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 "gromacs/legacyheaders/force.h"
  40
  41 #include "config.h"
  42
  43 #include <assert.h>
  44 #include <math.h>
  45 #include <string.h>
  46
  47 #include "gromacs/legacyheaders/coulomb.h"
  48 #include "gromacs/legacyheaders/domdec.h"
  49 #include "gromacs/legacyheaders/gmx_omp_nthreads.h"
  50 #include "gromacs/legacyheaders/macros.h"
  51 #include "gromacs/legacyheaders/mdrun.h"
  52 #include "gromacs/legacyheaders/names.h"
  53 #include "gromacs/legacyheaders/network.h"
  54 #include "gromacs/legacyheaders/nonbonded.h"
  55 #include "gromacs/legacyheaders/nrnb.h"
  56 #include "gromacs/legacyheaders/ns.h"
  57 #include "gromacs/legacyheaders/pme.h"
  58 #include "gromacs/legacyheaders/qmmm.h"
  59 #include "gromacs/legacyheaders/txtdump.h"
  60 #include "gromacs/legacyheaders/typedefs.h"
  61 #include "gromacs/legacyheaders/types/commrec.h"
  62 #include "gromacs/listed-forces/bonded.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     int nthreads_loop gmx_unused;
 122
 123     /* This reduction can run over any number of threads */
 124     nthreads_loop = gmx_omp_nthreads_get(emntBonded);
 125 #pragma omp parallel for num_threads(nthreads_loop) private(t) schedule(static)
 126     for (i = 0; i < n; i++)
 127     {
 128         for (t = 1; t < nthreads; t++)
 129         {
 130             rvec_inc(f[i], f_t[t].f[i]);
 131         }
 132     }
 133     for (t = 1; t < nthreads; t++)
 134     {
 135         *Vcorr_q  += f_t[t].Vcorr_q;
 136         *Vcorr_lj += f_t[t].Vcorr_lj;
 137         *dvdl_q   += f_t[t].dvdl[efptCOUL];
 138         *dvdl_lj  += f_t[t].dvdl[efptVDW];
 139         m_add(vir_q, f_t[t].vir_q, vir_q);
 140         m_add(vir_lj, f_t[t].vir_lj, vir_lj);
 141     }
 142 }
 143
 144 void do_force_lowlevel(t_forcerec *fr,      t_inputrec *ir,
 145                        t_idef     *idef,    t_commrec  *cr,
 146                        t_nrnb     *nrnb,    gmx_wallcycle_t wcycle,
 147                        t_mdatoms  *md,
 148                        rvec       x[],      history_t  *hist,
 149                        rvec       f[],
 150                        rvec       f_longrange[],
 151                        gmx_enerdata_t *enerd,
 152                        t_fcdata   *fcd,
 153                        gmx_localtop_t *top,
 154                        gmx_genborn_t *born,
 155                        gmx_bool       bBornRadii,
 156                        matrix     box,
 157                        t_lambda   *fepvals,
 158                        real       *lambda,
 159                        t_graph    *graph,
 160                        t_blocka   *excl,
 161                        rvec       mu_tot[],
 162                        int        flags,
 163                        float      *cycles_pme)
 164 {
 165     int         i, j;
 166     int         donb_flags;
 167     gmx_bool    bDoEpot, bSB;
 168     int         pme_flags;
 169     matrix      boxs;
 170     rvec        box_size;
 171     t_pbc       pbc;
 172     char        buf[22];
 173     double      clam_i, vlam_i;
 174     real        dvdl_dum[efptNR], dvdl_nb[efptNR], lam_i[efptNR];
 175     real        dvdl_q, dvdl_lj;
 176
 177 #ifdef GMX_MPI
 178     double  t0 = 0.0, t1, t2, t3; /* time measurement for coarse load balancing */
 179 #endif
 180
 181     set_pbc(&pbc, fr->ePBC, box);
 182
 183     /* reset free energy components */
 184     for (i = 0; i < efptNR; i++)
 185     {
 186         dvdl_nb[i]  = 0;
 187         dvdl_dum[i] = 0;
 188     }
 189
 190     /* Reset box */
 191     for (i = 0; (i < DIM); i++)
 192     {
 193         box_size[i] = box[i][i];
 194     }
 195
 196     debug_gmx();
 197
 198     /* do QMMM first if requested */
 199     if (fr->bQMMM)
 200     {
 201         enerd->term[F_EQM] = calculate_QMMM(cr, x, f, fr);
 202     }
 203
 204     /* Call the short range functions all in one go. */
 205
 206 #ifdef GMX_MPI
 207     /*#define TAKETIME ((cr->npmenodes) && (fr->timesteps < 12))*/
 208 #define TAKETIME FALSE
 209     if (TAKETIME)
 210     {
 211         MPI_Barrier(cr->mpi_comm_mygroup);
 212         t0 = MPI_Wtime();
 213     }
 214 #endif
 215
 216     if (ir->nwall)
 217     {
 218         /* foreign lambda component for walls */
 219         real dvdl_walls = do_walls(ir, fr, box, md, x, f, lambda[efptVDW],
 220                                    enerd->grpp.ener[egLJSR], nrnb);
 221         enerd->dvdl_lin[efptVDW] += dvdl_walls;
 222     }
 223
