORRECTION);
 455
 456                 if (fr->n_tpi > 0)
 457                 {
 458                     gmx_fatal(FARGS, "TPI with PME currently only works in a 3D geometry with tin-foil boundary conditions");
 459                 }
 460
 461                 nthreads = gmx_omp_nthreads_get(emntBonded);
 462 #pragma omp parallel for num_threads(nthreads) schedule(static)
 463                 for (t = 0; t < nthreads; t++)
 464                 {
 465                     int     s, e, i;
 466                     rvec   *fnv;
 467                     tensor *vir_q, *vir_lj;
 468                     real   *Vcorrt_q, *Vcorrt_lj, *dvdlt_q, *dvdlt_lj;
 469                     if (t == 0)
 470                     {
 471                         fnv       = fr->f_novirsum;
 472                         vir_q     = &fr->vir_el_recip;
 473                         vir_lj    = &fr->vir_lj_recip;
 474                         Vcorrt_q  = &Vcorr_q;
 475                         Vcorrt_lj = &Vcorr_lj;
 476                         dvdlt_q   = &dvdl_long_range_correction_q;
 477                         dvdlt_lj  = &dvdl_long_range_correction_lj;
 478                     }
 479                     else
 480                     {
 481                         fnv       = fr->f_t[t].f;
 482                         vir_q     = &fr->f_t[t].vir_q;
 483                         vir_lj    = &fr->f_t[t].vir_lj;
 484                         Vcorrt_q  = &fr->f_t[t].Vcorr_q;
 485                         Vcorrt_lj = &fr->f_t[t].Vcorr_lj;
 486                         dvdlt_q   = &fr->f_t[t].dvdl[efptCOUL];
 487                         dvdlt_lj  = &fr->f_t[t].dvdl[efptVDW];
 488                         for (i = 0; i < fr->natoms_force; i++)
 489                         {
 490                             clear_rvec(fnv[i]);
 491                         }
 492                         clear_mat(*vir_q);
 493                         clear_mat(*vir_lj);
 494                     }
 495                     *dvdlt_q  = 0;
 496                     *dvdlt_lj = 0;
 497
 498                     ewald_LRcorrection(fr->excl_load[t], fr->excl_load[t+1],
 499                                        cr, t, fr,
 500                                        md->chargeA,
 501                                        md->nChargePerturbed ? md->chargeB : NULL,
 502                                        md->sqrt_c6A,
 503                                        md->nTypePerturbed ? md->sqrt_c6B : NULL,
 504                                        md->sigmaA,
 505                                        md->nTypePerturbed ? md->sigmaB : NULL,
 506                                        md->sigma3A,
 507                                        md->nTypePerturbed ? md->sigma3B : NULL,
 508                                        ir->cutoff_scheme != ecutsVERLET,
 509                                        excl, x, bSB ? boxs : box, mu_tot,
 510                                        ir->ewald_geometry,
 511                                        ir->epsilon_surface,
 512                                        fnv, *vir_q, *vir_lj,
 513                                        Vcorrt_q, Vcorrt_lj,
 514                                        lambda[efptCOUL], lambda[efptVDW],
 515                                        dvdlt_q, dvdlt_lj);
 516                 }
 517                 if (nthreads > 1)
 518                 {
 519                     reduce_thread_forces(fr->natoms_force, fr->f_novirsum,
 520                                          fr->vir_el_recip, fr->vir_lj_recip,
 521                                          &Vcorr_q, &Vcorr_lj,
 522                                          &dvdl_long_range_correction_q,
 523                                          &dvdl_long_range_correction_lj,
 524                                          nthreads, fr->f_t);
 525                 }
 526                 wallcycle_sub_stop(wcycle, ewcsEWALD_CORRECTION);
 527             }
 528
 529             if (EEL_PME_EWALD(fr->eeltype) && fr->n_tpi == 0)
 530             {
 531                 Vcorr_q += ewald_charge_correction(cr, fr, lambda[efptCOUL], box,
 532                                                    &dvdl_long_range_correction_q,
 533                                                    fr->vir_el_recip);
 534             }
 535
 536             enerd->dvdl_lin[efptCOUL] += dvdl_long_range_correction_q;
 537             enerd->dvdl_lin[efptVDW]  += dvdl_long_range_correction_lj;
 538
 539             if ((EEL_PME(fr->eeltype) || EVDW_PME(fr->vdwtype)) && (cr->duty & DUTY_PME))
 540             {
 541                 /* Do reciprocal PME for Coulomb and/or LJ. */
 542                 assert(fr->n_tpi >= 0);
 543                 if (fr->n_tpi == 0 || (flags & GMX_FORCE_STATECHANGED))
 544                 {
 545                     pme_flags = GMX_PME_SPREAD | GMX_PME_SOLVE;
 546                     if (EEL_PME(fr->eeltype))
 547                     {
 548                         pme_flags     |= GMX_PME_DO_COULOMB;
 549                     }
 550                     if (EVDW_PME(fr->vdwtype))
