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 "gmxpre.h"
  38
  39 #include "config.h"
  40
  41 #include <assert.h>
  42 #include <math.h>
  43 #include <string.h>
  44
  45 #include "gromacs/legacyheaders/typedefs.h"
  46 #include "gromacs/legacyheaders/macros.h"
  47 #include "gromacs/legacyheaders/force.h"
  48 #include "gromacs/legacyheaders/nonbonded.h"
  49 #include "gromacs/legacyheaders/names.h"
  50 #include "gromacs/legacyheaders/network.h"
  51 #include "gromacs/legacyheaders/ns.h"
  52 #include "gromacs/legacyheaders/nrnb.h"
  53 #include "gromacs/legacyheaders/bondf.h"
  54 #include "gromacs/legacyheaders/txtdump.h"
  55 #include "gromacs/legacyheaders/coulomb.h"
  56 #include "gromacs/legacyheaders/pme.h"
  57 #include "gromacs/legacyheaders/mdrun.h"
  58 #include "gromacs/legacyheaders/domdec.h"
  59 #include "gromacs/legacyheaders/qmmm.h"
  60 #include "gromacs/legacyheaders/gmx_omp_nthreads.h"
  61
  62 #include "gromacs/legacyheaders/types/commrec.h"
  63 #include "gromacs/math/vec.h"
  64 #include "gromacs/pbcutil/ishift.h"
  65 #include "gromacs/pbcutil/mshift.h"
  66 #include "gromacs/pbcutil/pbc.h"
  67 #include "gromacs/timing/wallcycle.h"
  68 #include "gromacs/utility/fatalerror.h"
  69 #include "gromacs/utility/smalloc.h"
  70
  71 void ns(FILE              *fp,
  72         t_forcerec        *fr,
  73         matrix             box,
  74         gmx_groups_t      *groups,
  75         gmx_localtop_t    *top,
  76         t_mdatoms         *md,
  77         t_commrec         *cr,
  78         t_nrnb            *nrnb,
  79         gmx_bool           bFillGrid,
  80         gmx_bool           bDoLongRangeNS)
  81 {
  82     char   *ptr;
  83     int     nsearch;
  84
  85
  86     if (!fr->ns.nblist_initialized)
  87     {
  88         init_neighbor_list(fp, fr, md->homenr);
  89     }
  90
  91     if (fr->bTwinRange)
  92     {
  93         fr->nlr = 0;
  94     }
  95
  96     nsearch = search_neighbours(fp, fr, box, top, groups, cr, nrnb, md,
  97                                 bFillGrid, bDoLongRangeNS);
  98     if (debug)
  99     {
 100         fprintf(debug, "nsearch = %d\n", nsearch);
 101     }
 102
 103     /* Check whether we have to do dynamic load balancing */
 104     /*if ((nsb->nstDlb > 0) && (mod(step,nsb->nstDlb) == 0))
 105        count_nb(cr,nsb,&(top->blocks[ebCGS]),nns,fr->nlr,
 106        &(top->idef),opts->ngener);
 107      */
 108     if (fr->ns.dump_nl > 0)
 109     {
 110         dump_nblist(fp, cr, fr, fr->ns.dump_nl);
 111     }
 112 }
 113
 114 static void reduce_thread_forces(int n, rvec *f,
 115                                  tensor vir_q, tensor vir_lj,
 116                                  real *Vcorr_q, real *Vcorr_lj,
 117                                  real *dvdl_q, real *dvdl_lj,
 118                                  int nthreads, f_thread_t *f_t)
 119 {
 120     int t, i;
 121
 122     /* This reduction can run over any number of threads */
 123 #pragma omp parallel for num_threads(gmx_omp_nthreads_get(emntBonded)) private(t) schedule(static)
 124     for (i = 0; i < n; i++)
 125     {
 126         for (t = 1; t < nthreads; t++)
 127         {
 128             rvec_inc(f[i], f_t[t].f[i]);
 129         }
 130     }
 131     for (t = 1; t < nthreads; t++)
 132     {
 133         *Vcorr_q  += f_t[t].Vcorr_q;
 134         *Vcorr_lj += f_t[t].Vcorr_lj;
 135         *dvdl_q   += f_t[t].dvdl[efptCOUL];
 136         *dvdl_lj  += f_t[t].dvdl[efptVDW];
 137         m_add(vir_q, f_t[t].vir_q, vir_q);
 138         m_add(vir_lj, f_t[t].vir_lj, vir_lj);
 139     }
 140 }
 141
 142 void do_force_lowlevel(t_forcerec *fr,      t_inputrec *ir,
 143                        t_idef     *idef,    t_commrec  *cr,
 144                        t_nrnb     *nrnb,    gmx_wallcycle_t wcycle,
 145                        t_mdatoms  *md,
 146                        rvec       x[],      history_t  *hist,
 147                        rvec       f[],
 148                        rvec       f_longrange[],
 149                        gmx_enerdata_t *enerd,
 150                        t_fcdata   *fcd,
 151                        gmx_localtop_t *top,