 224     /* If doing GB, reset dvda and calculate the Born radii */
 225     if (ir->implicit_solvent)
 226     {
 227         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 228
 229         for (i = 0; i < born->nr; i++)
 230         {
 231             fr->dvda[i] = 0;
 232         }
 233
 234         if (bBornRadii)
 235         {
 236             calc_gb_rad(cr, fr, ir, top, x, &(fr->gblist), born, md, nrnb);
 237         }
 238
 239         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 240     }
 241
 242     where();
 243     /* We only do non-bonded calculation with group scheme here, the verlet
 244      * calls are done from do_force_cutsVERLET(). */
 245     if (fr->cutoff_scheme == ecutsGROUP && (flags & GMX_FORCE_NONBONDED))
 246     {
 247         donb_flags = 0;
 248         /* Add short-range interactions */
 249         donb_flags |= GMX_NONBONDED_DO_SR;
 250
 251         /* Currently all group scheme kernels always calculate (shift-)forces */
 252         if (flags & GMX_FORCE_FORCES)
 253         {
 254             donb_flags |= GMX_NONBONDED_DO_FORCE;
 255         }
 256         if (flags & GMX_FORCE_VIRIAL)
 257         {
 258             donb_flags |= GMX_NONBONDED_DO_SHIFTFORCE;
 259         }
 260         if (flags & GMX_FORCE_ENERGY)
 261         {
 262             donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 263         }
 264         if (flags & GMX_FORCE_DO_LR)
 265         {
 266             donb_flags |= GMX_NONBONDED_DO_LR;
 267         }
 268
 269         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 270         do_nonbonded(fr, x, f, f_longrange, md, excl,
 271                      &enerd->grpp, nrnb,
 272                      lambda, dvdl_nb, -1, -1, donb_flags);
 273
 274         /* If we do foreign lambda and we have soft-core interactions
 275          * we have to recalculate the (non-linear) energies contributions.
 276          */
 277         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 278         {
 279             for (i = 0; i < enerd->n_lambda; i++)
 280             {
 281                 for (j = 0; j < efptNR; j++)
 282                 {
 283                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 284                 }
 285                 reset_foreign_enerdata(enerd);
 286                 do_nonbonded(fr, x, f, f_longrange, md, excl,
 287                              &(enerd->foreign_grpp), nrnb,
 288                              lam_i, dvdl_dum, -1, -1,
 289                              (donb_flags & ~GMX_NONBONDED_DO_FORCE) | GMX_NONBONDED_DO_FOREIGNLAMBDA);
 290                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 291                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 292             }
 293         }
 294         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 295         where();
 296     }
 297
 298     /* If we are doing GB, calculate bonded forces and apply corrections
 299      * to the solvation forces */
 300     /* MRS: Eventually, many need to include free energy contribution here! */
 301     if (ir->implicit_solvent)
 302     {
 303         wallcycle_sub_start(wcycle, ewcsLISTED);
 304         calc_gb_forces(cr, md, born, top, x, f, fr, idef,
 305                        ir->gb_algorithm, ir->sa_algorithm, nrnb, &pbc, graph, enerd);
 306         wallcycle_sub_stop(wcycle, ewcsLISTED);
 307     }
 308
 309 #ifdef GMX_MPI
 310     if (TAKETIME)
 311     {
 312         t1          = MPI_Wtime();
 313         fr->t_fnbf += t1-t0;
 314     }
 315 #endif
 316
 317     if (fepvals->sc_alpha != 0)
 318     {
 319         enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
 320     }
 321     else
 322     {
 323         enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
 324     }
 325
 326     if (fepvals->sc_alpha != 0)
 327
 328     /* even though coulomb part is linear, we already added it, beacuse we
 329        need to go through the vdw calculation anyway */
 330     {
 331         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 332     }
 333     else
 334     {
 335         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 336     }
 337
 338     debug_gmx();
 339
 340
 341     if (debug)
 342     {
 343         pr_rvecs(debug, 0, "fshift after SR", fr->fshift, SHIFTS);
 344     }
 345
 346     /* Shift the coordinates. Must be done before listed forces and PPPM,
 347      * but is also necessary for SHAKE and update, therefore it can NOT
 348      * go when no listed forces have to be evaluated.