 551                     {
 552                         pme_flags |= GMX_PME_DO_LJ;
 553                     }
 554                     if (flags & GMX_FORCE_FORCES)
 555                     {
 556                         pme_flags |= GMX_PME_CALC_F;
 557                     }
 558                     if (flags & GMX_FORCE_VIRIAL)
 559                     {
 560                         pme_flags |= GMX_PME_CALC_ENER_VIR;
 561                     }
 562                     if (fr->n_tpi > 0)
 563                     {
 564                         /* We don't calculate f, but we do want the potential */
 565                         pme_flags |= GMX_PME_CALC_POT;
 566                     }
 567                     wallcycle_start(wcycle, ewcPMEMESH);
 568                     status = gmx_pme_do(fr->pmedata,
 569                                         0, md->homenr - fr->n_tpi,
 570                                         x, fr->f_novirsum,
 571                                         md->chargeA, md->chargeB,
 572                                         md->sqrt_c6A, md->sqrt_c6B,
 573                                         md->sigmaA, md->sigmaB,
 574                                         bSB ? boxs : box, cr,
 575                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_x(cr->dd) : 0,
 576                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_y(cr->dd) : 0,
 577                                         nrnb, wcycle,
 578                                         fr->vir_el_recip, fr->ewaldcoeff_q,
 579                                         fr->vir_lj_recip, fr->ewaldcoeff_lj,
 580                                         &Vlr_q, &Vlr_lj,
 581                                         lambda[efptCOUL], lambda[efptVDW],
 582                                         &dvdl_long_range_q, &dvdl_long_range_lj, pme_flags);
 583                     *cycles_pme = wallcycle_stop(wcycle, ewcPMEMESH);
 584                     if (status != 0)
 585                     {
 586                         gmx_fatal(FARGS, "Error %d in reciprocal PME routine", status);
 587                     }
 588                     /* We should try to do as little computation after
 589                      * this as possible, because parallel PME synchronizes
 590                      * the nodes, so we want all load imbalance of the
 591                      * rest of the force calculation to be before the PME
 592                      * call.  DD load balancing is done on the whole time
 593                      * of the force call (without PME).
 594                      */
 595                 }
 596                 if (fr->n_tpi > 0)
 597                 {
 598                     if (EVDW_PME(ir->vdwtype))
 599                     {
 600
 601                         gmx_fatal(FARGS, "Test particle insertion not implemented with LJ-PME");
 602                     }
 603                     /* Determine the PME grid energy of the test molecule
 604                      * with the PME grid potential of the other charges.
 605                      */
 606                     gmx_pme_calc_energy(fr->pmedata, fr->n_tpi,
 607                                         x + md->homenr - fr->n_tpi,
 608                                         md->chargeA + md->homenr - fr->n_tpi,
 609                                         &Vlr_q);
 610                 }
 611             }
 612         }
 613
 614         if (!EEL_PME(fr->eeltype) && EEL_PME_EWALD(fr->eeltype))
 615         {
 616             Vlr_q = do_ewald(ir, x, fr->f_novirsum,
 617                              md->chargeA, md->chargeB,
 618                              box_size, cr, md->homenr,
 619                              fr->vir_el_recip, fr->ewaldcoeff_q,
 620                              lambda[efptCOUL], &dvdl_long_range_q, fr->ewald_table);
 621         }
 622
 623         /* Note that with separate PME nodes we get the real energies later */
 624         enerd->dvdl_lin[efptCOUL] += dvdl_long_range_q;
 625         enerd->dvdl_lin[efptVDW]  += dvdl_long_range_lj;
 626         enerd->term[F_COUL_RECIP]  = Vlr_q + Vcorr_q;
 627         enerd->term[F_LJ_RECIP]    = Vlr_lj + Vcorr_lj;
 628         if (debug)
 629         {
 630             fprintf(debug, "Vlr_q = %g, Vcorr_q = %g, Vlr_corr_q = %g\n",
 631                     Vlr_q, Vcorr_q, enerd->term[F_COUL_RECIP]);
 632             pr_rvecs(debug, 0, "vir_el_recip after corr", fr->vir_el_recip, DIM);
 633             pr_rvecs(debug, 0, "fshift after LR Corrections", fr->fshift, SHIFTS);
 634             fprintf(debug, "Vlr_lj: %g, Vcorr_lj = %g, Vlr_corr_lj = %g\n",
 635                     Vlr_lj, Vcorr_lj, enerd->term[F_LJ_RECIP]);
 636             pr_rvecs(debug, 0, "vir_lj_recip after corr", fr->vir_lj_recip, DIM);
 637         }
 638     }
 639     else
 640     {
 641         /* Is there a reaction-field exclusion correction needed? */
 642         if (EEL_RF(fr->eeltype) && eelRF_NEC != fr->eeltype)
 643         {
 644             /* With the Verlet scheme, exclusion forces are calculated
 645              * in the non-bonded kernel.