 152                        gmx_genborn_t *born,
 153                        t_atomtypes *atype,
 154                        gmx_bool       bBornRadii,
 155                        matrix     box,
 156                        t_lambda   *fepvals,
 157                        real       *lambda,
 158                        t_graph    *graph,
 159                        t_blocka   *excl,
 160                        rvec       mu_tot[],
 161                        int        flags,
 162                        float      *cycles_pme)
 163 {
 164     int         i, j;
 165     int         donb_flags;
 166     gmx_bool    bDoEpot, bSB;
 167     int         pme_flags;
 168     matrix      boxs;
 169     rvec        box_size;
 170     t_pbc       pbc;
 171     char        buf[22];
 172     double      clam_i, vlam_i;
 173     real        dvdl_dum[efptNR], dvdl_nb[efptNR], lam_i[efptNR];
 174     real        dvdl_q, dvdl_lj;
 175
 176 #ifdef GMX_MPI
 177     double  t0 = 0.0, t1, t2, t3; /* time measurement for coarse load balancing */
 178 #endif
 179
 180     set_pbc(&pbc, fr->ePBC, box);
 181
 182     /* reset free energy components */
 183     for (i = 0; i < efptNR; i++)
 184     {
 185         dvdl_nb[i]  = 0;
 186         dvdl_dum[i] = 0;
 187     }
 188
 189     /* Reset box */
 190     for (i = 0; (i < DIM); i++)
 191     {
 192         box_size[i] = box[i][i];
 193     }
 194
 195     debug_gmx();
 196
 197     /* do QMMM first if requested */
 198     if (fr->bQMMM)
 199     {
 200         enerd->term[F_EQM] = calculate_QMMM(cr, x, f, fr);
 201     }
 202
 203     /* Call the short range functions all in one go. */
 204
 205 #ifdef GMX_MPI
 206     /*#define TAKETIME ((cr->npmenodes) && (fr->timesteps < 12))*/
 207 #define TAKETIME FALSE
 208     if (TAKETIME)
 209     {
 210         MPI_Barrier(cr->mpi_comm_mygroup);
 211         t0 = MPI_Wtime();
 212     }
 213 #endif
 214
 215     if (ir->nwall)
 216     {
 217         /* foreign lambda component for walls */
 218         real dvdl_walls = do_walls(ir, fr, box, md, x, f, lambda[efptVDW],
 219                                    enerd->grpp.ener[egLJSR], nrnb);
 220         enerd->dvdl_lin[efptVDW] += dvdl_walls;
 221     }
 222
 223     /* If doing GB, reset dvda and calculate the Born radii */
 224     if (ir->implicit_solvent)
 225     {
 226         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 227
 228         for (i = 0; i < born->nr; i++)
 229         {
 230             fr->dvda[i] = 0;
 231         }
 232
 233         if (bBornRadii)
 234         {
 235             calc_gb_rad(cr, fr, ir, top, x, &(fr->gblist), born, md, nrnb);
 236         }
 237
 238         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 239     }
 240
 241     where();
 242     /* We only do non-bonded calculation with group scheme here, the verlet
 243      * calls are done from do_force_cutsVERLET(). */
 244     if (fr->cutoff_scheme == ecutsGROUP && (flags & GMX_FORCE_NONBONDED))
 245     {
 246         donb_flags = 0;
 247         /* Add short-range interactions */
 248         donb_flags |= GMX_NONBONDED_DO_SR;
 249
 250         /* Currently all group scheme kernels always calculate (shift-)forces */
 251         if (flags & GMX_FORCE_FORCES)
 252         {
 253             donb_flags |= GMX_NONBONDED_DO_FORCE;
 254         }
 255         if (flags & GMX_FORCE_VIRIAL)
 256         {
 257             donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 258         }
 259         if (flags & GMX_FORCE_ENERGY)
 260         {
 261             donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 262         }
 263         if (flags & GMX_FORCE_DO_LR)
 264         {
 265             donb_flags |= GMX_NONBONDED_DO_LR;
 266         }
 267
 268         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 269         do_nonbonded(fr, x, f, f_longrange, md, excl,
 270                      &enerd->grpp, nrnb,
 271                      lambda, dvdl_nb, -1, -1, donb_flags);
 272
 273         /* If we do foreign lambda and we have soft-core interactions
 274          * we have to recalculate the (non-linear) energies contributions.