 349      */
 350
 351     /* Here sometimes we would not need to shift with NBFonly,
 352      * but we do so anyhow for consistency of the returned coordinates.
 353      */
 354     if (graph)
 355     {
 356         shift_self(graph, box, x);
 357         if (TRICLINIC(box))
 358         {
 359             inc_nrnb(nrnb, eNR_SHIFTX, 2*graph->nnodes);
 360         }
 361         else
 362         {
 363             inc_nrnb(nrnb, eNR_SHIFTX, graph->nnodes);
 364         }
 365     }
 366     /* Check whether we need to do listed interactions or correct for exclusions */
 367     if (fr->bMolPBC &&
 368         ((flags & GMX_FORCE_LISTED)
 369          || EEL_RF(fr->eeltype) || EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype)))
 370     {
 371         /* Since all atoms are in the rectangular or triclinic unit-cell,
 372          * only single box vector shifts (2 in x) are required.
 373          */
 374         set_pbc_dd(&pbc, fr->ePBC, cr->dd, TRUE, box);
 375     }
 376     debug_gmx();
 377
 378     if (flags & GMX_FORCE_LISTED)
 379     {
 380         wallcycle_sub_start(wcycle, ewcsLISTED);
 381         calc_bonds(cr->ms,
 382                    idef, (const rvec *) x, hist, f, fr, &pbc, graph, enerd, nrnb, lambda, md, fcd,
 383                    DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL,
 384                    flags);
 385
 386         /* Check if we have to determine energy differences
 387          * at foreign lambda's.
 388          */
 389         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) &&
 390             idef->ilsort != ilsortNO_FE)
 391         {
 392             if (idef->ilsort != ilsortFE_SORTED)
 393             {
 394                 gmx_incons("The bonded interactions are not sorted for free energy");
 395             }
 396             for (i = 0; i < enerd->n_lambda; i++)
 397             {
 398                 reset_foreign_enerdata(enerd);
 399                 for (j = 0; j < efptNR; j++)
 400                 {
 401                     lam_i[j] = (i == 0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 402                 }
 403                 calc_bonds_lambda(idef, (const rvec *) x, fr, &pbc, graph, &(enerd->foreign_grpp), enerd->foreign_term, nrnb, lam_i, md,
 404                                   fcd, DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL);
 405                 sum_epot(&(enerd->foreign_grpp), enerd->foreign_term);
 406                 enerd->enerpart_lambda[i] += enerd->foreign_term[F_EPOT];
 407             }
 408         }
 409         debug_gmx();
 410
 411         wallcycle_sub_stop(wcycle, ewcsLISTED);
 412     }
 413
 414     where();
 415
 416     *cycles_pme = 0;
 417     clear_mat(fr->vir_el_recip);
 418     clear_mat(fr->vir_lj_recip);
 419
 420     /* Do long-range electrostatics and/or LJ-PME, including related short-range
 421      * corrections.
 422      */
 423     if (EEL_FULL(fr->eeltype) || EVDW_PME(fr->vdwtype))
 424     {
 425         real Vlr               = 0, Vcorr = 0;
 426         real dvdl_long_range   = 0;
 427         int  status            = 0;
 428         real Vlr_q             = 0, Vlr_lj = 0, Vcorr_q = 0, Vcorr_lj = 0;
 429         real dvdl_long_range_q = 0, dvdl_long_range_lj = 0;
 430
 431         bSB = (ir->nwall == 2);
 432         if (bSB)
 433         {
 434             copy_mat(box, boxs);
 435             svmul(ir->wall_ewald_zfac, boxs[ZZ], boxs[ZZ]);
 436             box_size[ZZ] *= ir->wall_ewald_zfac;
 437         }
 438
 439         if (EEL_PME_EWALD(fr->eeltype) || EVDW_PME(fr->vdwtype))
 440         {
 441             real dvdl_long_range_correction_q   = 0;
 442             real dvdl_long_range_correction_lj  = 0;
 443             /* With the Verlet scheme exclusion forces are calculated
 444              * in the non-bonded kernel.
 445              */
 446             /* The TPI molecule does not have exclusions with the rest
 447              * of the system and no intra-molecular PME grid
 448              * contributions will be calculated in
 449              * gmx_pme_calc_energy.