 646              */
 647             if (ir->cutoff_scheme != ecutsVERLET)
 648             {
 649                 real dvdl_rf_excl      = 0;
 650                 enerd->term[F_RF_EXCL] =
 651                     RF_excl_correction(fr, graph, md, excl, x, f,
 652                                        fr->fshift, &pbc, lambda[efptCOUL], &dvdl_rf_excl);
 653
 654                 enerd->dvdl_lin[efptCOUL] += dvdl_rf_excl;
 655             }
 656         }
 657     }
 658     where();
 659     debug_gmx();
 660
 661     if (debug)
 662     {
 663         print_nrnb(debug, nrnb);
 664     }
 665     debug_gmx();
 666
 667 #ifdef GMX_MPI
 668     if (TAKETIME)
 669     {
 670         t2 = MPI_Wtime();
 671         MPI_Barrier(cr->mpi_comm_mygroup);
 672         t3          = MPI_Wtime();
 673         fr->t_wait += t3-t2;
 674         if (fr->timesteps == 11)
 675         {
 676             fprintf(stderr, "* PP load balancing info: rank %d, step %s, rel wait time=%3.0f%% , load string value: %7.2f\n",
 677                     cr->nodeid, gmx_step_str(fr->timesteps, buf),
 678                     100*fr->t_wait/(fr->t_wait+fr->t_fnbf),
 679                     (fr->t_fnbf+fr->t_wait)/fr->t_fnbf);
 680         }
 681         fr->timesteps++;
 682     }
 683 #endif
 684
 685     if (debug)
 686     {
 687         pr_rvecs(debug, 0, "fshift after bondeds", fr->fshift, SHIFTS);
 688     }
 689
 690 }
 691
 692 void init_enerdata(int ngener, int n_lambda, gmx_enerdata_t *enerd)
 693 {
 694     int i, n2;
 695
 696     for (i = 0; i < F_NRE; i++)
 697     {
 698         enerd->term[i]         = 0;
 699         enerd->foreign_term[i] = 0;
 700     }
 701
 702
 703     for (i = 0; i < efptNR; i++)
 704     {
 705         enerd->dvdl_lin[i]     = 0;
 706         enerd->dvdl_nonlin[i]  = 0;
 707     }
 708
 709     n2 = ngener*ngener;
 710     if (debug)
 711     {
 712         fprintf(debug, "Creating %d sized group matrix for energies\n", n2);
 713     }
 714     enerd->grpp.nener         = n2;
 715     enerd->foreign_grpp.nener = n2;
 716     for (i = 0; (i < egNR); i++)
 717     {
 718         snew(enerd->grpp.ener[i], n2);
 719         snew(enerd->foreign_grpp.ener[i], n2);
 720     }
 721
 722     if (n_lambda)
 723     {
 724         enerd->n_lambda = 1 + n_lambda;
 725         snew(enerd->enerpart_lambda, enerd->n_lambda);
 726     }
 727     else
 728     {
 729         enerd->n_lambda = 0;
 730     }
 731 }
 732
 733 void destroy_enerdata(gmx_enerdata_t *enerd)
 734 {
 735     int i;
 736
 737     for (i = 0; (i < egNR); i++)
 738     {
 739         sfree(enerd->grpp.ener[i]);
 740     }
 741
 742     for (i = 0; (i < egNR); i++)
 743     {
 744         sfree(enerd->foreign_grpp.ener[i]);
 745     }
 746
 747     if (enerd->n_lambda)
 748     {
 749         sfree(enerd->enerpart_lambda);
 750     }
 751 }
 752
 753 static real sum_v(int n, real v[])
 754 {
 755     real t;
 756     int  i;
 757
 758     t = 0.0;
 759     for (i = 0; (i < n); i++)
 760     {
 761         t = t + v[i];
 762     }
 763
 764     return t;
 765 }
 766
 767 void sum_epot(gmx_grppairener_t *grpp, real *epot)
 768 {
 769     int i;
 770
 771     /* Accumulate energies */
 772     epot[F_COUL_SR]  = sum_v(grpp->nener, grpp->ener[egCOULSR]);
 773     epot[F_LJ]       = sum_v(grpp->nener, grpp->ener[egLJSR]);
 774     epot[F_LJ14]     = sum_v(grpp->nener, grpp->ener[egLJ14]);
 775     epot[F_COUL14]   = sum_v(grpp->nener, grpp->ener[egCOUL14]);