 275          */
 276         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 277         {
 278             for (i = 0; i < enerd->n_lambda; i++)
 279             {
 280                 for (j = 0; j < efptNR; j++)
 281                 {
 282                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 283                 }
 284                 reset_foreign_enerdata(enerd);
 285                 do_nonbonded(fr, x, f, f_longrange, md, excl,
 286                              &(enerd->foreign_grpp), nrnb,
 287                              lam_i, dvdl_dum, -1, -1,
 288                              (donb_flags & ~GMX_NONBONDED_DO_FORCE) | GMX_NONBONDED_DO_FOREIGNLAMBDA);
 289                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 290                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 291             }
 292         }
 293         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 294         where();
 295     }
 296
 297     /* If we are doing GB, calculate bonded forces and apply corrections
 298      * to the solvation forces */
 299     /* MRS: Eventually, many need to include free energy contribution here! */
 300     if (ir->implicit_solvent)
 301     {
 302         wallcycle_sub_start(wcycle, ewcsBONDED);
 303         calc_gb_forces(cr, md, born, top, x, f, fr, idef,
 304                        ir->gb_algorithm, ir->sa_algorithm, nrnb, &pbc, graph, enerd);
 305         wallcycle_sub_stop(wcycle, ewcsBONDED);
 306     }
 307
 308 #ifdef GMX_MPI
 309     if (TAKETIME)
 310     {
 311         t1          = MPI_Wtime();
 312         fr->t_fnbf += t1-t0;
 313     }
 314 #endif
 315
 316     if (fepvals->sc_alpha != 0)
 317     {
 318         enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
 319     }
 320     else
 321     {
 322         enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
 323     }
 324
 325     if (fepvals->sc_alpha != 0)
 326
 327     /* even though coulomb part is linear, we already added it, beacuse we
 328        need to go through the vdw calculation anyway */
 329     {
 330         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 331     }
 332     else
 333     {
 334         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 335     }
 336
 337     debug_gmx();
 338
 339
 340     if (debug)
 341     {
 342         pr_rvecs(debug, 0, "fshift after SR", fr->fshift, SHIFTS);
 343     }
 344
 345     /* Shift the coordinates. Must be done before bonded forces and PPPM,
 346      * but is also necessary for SHAKE and update, therefore it can NOT
 347      * go when no bonded forces have to be evaluated.
 348      */
 349
 350     /* Here sometimes we would not need to shift with NBFonly,
 351      * but we do so anyhow for consistency of the returned coordinates.
 352      */
 353     if (graph)
 354     {
 355         shift_self(graph, box, x);
 356         if (TRICLINIC(box))
 357         {
 358             inc_nrnb(nrnb, eNR_SHIFTX, 2*graph->nnodes);
 359         }
 360         else
 361         {
 362             inc_nrnb(nrnb, eNR_SHIFTX, graph->nnodes);
 363         }
 364     }
 365     /* Check whether we need to do bondeds or correct for exclusions */
 366     if (fr->bMolPBC &&
 367         ((flags & GMX_FORCE_BONDED)
 368          || EEL_RF(fr->eeltype) || EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype)))
 369     {
 370         /* Since all atoms are in the rectangular or triclinic unit-cell,
 371          * only single box vector shifts (2 in x) are required.
 372          */
 373         set_pbc_dd(&pbc, fr->ePBC, cr->dd, TRUE, box);
 374     }
 375     debug_gmx();
 376
 377     if (flags & GMX_FORCE_BONDED)
 378     {
 379         wallcycle_sub_start(wcycle, ewcsBONDED);
 380         calc_bonds(cr->ms,
 381                    idef, x, hist, f, fr, &pbc, graph, enerd, nrnb, lambda, md, fcd,
 382                    DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL, atype, born,
 383                    flags);
 384
 385         /* Check if we have to determine energy differences
 386          * at foreign lambda's.
 387          */
 388         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) &&
 389             idef->ilsort != ilsortNO_FE)
 390         {
 391             if (idef->ilsort != ilsortFE_SORTED)
 392             {
 393                 gmx_incons("The bonded interactions are not sorted for free energy");
 394             }
 395             for (i = 0; i < enerd->n_lambda; i++)
 396             {
 397                 reset_foreign_enerdata(enerd);
 398                 for (j = 0; j < efptNR; j++)
 399                 {
 400                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 401                 }
 402                 calc_bonds_lambda(idef, x, fr, &pbc, graph, &(enerd->foreign_grpp), enerd->foreign_term, nrnb, lam_i, md,
 403                                   fcd, DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL);
 404                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 405                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 406             }
 407         }
 408         debug_gmx();
 409
 410         wallcycle_sub_stop(wcycle, ewcsBONDED);
 411     }
 412
 413     where();
 414
 415     *cycles_pme = 0;
 416     clear_mat(fr->vir_el_recip);
 417     clear_mat(fr->vir_lj_recip);
 418
 419     /* Do long-range electrostatics and/or LJ-PME, including related short-range
 420      * corrections.