 450              */
 451             if ((ir->cutoff_scheme == ecutsGROUP && fr->n_tpi == 0) ||
 452                 ir->ewald_geometry != eewg3D ||
 453                 ir->epsilon_surface != 0)
 454             {
 455                 int nthreads, t;
 456
 457                 wallcycle_sub_start(wcycle, ewcsEWALD_CORRECTION);
 458
 459                 if (fr->n_tpi > 0)
 460                 {
 461                     gmx_fatal(FARGS, "TPI with PME currently only works in a 3D geometry with tin-foil boundary conditions");
 462                 }
 463
 464                 nthreads = gmx_omp_nthreads_get(emntBonded);
 465 #pragma omp parallel for num_threads(nthreads) schedule(static)
 466                 for (t = 0; t < nthreads; t++)
 467                 {
 468                     int     s, e, i;
 469                     rvec   *fnv;
 470                     tensor *vir_q, *vir_lj;
 471                     real   *Vcorrt_q, *Vcorrt_lj, *dvdlt_q, *dvdlt_lj;
 472                     if (t == 0)
 473                     {
 474                         fnv       = fr->f_novirsum;
 475                         vir_q     = &fr->vir_el_recip;
 476                         vir_lj    = &fr->vir_lj_recip;
 477                         Vcorrt_q  = &Vcorr_q;
 478                         Vcorrt_lj = &Vcorr_lj;
 479                         dvdlt_q   = &dvdl_long_range_correction_q;
 480                         dvdlt_lj  = &dvdl_long_range_correction_lj;
 481                     }
 482                     else
 483                     {
 484                         fnv       = fr->f_t[t].f;
 485                         vir_q     = &fr->f_t[t].vir_q;
 486                         vir_lj    = &fr->f_t[t].vir_lj;
 487                         Vcorrt_q  = &fr->f_t[t].Vcorr_q;
 488                         Vcorrt_lj = &fr->f_t[t].Vcorr_lj;
 489                         dvdlt_q   = &fr->f_t[t].dvdl[efptCOUL];
 490                         dvdlt_lj  = &fr->f_t[t].dvdl[efptVDW];
 491                         for (i = 0; i < fr->natoms_force; i++)
 492                         {
 493                             clear_rvec(fnv[i]);
 494                         }
 495                         clear_mat(*vir_q);
 496                         clear_mat(*vir_lj);
 497                     }
 498                     *dvdlt_q  = 0;
 499                     *dvdlt_lj = 0;
 500
 501                     ewald_LRcorrection(fr->excl_load[t], fr->excl_load[t+1],
 502                                        cr, t, fr,
 503                                        md->chargeA,
 504                                        md->nChargePerturbed ? md->chargeB : NULL,
 505                                        md->sqrt_c6A,
 506                                        md->nTypePerturbed ? md->sqrt_c6B : NULL,
 507                                        md->sigmaA,
 508                                        md->nTypePerturbed ? md->sigmaB : NULL,
 509                                        md->sigma3A,
 510                                        md->nTypePerturbed ? md->sigma3B : NULL,
 511                                        ir->cutoff_scheme != ecutsVERLET,
 512                                        excl, x, bSB ? boxs : box, mu_tot,
 513                                        ir->ewald_geometry,
 514                                        ir->epsilon_surface,
 515                                        fnv, *vir_q, *vir_lj,
 516                                        Vcorrt_q, Vcorrt_lj,
 517                                        lambda[efptCOUL], lambda[efptVDW],
 518                                        dvdlt_q, dvdlt_lj);
 519                 }
 520                 if (nthreads > 1)
 521                 {
 522                     reduce_thread_forces(fr->natoms_force, fr->f_novirsum,
 523                                          fr->vir_el_recip, fr->vir_lj_recip,
 524                                          &Vcorr_q, &Vcorr_lj,
 525                                          &dvdl_long_range_correction_q,
 526                                          &dvdl_long_range_correction_lj,
 527                                          nthreads, fr->f_t);
 528                 }
 529                 wallcycle_sub_stop(wcycle, ewcsEWALD_CORRECTION);
