src/gromacs/mdlib/force.c

   1 /* -*- mode: c; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; c-file-style: "stroustrup"; -*-
   2  *
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  *          GROningen MAchine for Chemical Simulations
   9  *
  10  *                        VERSION 3.2.0
  11  * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
  12  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
  13  * Copyright (c) 2001-2004, The GROMACS development team,
  14  * check out http://www.gromacs.org for more information.
  15
  16  * This program is free software; you can redistribute it and/or
  17  * modify it under the terms of the GNU General Public License
  18  * as published by the Free Software Foundation; either version 2
  19  * of the License, or (at your option) any later version.
  20  *
  21  * If you want to redistribute modifications, please consider that
  22  * scientific software is very special. Version control is crucial -
  23  * bugs must be traceable. We will be happy to consider code for
  24  * inclusion in the official distribution, but derived work must not
  25  * be called official GROMACS. Details are found in the README & COPYING
  26  * files - if they are missing, get the official version at www.gromacs.org.
  27  *
  28  * To help us fund GROMACS development, we humbly ask that you cite
  29  * the papers on the package - you can find them in the top README file.
  30  *
  31  * For more info, check our website at http://www.gromacs.org
  32  *
  33  * And Hey:
  34  * GROwing Monsters And Cloning Shrimps
  35  */
  36 #ifdef HAVE_CONFIG_H
  37 #include <config.h>
  38 #endif
  39
  40 #include <math.h>
  41 #include <string.h>
  42 #include <assert.h>
  43 #include "sysstuff.h"
  44 #include "typedefs.h"
  45 #include "macros.h"
  46 #include "smalloc.h"
  47 #include "macros.h"
  48 #include "physics.h"
  49 #include "force.h"
  50 #include "nonbonded.h"
  51 #include "names.h"
  52 #include "network.h"
  53 #include "pbc.h"
  54 #include "ns.h"
  55 #include "nrnb.h"
  56 #include "bondf.h"
  57 #include "mshift.h"
  58 #include "txtdump.h"
  59 #include "coulomb.h"
  60 #include "pme.h"
  61 #include "mdrun.h"
  62 #include "domdec.h"
  63 #include "partdec.h"
  64 #include "qmmm.h"
  65 #include "gmx_omp_nthreads.h"
  66
  67
  68 void ns(FILE *fp,
  69         t_forcerec *fr,
  70         rvec       x[],
  71         matrix     box,
  72         gmx_groups_t *groups,
  73         t_grpopts  *opts,
  74         gmx_localtop_t *top,
  75         t_mdatoms  *md,
  76         t_commrec  *cr,
  77         t_nrnb     *nrnb,
  78         real       *lambda,
  79         real       *dvdlambda,
  80         gmx_grppairener_t *grppener,
  81         gmx_bool       bFillGrid,
  82         gmx_bool       bDoLongRangeNS)
  83 {
  84   char   *ptr;
  85   int    nsearch;
  86
  87
  88   if (!fr->ns.nblist_initialized)
  89   {
  90       init_neighbor_list(fp, fr, md->homenr);
  91   }
  92
  93   if (fr->bTwinRange)
  94     fr->nlr=0;
  95
  96     nsearch = search_neighbours(fp,fr,x,box,top,groups,cr,nrnb,md,
  97                                 lambda,dvdlambda,grppener,
  98                                 bFillGrid,bDoLongRangeNS,TRUE);
  99   if (debug)
 100     fprintf(debug,"nsearch = %d\n",nsearch);
 101
 102   /* Check whether we have to do dynamic load balancing */
 103   /*if ((nsb->nstDlb > 0) && (mod(step,nsb->nstDlb) == 0))
 104     count_nb(cr,nsb,&(top->blocks[ebCGS]),nns,fr->nlr,
 105     &(top->idef),opts->ngener);
 106   */
 107   if (fr->ns.dump_nl > 0)
 108     dump_nblist(fp,cr,fr,fr->ns.dump_nl);
 109 }
 110
 111 static void reduce_thread_forces(int n,rvec *f,
 112                                  tensor vir,
 113                                  real *Vcorr,
 114                                  int efpt_ind,real *dvdl,
 115                                  int nthreads,f_thread_t *f_t)
 116 {
 117     int t,i;
 118
 119     /* This reduction can run over any number of threads */
 120 #pragma omp parallel for num_threads(gmx_omp_nthreads_get(emntBonded)) private(t) schedule(static)
 121     for(i=0; i<n; i++)
 122     {
 123         for(t=1; t<nthreads; t++)
 124         {
 125             rvec_inc(f[i],f_t[t].f[i]);
 126         }
 127     }
 128     for(t=1; t<nthreads; t++)
 129     {
 130         *Vcorr += f_t[t].Vcorr;
 131         *dvdl  += f_t[t].dvdl[efpt_ind];
 132         m_add(vir,f_t[t].vir,vir);
 133     }
 134 }
 135
 136 void do_force_lowlevel(FILE       *fplog,   gmx_large_int_t step,
 137                        t_forcerec *fr,      t_inputrec *ir,
 138                        t_idef     *idef,    t_commrec  *cr,