 421      */
 422     if (EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype))
 423     {
 424         real Vlr               = 0, Vcorr = 0;
 425         real dvdl_long_range   = 0;
 426         int  status            = 0;
 427         real Vlr_q             = 0, Vlr_lj = 0, Vcorr_q = 0, Vcorr_lj = 0;
 428         real dvdl_long_range_q = 0, dvdl_long_range_lj = 0;
 429
 430         bSB = (ir->nwall == 2);
 431         if (bSB)
 432         {
 433             copy_mat(box, boxs);
 434             svmul(ir->wall_ewald_zfac, boxs[ZZ], boxs[ZZ]);
 435             box_size[ZZ] *= ir->wall_ewald_zfac;
 436         }
 437
 438         if (EEL_PME_EWALD(fr->eeltype) || EVDW_PME(fr->vdwtype))
 439         {
 440             real dvdl_long_range_correction_q   = 0;
 441             real dvdl_long_range_correction_lj  = 0;
 442             /* With the Verlet scheme exclusion forces are calculated
 443              * in the non-bonded kernel.
 444              */
 445             /* The TPI molecule does not have exclusions with the rest
 446              * of the system and no intra-molecular PME grid
 447              * contributions will be calculated in
 448              * gmx_pme_calc_energy.
 449              */
 450             if ((ir->cutoff_scheme == ecutsGROUP && fr->n_tpi == 0) ||
 451                 ir->ewald_geometry != eewg3D ||
 452                 ir->epsilon_surface != 0)
 453             {
 454                 int nthreads, t;
 455
 456                 wallcycle_sub_start(wcycle, ewcsEWALD_CORRECTION);
 457
 458                 if (fr->n_tpi > 0)
 459                 {
 460                     gmx_fatal(FARGS, "TPI with PME currently only works in a 3D geometry with tin-foil boundary conditions");
 461                 }
 462
 463                 nthreads = gmx_omp_nthreads_get(emntBonded);
 464 #pragma omp parallel for num_threads(nthreads) schedule(static)
 465                 for (t = 0; t < nthreads; t++)
 466                 {
 467                     int     s, e, i;
 468                     rvec   *fnv;
 469                     tensor *vir_q, *vir_lj;
 470                     real   *Vcorrt_q, *Vcorrt_lj, *dvdlt_q, *dvdlt_lj;
 471                     if (t == 0)
 472                     {
 473                         fnv       = fr->f_novirsum;
 474                         vir_q     = &fr->vir_el_recip;
 475                         vir_lj    = &fr->vir_lj_recip;
 476                         Vcorrt_q  = &Vcorr_q;
 477                         Vcorrt_lj = &Vcorr_lj;
 478                         dvdlt_q   = &dvdl_long_range_correction_q;
 479                         dvdlt_lj  = &dvdl_long_range_correction_lj;
 480                     }
 481                     else
 482                     {
 483                         fnv       = fr->f_t[t].f;
 484                         vir_q     = &fr->f_t[t].vir_q;
 485                         vir_lj    = &fr->f_t[t].vir_lj;
 486                         Vcorrt_q  = &fr->f_t[t].Vcorr_q;
 487                         Vcorrt_lj = &fr->f_t[t].Vcorr_lj;
 488                         dvdlt_q   = &fr->f_t[t].dvdl[efptCOUL];
 489                         dvdlt_lj  = &fr->f_t[t].dvdl[efptVDW];
 490                         for (i = 0; i < fr->natoms_force; i++)
 491                         {
 492                             clear_rvec(fnv[i]);
 493                         }
 494                         clear_mat(*vir_q);
 495                         clear_mat(*vir_lj);
 496                     }
 497                     *dvdlt_q  = 0;
 498                     *dvdlt_lj = 0;
 499
 500                     ewald_LRcorrection(fr->excl_load[t], fr->excl_load[t+1],
 501                                        cr, t, fr,
 502                                        md->chargeA,
 503                                        md->nChargePerturbed ? md->chargeB : NULL,
 504                                        md->sqrt_c6A,
 505                                        md->nTypePerturbed ? md->sqrt_c6B : NULL,
 506                                        md->sigmaA,
 507                                        md->nTypePerturbed ? md->sigmaB : NULL,
 508                                        md->sigma3A,
 509                                        md->nTypePerturbed ? md->sigma3B : NULL,
 510                                        ir->cutoff_scheme != ecutsVERLET,
 511                                        excl, x, bSB ? boxs : box, mu_tot,
 512                                        ir->ewald_geometry,
 513                                        ir->epsilon_surface,
 514                                        fnv, *vir_q, *vir_lj,
 515                                        Vcorrt_q, Vcorrt_lj,
 516                                        lambda[efptCOUL], lambda[efptVDW],
 517                                        dvdlt_q, dvdlt_lj);