 776     epot[F_COUL_LR]  = sum_v(grpp->nener, grpp->ener[egCOULLR]);
 777     epot[F_LJ_LR]    = sum_v(grpp->nener, grpp->ener[egLJLR]);
 778     /* We have already added 1-2,1-3, and 1-4 terms to F_GBPOL */
 779     epot[F_GBPOL]   += sum_v(grpp->nener, grpp->ener[egGB]);
 780
 781 /* lattice part of LR doesnt belong to any group
 782  * and has been added earlier
 783  */
 784     epot[F_BHAM]     = sum_v(grpp->nener, grpp->ener[egBHAMSR]);
 785     epot[F_BHAM_LR]  = sum_v(grpp->nener, grpp->ener[egBHAMLR]);
 786
 787     epot[F_EPOT] = 0;
 788     for (i = 0; (i < F_EPOT); i++)
 789     {
 790         if (i != F_DISRESVIOL && i != F_ORIRESDEV)
 791         {
 792             epot[F_EPOT] += epot[i];
 793         }
 794     }
 795 }
 796
 797 void sum_dhdl(gmx_enerdata_t *enerd, real *lambda, t_lambda *fepvals)
 798 {
 799     int    i, j, index;
 800     double dlam;
 801
 802     enerd->dvdl_lin[efptVDW] += enerd->term[F_DVDL_VDW];  /* include dispersion correction */
 803     enerd->term[F_DVDL]       = 0.0;
 804     for (i = 0; i < efptNR; i++)
 805     {
 806         if (fepvals->separate_dvdl[i])
 807         {
 808             /* could this be done more readably/compactly? */
 809             switch (i)
 810             {
 811                 case (efptMASS):
 812                     index = F_DKDL;
 813                     break;
 814                 case (efptCOUL):
 815                     index = F_DVDL_COUL;
 816                     break;
 817                 case (efptVDW):
 818                     index = F_DVDL_VDW;
 819                     break;
 820                 case (efptBONDED):
 821                     index = F_DVDL_BONDED;
 822                     break;
 823                 case (efptRESTRAINT):
 824                     index = F_DVDL_RESTRAINT;
 825                     break;
 826                 default:
 827                     index = F_DVDL;
 828                     break;
 829             }
 830             enerd->term[index] = enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 831             if (debug)
 832             {
 833                 fprintf(debug, "dvdl-%s[%2d]: %f: non-linear %f + linear %f\n",
 834                         efpt_names[i], i, enerd->term[index], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 835             }
 836         }
 837         else
 838         {
 839             enerd->term[F_DVDL] += enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 840             if (debug)
 841             {
 842                 fprintf(debug, "dvd-%sl[%2d]: %f: non-linear %f + linear %f\n",
 843                         efpt_names[0], i, enerd->term[F_DVDL], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 844             }
 845         }
 846     }
 847
 848     /* Notes on the foreign lambda free energy difference evaluation:
 849      * Adding the potential and ekin terms that depend linearly on lambda
 850      * as delta lam * dvdl to the energy differences is exact.
 851      * For the constraints this is not exact, but we have no other option
 852      * without literally changing the lengths and reevaluating the energies at each step.
 853      * (try to remedy this post 4.6 - MRS)
 854      * For the non-bonded LR term we assume that the soft-core (if present)
 855      * no longer affects the energy beyond the short-range cut-off,
 856      * which is a very good approximation (except for exotic settings).