 530             }
 531
 532             if (EEL_PME_EWALD(fr->eeltype) && fr->n_tpi == 0)
 533             {
 534                 Vcorr_q += ewald_charge_correction(cr, fr, lambda[efptCOUL], box,
 535                                                    &dvdl_long_range_correction_q,
 536                                                    fr->vir_el_recip);
 537             }
 538
 539             enerd->dvdl_lin[efptCOUL] += dvdl_long_range_correction_q;
 540             enerd->dvdl_lin[efptVDW]  += dvdl_long_range_correction_lj;
 541
 542             if ((EEL_PME(fr->eeltype) || EVDW_PME(fr->vdwtype)) && (cr->duty & DUTY_PME))
 543             {
 544                 /* Do reciprocal PME for Coulomb and/or LJ. */
 545                 assert(fr->n_tpi >= 0);
 546                 if (fr->n_tpi == 0 || (flags & GMX_FORCE_STATECHANGED))
 547                 {
 548                     pme_flags = GMX_PME_SPREAD | GMX_PME_SOLVE;
 549                     if (EEL_PME(fr->eeltype))
 550                     {
 551                         pme_flags     |= GMX_PME_DO_COULOMB;
 552                     }
 553                     if (EVDW_PME(fr->vdwtype))
 554                     {
 555                         pme_flags |= GMX_PME_DO_LJ;
 556                     }
 557                     if (flags & GMX_FORCE_FORCES)
 558                     {
 559                         pme_flags |= GMX_PME_CALC_F;
 560                     }
 561                     if (flags & GMX_FORCE_VIRIAL)
 562                     {
 563                         pme_flags |= GMX_PME_CALC_ENER_VIR;
 564                     }
 565                     if (fr->n_tpi > 0)
 566                     {
 567                         /* We don't calculate f, but we do want the potential */
 568                         pme_flags |= GMX_PME_CALC_POT;
 569                     }
 570                     wallcycle_start(wcycle, ewcPMEMESH);
 571                     status = gmx_pme_do(fr->pmedata,
 572                                         0, md->homenr - fr->n_tpi,
 573                                         x, fr->f_novirsum,
 574                                         md->chargeA, md->chargeB,
 575                                         md->sqrt_c6A, md->sqrt_c6B,
 576                                         md->sigmaA, md->sigmaB,
 577                                         bSB ? boxs : box, cr,
 578                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_x(cr->dd) : 0,
 579                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_y(cr->dd) : 0,
 580                                         nrnb, wcycle,
 581                                         fr->vir_el_recip, fr->ewaldcoeff_q,
 582                                         fr->vir_lj_recip, fr->ewaldcoeff_lj,
 583                                         &Vlr_q, &Vlr_lj,
 584                                         lambda[efptCOUL], lambda[efptVDW],
 585                                         &dvdl_long_range_q, &dvdl_long_range_lj, pme_flags);
 586                     *cycles_pme = wallcycle_stop(wcycle, ewcPMEMESH);
 587                     if (status != 0)
 588                     {
 589                         gmx_fatal(FARGS, "Error %d in reciprocal PME routine", status);
 590                     }
 591                     /* We should try to do as little computation after
 592                      * this as possible, because parallel PME synchronizes
 593                      * the nodes, so we want all load imbalance of the
 594                      * rest of the force calculation to be before the PME
 595                      * call.  DD load balancing is done on the whole time
 596                      * of the force call (without PME).
 597                      */
 598                 }
 599                 if (fr->n_tpi > 0)
 600                 {
 601                     if (EVDW_PME(ir->vdwtype))
 602                     {
 603
 604                         gmx_fatal(FARGS, "Test particle insertion not implemented with LJ-PME");
 605                     }
 606                     /* Determine the PME grid energy of the test molecule
 607                      * with the PME grid potential of the other charges.