 518                 }
 519                 if (nthreads > 1)
 520                 {
 521                     reduce_thread_forces(fr->natoms_force, fr->f_novirsum,
 522                                          fr->vir_el_recip, fr->vir_lj_recip,
 523                                          &Vcorr_q, &Vcorr_lj,
 524                                          &dvdl_long_range_correction_q,
 525                                          &dvdl_long_range_correction_lj,
 526                                          nthreads, fr->f_t);
 527                 }
 528                 wallcycle_sub_stop(wcycle, ewcsEWALD_CORRECTION);
 529             }
 530
 531             if (EEL_PME_EWALD(fr->eeltype) && fr->n_tpi == 0)
 532             {
 533                 Vcorr_q += ewald_charge_correction(cr, fr, lambda[efptCOUL], box,
 534                                                    &dvdl_long_range_correction_q,
 535                                                    fr->vir_el_recip);
 536             }
 537
 538             enerd->dvdl_lin[efptCOUL] += dvdl_long_range_correction_q;
 539             enerd->dvdl_lin[efptVDW]  += dvdl_long_range_correction_lj;
 540
 541             if ((EEL_PME(fr->eeltype) || EVDW_PME(fr->vdwtype)) && (cr->duty & DUTY_PME))
 542             {
 543                 /* Do reciprocal PME for Coulomb and/or LJ. */
 544                 assert(fr->n_tpi >= 0);
 545                 if (fr->n_tpi == 0 || (flags & GMX_FORCE_STATECHANGED))
 546                 {
 547                     pme_flags = GMX_PME_SPREAD | GMX_PME_SOLVE;
 548                     if (EEL_PME(fr->eeltype))
 549                     {
 550                         pme_flags     |= GMX_PME_DO_COULOMB;
 551                     }
 552                     if (EVDW_PME(fr->vdwtype))
 553                     {
 554                         pme_flags |= GMX_PME_DO_LJ;
 555                     }
 556                     if (flags & GMX_FORCE_FORCES)
 557                     {
 558                         pme_flags |= GMX_PME_CALC_F;
 559                     }
 560                     if (flags & GMX_FORCE_VIRIAL)
 561                     {
 562                         pme_flags |= GMX_PME_CALC_ENER_VIR;
 563                     }
 564                     if (fr->n_tpi > 0)
 565                     {
 566                         /* We don't calculate f, but we do want the potential */
 567                         pme_flags |= GMX_PME_CALC_POT;
 568                     }
 569                     wallcycle_start(wcycle, ewcPMEMESH);
 570                     status = gmx_pme_do(fr->pmedata,
 571                                         0, md->homenr - fr->n_tpi,
 572                                         x, fr->f_novirsum,
 573                                         md->chargeA, md->chargeB,
 574                                         md->sqrt_c6A, md->sqrt_c6B,
 575                                         md->sigmaA, md->sigmaB,
 576                                         bSB ? boxs : box, cr,
 577                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_x(cr->dd) : 0,
 578                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_y(cr->dd) : 0,
 579                                         nrnb, wcycle,
 580                                         fr->vir_el_recip, fr->ewaldcoeff_q,
 581                                         fr->vir_lj_recip, fr->ewaldcoeff_lj,
 582                                         &Vlr_q, &Vlr_lj,
 583                                         lambda[efptCOUL], lambda[efptVDW],
 584                                         &dvdl_long_range_q, &dvdl_long_range_lj, pme_flags);
 585                     *cycles_pme = wallcycle_stop(wcycle, ewcPMEMESH);
 586                     if (status != 0)
 587                     {
 588                         gmx_fatal(FARGS, "Error %d in reciprocal PME routine", status);
 589                     }
 590                     /* We should try to do as little computation after
 591                      * this as possible, because parallel PME synchronizes
 592                      * the nodes, so we want all load imbalance of the
 593                      * rest of the force calculation to be before the PME
 594                      * call.  DD load balancing is done on the whole time
 595                      * of the force call (without PME).
 596                      */
 597                 }
 598                 if (fr->n_tpi > 0)
 599                 {
 600                     if (EVDW_PME(ir->vdwtype))
 601                     {
 602
 603                         gmx_fatal(FARGS, "Test particle insertion not implemented with LJ-PME");
 604                     }
 605                     /* Determine the PME grid energy of the test molecule
 606                      * with the PME grid potential of the other charges.