 857      * (investigate how to overcome this post 4.6 - MRS)
 858      */
 859     if (fepvals->separate_dvdl[efptBONDED])
 860     {
 861         enerd->term[F_DVDL_BONDED] += enerd->term[F_DVDL_CONSTR];
 862     }
 863     else
 864     {
 865         enerd->term[F_DVDL] += enerd->term[F_DVDL_CONSTR];
 866     }
 867     enerd->term[F_DVDL_CONSTR] = 0;
 868
 869     for (i = 0; i < fepvals->n_lambda; i++)
 870     {
 871         /* note we are iterating over fepvals here!
 872            For the current lam, dlam = 0 automatically,
 873            so we don't need to add anything to the
 874            enerd->enerpart_lambda[0] */
 875
 876         /* we don't need to worry about dvdl_lin contributions to dE at
 877            current lambda, because the contributions to the current
 878            lambda are automatically zeroed */
 879
 880         for (j = 0; j < efptNR; j++)
 881         {
 882             /* Note that this loop is over all dhdl components, not just the separated ones */
 883             dlam = (fepvals->all_lambda[j][i]-lambda[j]);
 884             enerd->enerpart_lambda[i+1] += dlam*enerd->dvdl_lin[j];
 885             if (debug)
 886             {
 887                 fprintf(debug, "enerdiff lam %g: (%15s), non-linear %f linear %f*%f\n",
 888                         fepvals->all_lambda[j][i], efpt_names[j],
 889                         (enerd->enerpart_lambda[i+1] - enerd->enerpart_lambda[0]),
 890                         dlam, enerd->dvdl_lin[j]);
 891             }
 892         }
 893     }
 894 }
 895
 896
 897 void reset_foreign_enerdata(gmx_enerdata_t *enerd)
 898 {
 899     int  i, j;
 900
 901     /* First reset all foreign energy components.  Foreign energies always called on
 902        neighbor search steps */
 903     for (i = 0; (i < egNR); i++)
 904     {
 905         for (j = 0; (j < enerd->grpp.nener); j++)
 906         {
 907             enerd->foreign_grpp.ener[i][j] = 0.0;
 908         }
 909     }
 910
 911     /* potential energy components */
 912     for (i = 0; (i <= F_EPOT); i++)
 913     {
 914         enerd->foreign_term[i] = 0.0;
 915     }
 916 }
 917
 918 void reset_enerdata(t_forcerec *fr, gmx_bool bNS,
 919                     gmx_enerdata_t *enerd,
 920                     gmx_bool bMaster)
 921 {
 922     gmx_bool bKeepLR;
 923     int      i, j;
 924
 925     /* First reset all energy components, except for the long range terms
 926      * on the master at non neighbor search steps, since the long range
 927      * terms have already been summed at the last neighbor search step.
 928      */
 929     bKeepLR = (fr->bTwinRange && !bNS);
 930     for (i = 0; (i < egNR); i++)
 931     {
 932         if (!(bKeepLR && bMaster && (i == egCOULLR || i == egLJLR)))
 933         {
 934             for (j = 0; (j < enerd->grpp.nener); j++)
 935             {
 936                 enerd->grpp.ener[i][j] = 0.0;
 937             }
 938         }
 939     }
 940     for (i = 0; i < efptNR; i++)
 941     {
 942         enerd->dvdl_lin[i]    = 0.0;
 943         enerd->dvdl_nonlin[i] = 0.0;
 944     }
 945
 946     /* Normal potential energy components */
 947     for (i = 0; (i <= F_EPOT); i++)
 948     {
 949         enerd->term[i] = 0.0;
 950     }
 951     /* Initialize the dVdlambda term with the long range contribution */
 952     /* Initialize the dvdl term with the long range contribution */
 953     enerd->term[F_DVDL]            = 0.0;
 954     enerd->term[F_DVDL_COUL]       = 0.0;
 955     enerd->term[F_DVDL_VDW]        = 0.0;
 956     enerd->term[F_DVDL_BONDED]     = 0.0;
 957     enerd->term[F_DVDL_RESTRAINT]  = 0.0;
 958     enerd->term[F_DKDL]            = 0.0;
 959     if (enerd->n_lambda > 0)
 960     {
 961         for (i = 0; i < enerd->n_lambda; i++)
 962         {
 963             enerd->enerpart_lambda[i] = 0.0;
 964         }
 965     }
 966     /* reset foreign energy data - separate function since we also call it elsewhere */
 967     reset_foreign_enerdata(enerd);
 968 }