 608                      */
 609                     gmx_pme_calc_energy(fr->pmedata, fr->n_tpi,
 610                                         x + md->homenr - fr->n_tpi,
 611                                         md->chargeA + md->homenr - fr->n_tpi,
 612                                         &Vlr_q);
 613                 }
 614             }
 615         }
 616
 617         if (!EEL_PME(fr->eeltype) && EEL_PME_EWALD(fr->eeltype))
 618         {
 619             Vlr_q = do_ewald(ir, x, fr->f_novirsum,
 620                              md->chargeA, md->chargeB,
 621                              box_size, cr, md->homenr,
 622                              fr->vir_el_recip, fr->ewaldcoeff_q,
 623                              lambda[efptCOUL], &dvdl_long_range_q, fr->ewald_table);
 624         }
 625
 626         /* Note that with separate PME nodes we get the real energies later */
 627         enerd->dvdl_lin[efptCOUL] += dvdl_long_range_q;
 628         enerd->dvdl_lin[efptVDW]  += dvdl_long_range_lj;
 629         enerd->term[F_COUL_RECIP]  = Vlr_q + Vcorr_q;
 630         enerd->term[F_LJ_RECIP]    = Vlr_lj + Vcorr_lj;
 631         if (debug)
 632         {
 633             fprintf(debug, "Vlr_q = %g, Vcorr_q = %g, Vlr_corr_q = %g\n",
 634                     Vlr_q, Vcorr_q, enerd->term[F_COUL_RECIP]);
 635             pr_rvecs(debug, 0, "vir_el_recip after corr", fr->vir_el_recip, DIM);
 636             pr_rvecs(debug, 0, "fshift after LR Corrections", fr->fshift, SHIFTS);
 637             fprintf(debug, "Vlr_lj: %g, Vcorr_lj = %g, Vlr_corr_lj = %g\n",
 638                     Vlr_lj, Vcorr_lj, enerd->term[F_LJ_RECIP]);
 639             pr_rvecs(debug, 0, "vir_lj_recip after corr", fr->vir_lj_recip, DIM);
 640         }
 641     }
 642     else
 643     {
 644         /* Is there a reaction-field exclusion correction needed? */
 645         if (EEL_RF(fr->eeltype) && eelRF_NEC != fr->eeltype)
 646         {
 647             /* With the Verlet scheme, exclusion forces are calculated
 648              * in the non-bonded kernel.
 649              */
 650             if (ir->cutoff_scheme != ecutsVERLET)
 651             {
 652                 real dvdl_rf_excl      = 0;
 653                 enerd->term[F_RF_EXCL] =
 654                     RF_excl_correction(fr, graph, md, excl, x, f,
 655                                        fr->fshift, &pbc, lambda[efptCOUL], &dvdl_rf_excl);
 656
 657                 enerd->dvdl_lin[efptCOUL] += dvdl_rf_excl;
 658             }
 659         }
 660     }
 661     where();
 662     debug_gmx();
 663
 664     if (debug)
 665     {
 666         print_nrnb(debug, nrnb);
 667     }
 668     debug_gmx();
 669
 670 #ifdef GMX_MPI
 671     if (TAKETIME)
 672     {
 673         t2 = MPI_Wtime();
 674         MPI_Barrier(cr->mpi_comm_mygroup);
 675         t3          = MPI_Wtime();
 676         fr->t_wait += t3-t2;
 677         if (fr->timesteps == 11)
 678         {
 679             fprintf(stderr, "* PP load balancing info: rank %d, step %s, rel wait time=%3.0f%% , load string value: %7.2f\n",
 680                     cr->nodeid, gmx_step_str(fr->timesteps, buf),
 681                     100*fr->t_wait/(fr->t_wait+fr->t_fnbf),
 682                     (fr->t_fnbf+fr->t_wait)/fr->t_fnbf);
 683         }
 684         fr->timesteps++;
 685     }
 686 #endif
 687
 688     if (debug)
 689     {
 690         pr_rvecs(debug, 0, "fshift after bondeds", fr->fshift, SHIFTS);
 691     }
 692
 693 }
 694
 695 void init_enerdata(int ngener, int n_lambda, gmx_enerdata_t *enerd)
 696 {
 697     int i, n2;
 698
 699     for (i = 0; i < F_NRE; i++)
 700     {
 701         enerd->term[i]         = 0;
 702         enerd->foreign_term[i] = 0;
 703     }
 704
 705
 706     for (i = 0; i < efptNR; i++)
 707     {
 708         enerd->dvdl_lin[i]     = 0;
 709         enerd->dvdl_nonlin[i]  = 0;
 710     }
 711
 712     n2 = ngener*ngener;
 713     if (debug)
 714     {
 715         fprintf(debug, "Creating %d sized group matrix for energies\n", n2);
 716     }
 717     enerd->grpp.nener         = n2;
 718     enerd->foreign_grpp.nener = n2;
 719     for (i = 0; (i < egNR); i++)
 720     {
 721         snew(enerd->grpp.ener[i], n2);
 722         snew(enerd->foreign_grpp.ener[i], n2);
 723     }
 724
 725     if (n_lambda)
 726     {
 727         enerd->n_lambda = 1 + n_lambda;
 728         snew(enerd->enerpart_lambda, enerd->n_lambda);
 729     }
 730     else
 731     {
 732         enerd->n_lambda = 0;