 607                      */
 608                     gmx_pme_calc_energy(fr->pmedata, fr->n_tpi,
 609                                         x + md->homenr - fr->n_tpi,
 610                                         md->chargeA + md->homenr - fr->n_tpi,
 611                                         &Vlr_q);
 612                 }
 613             }
 614         }
 615
 616         if (!EEL_PME(fr->eeltype) && EEL_PME_EWALD(fr->eeltype))
 617         {
 618             Vlr_q = do_ewald(ir, x, fr->f_novirsum,
 619                              md->chargeA, md->chargeB,
 620                              box_size, cr, md->homenr,
 621                              fr->vir_el_recip, fr->ewaldcoeff_q,
 622                              lambda[efptCOUL], &dvdl_long_range_q, fr->ewald_table);
 623         }
 624
 625         /* Note that with separate PME nodes we get the real energies later */
 626         enerd->dvdl_lin[efptCOUL] += dvdl_long_range_q;
 627         enerd->dvdl_lin[efptVDW]  += dvdl_long_range_lj;
 628         enerd->term[F_COUL_RECIP]  = Vlr_q + Vcorr_q;
 629         enerd->term[F_LJ_RECIP]    = Vlr_lj + Vcorr_lj;
 630         if (debug)
 631         {
 632             fprintf(debug, "Vlr_q = %g, Vcorr_q = %g, Vlr_corr_q = %g\n",
 633                     Vlr_q, Vcorr_q, enerd->term[F_COUL_RECIP]);
 634             pr_rvecs(debug, 0, "vir_el_recip after corr", fr->vir_el_recip, DIM);
 635             pr_rvecs(debug, 0, "fshift after LR Corrections", fr->fshift, SHIFTS);
 636             fprintf(debug, "Vlr_lj: %g, Vcorr_lj = %g, Vlr_corr_lj = %g\n",
 637                     Vlr_lj, Vcorr_lj, enerd->term[F_LJ_RECIP]);
 638             pr_rvecs(debug, 0, "vir_lj_recip after corr", fr->vir_lj_recip, DIM);
 639         }
 640     }
 641     else
 642     {
 643         /* Is there a reaction-field exclusion correction needed? */
 644         if (EEL_RF(fr->eeltype) && eelRF_NEC != fr->eeltype)
 645         {
 646             /* With the Verlet scheme, exclusion forces are calculated
 647              * in the non-bonded kernel.
 648              */
 649             if (ir->cutoff_scheme != ecutsVERLET)
 650             {
 651                 real dvdl_rf_excl      = 0;
 652                 enerd->term[F_RF_EXCL] =
 653                     RF_excl_correction(fr, graph, md, excl, x, f,
 654                                        fr->fshift, &pbc, lambda[efptCOUL], &dvdl_rf_excl);
 655
 656                 enerd->dvdl_lin[efptCOUL] += dvdl_rf_excl;
 657             }
 658         }
 659     }
 660     where();
 661     debug_gmx();
 662
 663     if (debug)
 664     {
 665         print_nrnb(debug, nrnb);
 666     }
 667     debug_gmx();
 668
 669 #ifdef GMX_MPI
 670     if (TAKETIME)
 671     {
 672         t2 = MPI_Wtime();
 673         MPI_Barrier(cr->mpi_comm_mygroup);
 674         t3          = MPI_Wtime();
 675         fr->t_wait += t3-t2;
 676         if (fr->timesteps == 11)
 677         {
 678             fprintf(stderr, "* PP load balancing info: rank %d, step %s, rel wait time=%3.0f%% , load string value: %7.2f\n",
 679                     cr->nodeid, gmx_step_str(fr->timesteps, buf),
 680                     100*fr->t_wait/(fr->t_wait+fr->t_fnbf),
 681                     (fr->t_fnbf+fr->t_wait)/fr->t_fnbf);
 682         }
 683         fr->timesteps++;
 684     }
 685 #endif
 686
 687     if (debug)
 688     {
 689         pr_rvecs(debug, 0, "fshift after bondeds", fr->fshift, SHIFTS);
 690     }
 691
 692 }
 693
 694 void init_enerdata(int ngener, int n_lambda, gmx_enerdata_t *enerd)
 695 {
 696     int i, n2;
 697
 698     for (i = 0; i < F_NRE; i++)
 699     {
 700         enerd->term[i]         = 0;
 701         enerd->foreign_term[i] = 0;
 702     }
 703
 704
 705     for (i = 0; i < efptNR; i++)
 706     {
 707         enerd->dvdl_lin[i]     = 0;
 708         enerd->dvdl_nonlin[i]  = 0;
 709     }
 710
 711     n2 = ngener*ngener;
 712     if (debug)
 713     {
 714         fprintf(debug, "Creating %d sized group matrix for energies\n", n2);
 715     }
 716     enerd->grpp.nener         = n2;
 717     enerd->foreign_grpp.nener = n2;
 718     for (i = 0; (i < egNR); i++)
 719     {
 720         snew(enerd->grpp.ener[i], n2);
 721         snew(enerd->foreign_grpp.ener[i], n2);
 722     }
 723
 724     if (n_lambda)
 725     {
 726         enerd->n_lambda = 1 + n_lambda;