 733     }
 734 }
 735
 736 void destroy_enerdata(gmx_enerdata_t *enerd)
 737 {
 738     int i;
 739
 740     for (i = 0; (i < egNR); i++)
 741     {
 742         sfree(enerd->grpp.ener[i]);
 743     }
 744
 745     for (i = 0; (i < egNR); i++)
 746     {
 747         sfree(enerd->foreign_grpp.ener[i]);
 748     }
 749
 750     if (enerd->n_lambda)
 751     {
 752         sfree(enerd->enerpart_lambda);
 753     }
 754 }
 755
 756 static real sum_v(int n, real v[])
 757 {
 758     real t;
 759     int  i;
 760
 761     t = 0.0;
 762     for (i = 0; (i < n); i++)
 763     {
 764         t = t + v[i];
 765     }
 766
 767     return t;
 768 }
 769
 770 void sum_epot(gmx_grppairener_t *grpp, real *epot)
 771 {
 772     int i;
 773
 774     /* Accumulate energies */
 775     epot[F_COUL_SR]  = sum_v(grpp->nener, grpp->ener[egCOULSR]);
 776     epot[F_LJ]       = sum_v(grpp->nener, grpp->ener[egLJSR]);
 777     epot[F_LJ14]     = sum_v(grpp->nener, grpp->ener[egLJ14]);
 778     epot[F_COUL14]   = sum_v(grpp->nener, grpp->ener[egCOUL14]);
 779     epot[F_COUL_LR]  = sum_v(grpp->nener, grpp->ener[egCOULLR]);
 780     epot[F_LJ_LR]    = sum_v(grpp->nener, grpp->ener[egLJLR]);
 781     /* We have already added 1-2,1-3, and 1-4 terms to F_GBPOL */
 782     epot[F_GBPOL]   += sum_v(grpp->nener, grpp->ener[egGB]);
 783
 784 /* lattice part of LR doesnt belong to any group
 785  * and has been added earlier
 786  */
 787     epot[F_BHAM]     = sum_v(grpp->nener, grpp->ener[egBHAMSR]);
 788     epot[F_BHAM_LR]  = sum_v(grpp->nener, grpp->ener[egBHAMLR]);
 789
 790     epot[F_EPOT] = 0;
 791     for (i = 0; (i < F_EPOT); i++)
 792     {
 793         if (i != F_DISRESVIOL && i != F_ORIRESDEV)
 794         {
 795             epot[F_EPOT] += epot[i];
 796         }
 797     }
 798 }
 799
 800 void sum_dhdl(gmx_enerdata_t *enerd, real *lambda, t_lambda *fepvals)
 801 {
 802     int    i, j, index;
 803     double dlam;
 804
 805     enerd->dvdl_lin[efptVDW] += enerd->term[F_DVDL_VDW];  /* include dispersion correction */
 806     enerd->term[F_DVDL]       = 0.0;
 807     for (i = 0; i < efptNR; i++)
 808     {
 809         if (fepvals->separate_dvdl[i])
 810         {
 811             /* could this be done more readably/compactly? */
 812             switch (i)
 813             {
 814                 case (efptMASS):
 815                     index = F_DKDL;
 816                     break;
 817                 case (efptCOUL):
 818                     index = F_DVDL_COUL;
 819                     break;
 820                 case (efptVDW):
 821                     index = F_DVDL_VDW;
 822                     break;
 823                 case (efptBONDED):
 824                     index = F_DVDL_BONDED;
 825                     break;
 826                 case (efptRESTRAINT):
 827                     index = F_DVDL_RESTRAINT;
 828                     break;
 829                 default:
 830                     index = F_DVDL;
 831                     break;
 832             }
 833             enerd->term[index] = enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 834             if (debug)
 835             {
 836                 fprintf(debug, "dvdl-%s[%2d]: %f: non-linear %f + linear %f\n",
 837                         efpt_names[i], i, enerd->term[index], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 838             }
 839         }
 840         else
 841         {
 842             enerd->term[F_DVDL] += enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 843             if (debug)
 844             {
 845                 fprintf(debug, "dvd-%sl[%2d]: %f: non-linear %f + linear %f\n",
 846                         efpt_names[0], i, enerd->term[F_DVDL], enerd->dvdl_nonlin[i], enerd->dvdl_lin[i]);
 847             }
 848         }
 849     }
 850
 851     /* Notes on the foreign lambda free energy difference evaluation:
 852      * Adding the potential and ekin terms that depend linearly on lambda
 853      * as delta lam * dvdl to the energy differences is exact.
 854      * For the constraints this is not exact, but we have no other option
 855      * without literally changing the lengths and reevaluating the energies at each step.
 856      * (try to remedy this post 4.6 - MRS)
 857      * For the non-bonded LR term we assume that the soft-core (if present)
 858      * no longer affects the energy beyond the short-range cut-off,
 859      * which is a very good approximation (except for exotic settings).