 727         snew(enerd->enerpart_lambda, enerd->n_lambda);
 728     }
 729     else
 730     {
 731         enerd->n_lambda = 0;
 732     }
 733 }
 734
 735 void destroy_enerdata(gmx_enerdata_t *enerd)
 736 {
 737     int i;
 738
 739     for (i = 0; (i < egNR); i++)
 740     {
 741         sfree(enerd->grpp.ener[i]);
 742     }
 743
 744     for (i = 0; (i < egNR); i++)
 745     {
 746         sfree(enerd->foreign_grpp.ener[i]);
 747     }
 748
 749     if (enerd->n_lambda)
 750     {
 751         sfree(enerd->enerpart_lambda);
 752     }
 753 }
 754
 755 static real sum_v(int n, real v[])
 756 {
 757     real t;
 758     int  i;
 759
 760     t = 0.0;
 761     for (i = 0; (i < n); i++)
 762     {
 763         t = t + v[i];
 764     }
 765
 766     return t;
 767 }
 768
 769 void sum_epot(gmx_grppairener_t *grpp, real *epot)
 770 {
 771     int i;
 772
 773     /* Accumulate energies */
 774     epot[F_COUL_SR]  = sum_v(grpp->nener, grpp->ener[egCOULSR]);
 775     epot[F_LJ]       = sum_v(grpp->nener, grpp->ener[egLJSR]);
 776     epot[F_LJ14]     = sum_v(grpp->nener, grpp->ener[egLJ14]);
 777     epot[F_COUL14]   = sum_v(grpp->nener, grpp->ener[egCOUL14]);
 778     epot[F_COUL_LR]  = sum_v(grpp->nener, grpp->ener[egCOULLR]);
 779     epot[F_LJ_LR]    = sum_v(grpp->nener, grpp->ener[egLJLR]);
 780     /* We have already added 1-2,1-3, and 1-4 terms to F_GBPOL */
 781     epot[F_GBPOL]   += sum_v(grpp->nener, grpp->ener[egGB]);
 782
 783 /* lattice part of LR doesnt belong to any group
 784  * and has been added earlier
 785  */
 786     epot[F_BHAM]     = sum_v(grpp->nener, grpp->ener[egBHAMSR]);
 787     epot[F_BHAM_LR]  = sum_v(grpp->nener, grpp->ener[egBHAMLR]);
 788
 789     epot[F_EPOT] = 0;
 790     for (i = 0; (i < F_EPOT); i++)
 791     {
 792         if (i != F_DISRESVIOL && i != F_ORIRESDEV)
 793         {
 794             epot[F_EPOT] += epot[i];
 795         }
 796     }
 797 }
 798
 799 void sum_dhdl(gmx_enerdata_t *enerd, real *lambda, t_lambda *fepvals)
 800 {
 801     int    i, j, index;
 802     double dlam;
 803
 804     enerd->dvdl_lin[efptVDW] += enerd->term[F_DVDL_VDW];  /* include dispersion correction */
 805     enerd->term[F_DVDL]       = 0.0;
 806     for (i = 0; i < efptNR; i++)
 807     {
 808         if (fepvals->separate_dvdl[i])
 809         {
 810             /* could this be done more readably/compactly? */
 811             switch (i)
 812             {
 813                 case (efptMASS):
 814                     index = F_DKDL;
 815                     break;
 816                 case (efptCOUL):
 817                     index = F_DVDL_COUL;
 818                     break;
 819                 case (efptVDW):
 820                     index = F_DVDL_VDW;
 821                     break;
 822                 case (efptBONDED):
 823                     index = F_DVDL_BONDED;
 824                     break;
 825                 case (efptRESTRAINT):
 826                     index = F_DVDL_RESTRAINT;
 827                     break;
 828                 default:
 829                     index = F_DVDL;
 830                     break;
 831             }
 832             enerd->term[index] = enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 833             if (debug)
 834             {
 835                 fprintf(debug, "dvdl-%s[%2d]: %f: non-linear %f + linear %f\n",
 836                         efpt_names[i], i, enerd->term[index], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 837             }
 838         }
 839         else
 840         {
 841             enerd->term[F_DVDL] += enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 842             if (debug)
 843             {
 844                 fprintf(debug, "dvd-%sl[%2d]: %f: non-linear %f + linear %f\n",
 845                         efpt_names[0], i, enerd->term[F_DVDL], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 846             }
 847         }
 848     }
 849
 850     /* Notes on the foreign lambda free energy difference evaluation:
 851      * Adding the potential and ekin terms that depend linearly on lambda
 852      * as delta lam * dvdl to the energy differences is exact.
 853      * For the constraints this is not exact, but we have no other option
 854      * without literally changing the lengths and reevaluating the energies at each step.