 860      * (investigate how to overcome this post 4.6 - MRS)
 861      */
 862     if (fepvals->separate_dvdl[efptBONDED])
 863     {
 864         enerd->term[F_DVDL_BONDED] += enerd->term[F_DVDL_CONSTR];
 865     }
 866     else
 867     {
 868         enerd->term[F_DVDL] += enerd->term[F_DVDL_CONSTR];
 869     }
 870     enerd->term[F_DVDL_CONSTR] = 0;
 871
 872     for (i = 0; i < fepvals->n_lambda; i++)
 873     {
 874         /* note we are iterating over fepvals here!
 875            For the current lam, dlam = 0 automatically,
 876            so we don't need to add anything to the
 877            enerd->enerpart_lambda[0] */
 878
 879         /* we don't need to worry about dvdl_lin contributions to dE at
 880            current lambda, because the contributions to the current
 881            lambda are automatically zeroed */
 882
 883         for (j = 0; j < efptNR; j++)
 884         {
 885             /* Note that this loop is over all dhdl components, not just the separated ones */
 886             dlam = (fepvals->all_lambda[j][i]-lambda[j]);
 887             enerd->enerpart_lambda[i+1] += dlam*enerd->dvdl_lin[j];
 888             if (debug)
 889             {
 890                 fprintf(debug, "enerdiff lam %g: (%15s), non-linear %f linear %f*%f\n",
 891                         fepvals->all_lambda[j][i], efpt_names[j],
 892                         (enerd->enerpart_lambda[i+1] - enerd->enerpart_lambda[0]),
 893                         dlam, enerd->dvdl_lin[j]);
 894             }
 895         }
 896     }
 897 }
 898
 899
 900 void reset_foreign_enerdata(gmx_enerdata_t *enerd)
 901 {
 902     int  i, j;
 903
 904     /* First reset all foreign energy components.  Foreign energies always called on
 905        neighbor search steps */
 906     for (i = 0; (i < egNR); i++)
 907     {
 908         for (j = 0; (j < enerd->grpp.nener); j++)
 909         {
 910             enerd->foreign_grpp.ener[i][j] = 0.0;
 911         }
 912     }
 913
 914     /* potential energy components */
 915     for (i = 0; (i <= F_EPOT); i++)
 916     {
 917         enerd->foreign_term[i] = 0.0;
 918     }
 919 }
 920
 921 void reset_enerdata(t_forcerec *fr, gmx_bool bNS,
 922                     gmx_enerdata_t *enerd,
 923                     gmx_bool bMaster)
 924 {
 925     gmx_bool bKeepLR;
 926     int      i, j;
 927
 928     /* First reset all energy components, except for the long range terms
 929      * on the master at non neighbor search steps, since the long range
 930      * terms have already been summed at the last neighbor search step.
 931      */
 932     bKeepLR = (fr->bTwinRange && !bNS);
 933     for (i = 0; (i < egNR); i++)
 934     {
 935         if (!(bKeepLR && bMaster && (i == egCOULLR || i == egLJLR)))
 936         {
 937             for (j = 0; (j < enerd->grpp.nener); j++)
 938             {
 939                 enerd->grpp.ener[i][j] = 0.0;
 940             }
 941         }
 942     }
 943     for (i = 0; i < efptNR; i++)
 944     {
 945         enerd->dvdl_lin[i]    = 0.0;
 946         enerd->dvdl_nonlin[i] = 0.0;
 947     }
 948
 949     /* Normal potential energy components */
 950     for (i = 0; (i <= F_EPOT); i++)
 951     {
 952         enerd->term[i] = 0.0;
 953     }
 954     /* Initialize the dVdlambda term with the long range contribution */
 955     /* Initialize the dvdl term with the long range contribution */
 956     enerd->term[F_DVDL]            = 0.0;
 957     enerd->term[F_DVDL_COUL]       = 0.0;
 958     enerd->term[F_DVDL_VDW]        = 0.0;
 959     enerd->term[F_DVDL_BONDED]     = 0.0;
 960     enerd->term[F_DVDL_RESTRAINT]  = 0.0;
 961     enerd->term[F_DKDL]            = 0.0;
 962     if (enerd->n_lambda > 0)
 963     {
 964         for (i = 0; i < enerd->n_lambda; i++)
 965         {
 966             enerd->enerpart_lambda[i] = 0.0;
 967         }
 968     }
 969     /* reset foreign energy data - separate function since we also call it elsewhere */
 970     reset_foreign_enerdata(enerd);
 971 }