 855      * (try to remedy this post 4.6 - MRS)
 856      * For the non-bonded LR term we assume that the soft-core (if present)
 857      * no longer affects the energy beyond the short-range cut-off,
 858      * which is a very good approximation (except for exotic settings).
 859      * (investigate how to overcome this post 4.6 - MRS)
 860      */
 861     if (fepvals->separate_dvdl[efptBONDED])
 862     {
 863         enerd->term[F_DVDL_BONDED] += enerd->term[F_DVDL_CONSTR];
 864     }
 865     else
 866     {
 867         enerd->term[F_DVDL] += enerd->term[F_DVDL_CONSTR];
 868     }
 869     enerd->term[F_DVDL_CONSTR] = 0;
 870
 871     for (i = 0; i < fepvals->n_lambda; i++)
 872     {
 873         /* note we are iterating over fepvals here!
 874            For the current lam, dlam = 0 automatically,
 875            so we don't need to add anything to the
 876            enerd->enerpart_lambda[0] */
 877
 878         /* we don't need to worry about dvdl_lin contributions to dE at
 879            current lambda, because the contributions to the current
 880            lambda are automatically zeroed */
 881
 882         for (j = 0; j < efptNR; j++)
 883         {
 884             /* Note that this loop is over all dhdl components, not just the separated ones */
 885             dlam = (fepvals->all_lambda[j][i]-lambda[j]);
 886             enerd->enerpart_lambda[i+1] += dlam*enerd->dvdl_lin[j];
 887             if (debug)
 888             {
 889                 fprintf(debug, "enerdiff lam %g: (%15s), non-linear %f linear %f*%f\n",
 890                         fepvals->all_lambda[j][i], efpt_names[j],
 891                         (enerd->enerpart_lambda[i+1] - enerd->enerpart_lambda[0]),
 892                         dlam, enerd->dvdl_lin[j]);
 893             }
 894         }
 895     }
 896 }
 897
 898
 899 void reset_foreign_enerdata(gmx_enerdata_t *enerd)
 900 {
 901     int  i, j;
 902
 903     /* First reset all foreign energy components.  Foreign energies always called on
 904        neighbor search steps */
 905     for (i = 0; (i < egNR); i++)
 906     {
 907         for (j = 0; (j < enerd->grpp.nener); j++)
 908         {
 909             enerd->foreign_grpp.ener[i][j] = 0.0;
 910         }
 911     }
 912
 913     /* potential energy components */
 914     for (i = 0; (i <= F_EPOT); i++)
 915     {
 916         enerd->foreign_term[i] = 0.0;
 917     }
 918 }
 919
 920 void reset_enerdata(t_forcerec *fr, gmx_bool bNS,
 921                     gmx_enerdata_t *enerd,
 922                     gmx_bool bMaster)
 923 {
 924     gmx_bool bKeepLR;
 925     int      i, j;
 926
 927     /* First reset all energy components, except for the long range terms
 928      * on the master at non neighbor search steps, since the long range
 929      * terms have already been summed at the last neighbor search step.
 930      */
 931     bKeepLR = (fr->bTwinRange && !bNS);
 932     for (i = 0; (i < egNR); i++)
 933     {
 934         if (!(bKeepLR && bMaster && (i == egCOULLR || i == egLJLR)))
 935         {
 936             for (j = 0; (j < enerd->grpp.nener); j++)
 937             {
 938                 enerd->grpp.ener[i][j] = 0.0;
 939             }
 940         }
 941     }
 942     for (i = 0; i < efptNR; i++)
 943     {
 944         enerd->dvdl_lin[i]    = 0.0;
 945         enerd->dvdl_nonlin[i] = 0.0;
 946     }
 947
 948     /* Normal potential energy components */
 949     for (i = 0; (i <= F_EPOT); i++)
 950     {
 951         enerd->term[i] = 0.0;
 952     }
 953     /* Initialize the dVdlambda term with the long range contribution */
 954     /* Initialize the dvdl term with the long range contribution */
 955     enerd->term[F_DVDL]            = 0.0;
 956     enerd->term[F_DVDL_COUL]       = 0.0;
 957     enerd->term[F_DVDL_VDW]        = 0.0;
 958     enerd->term[F_DVDL_BONDED]     = 0.0;
 959     enerd->term[F_DVDL_RESTRAINT]  = 0.0;
 960     enerd->term[F_DKDL]            = 0.0;
 961     if (enerd->n_lambda > 0)
 962     {
 963         for (i = 0; i < enerd->n_lambda; i++)
 964         {
 965             enerd->enerpart_lambda[i] = 0.0;
 966         }
 967     }
 968     /* reset foreign energy data - separate function since we also call it elsewhere */
 969     reset_foreign_enerdata(enerd);
 970 }