 139                        t_nrnb     *nrnb,    gmx_wallcycle_t wcycle,
 140                        t_mdatoms  *md,
 141                        t_grpopts  *opts,
 142                        rvec       x[],      history_t  *hist,
 143                        rvec       f[],
 144                        rvec       f_longrange[],
 145                        gmx_enerdata_t *enerd,
 146                        t_fcdata   *fcd,
 147                        gmx_mtop_t     *mtop,
 148                        gmx_localtop_t *top,
 149                        gmx_genborn_t *born,
 150                        t_atomtypes *atype,
 151                        gmx_bool       bBornRadii,
 152                        matrix     box,
 153                        t_lambda   *fepvals,
 154                        real       *lambda,
 155                        t_graph    *graph,
 156                        t_blocka   *excl,
 157                        rvec       mu_tot[],
 158                        int        flags,
 159                        float      *cycles_pme)
 160 {
 161     int     i,j,status;
 162     int     donb_flags;
 163     gmx_bool    bDoEpot,bSepDVDL,bSB;
 164     int     pme_flags;
 165     matrix  boxs;
 166     rvec    box_size;
 167     real    Vsr,Vlr,Vcorr=0;
 168     t_pbc   pbc;
 169     real    dvdgb;
 170     char    buf[22];
 171     gmx_enerdata_t ed_lam;
 172     double  clam_i,vlam_i;
 173     real    dvdl_dum[efptNR], dvdl, dvdl_nb[efptNR], lam_i[efptNR];
 174     real    dvdlsum;
 175
 176 #ifdef GMX_MPI
 177     double  t0=0.0,t1,t2,t3; /* time measurement for coarse load balancing */
 178 #endif
 179
 180 #define PRINT_SEPDVDL(s,v,dvdlambda) if (bSepDVDL) fprintf(fplog,sepdvdlformat,s,v,dvdlambda);
 181
 182
 183     set_pbc(&pbc,fr->ePBC,box);
 184
 185     /* reset free energy components */
 186     for (i=0;i<efptNR;i++)
 187     {
 188         dvdl_nb[i]  = 0;
 189         dvdl_dum[i] = 0;
 190     }
 191
 192     /* Reset box */
 193     for(i=0; (i<DIM); i++)
 194     {
 195         box_size[i]=box[i][i];
 196     }
 197
 198     bSepDVDL=(fr->bSepDVDL && do_per_step(step,ir->nstlog));
 199     debug_gmx();
 200
 201     /* do QMMM first if requested */
 202     if(fr->bQMMM)
 203     {
 204         enerd->term[F_EQM] = calculate_QMMM(cr,x,f,fr,md);
 205     }
 206
 207     if (bSepDVDL)
 208     {
 209         fprintf(fplog,"Step %s: non-bonded V and dVdl for node %d:\n",
 210                 gmx_step_str(step,buf),cr->nodeid);
 211     }
 212
 213     /* Call the short range functions all in one go. */
 214
 215 #ifdef GMX_MPI
 216     /*#define TAKETIME ((cr->npmenodes) && (fr->timesteps < 12))*/
 217 #define TAKETIME FALSE
 218     if (TAKETIME)
 219     {
 220         MPI_Barrier(cr->mpi_comm_mygroup);
 221         t0=MPI_Wtime();
 222     }
 223 #endif
 224
 225     if (ir->nwall)
 226     {
 227         /* foreign lambda component for walls */
 228         dvdl = do_walls(ir,fr,box,md,x,f,lambda[efptVDW],
 229                         enerd->grpp.ener[egLJSR],nrnb);
 230         PRINT_SEPDVDL("Walls",0.0,dvdl);
 231         enerd->dvdl_lin[efptVDW] += dvdl;
 232     }
 233
 234         /* If doing GB, reset dvda and calculate the Born radii */
 235         if (ir->implicit_solvent)
 236         {
 237         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 238
 239                 for(i=0;i<born->nr;i++)
 240                 {
 241                         fr->dvda[i]=0;
 242                 }
 243
 244                 if(bBornRadii)
 245                 {
 246                         calc_gb_rad(cr,fr,ir,top,atype,x,&(fr->gblist),born,md,nrnb);
 247                 }
 248
 249         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 250         }
 251
 252     where();
 253     /* We only do non-bonded calculation with group scheme here, the verlet
 254      * calls are done from do_force_cutsVERLET(). */
 255     if (fr->cutoff_scheme == ecutsGROUP && (flags & GMX_FORCE_NONBONDED))
 256     {
 257         donb_flags = 0;
 258         /* Add short-range interactions */
 259         donb_flags |= GMX_NONBONDED_DO_SR;
 260
 261         if (flags & GMX_FORCE_FORCES)
 262         {
 263             donb_flags |= GMX_NONBONDED_DO_FORCE;
 264         }
 265         if (flags & GMX_FORCE_ENERGY)
 266         {
 267             donb_flags |= GMX_NONBONDED_DO_POTENTIAL;
 268         }
 269         if (flags & GMX_FORCE_DO_LR)
 270         {
 271             donb_flags |= GMX_NONBONDED_DO_LR;
 272         }
 273
 274         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 275         do_nonbonded(cr,fr,x,f,f_longrange,md,excl,
 276                     &enerd->grpp,box_size,nrnb,
 277                     lambda,dvdl_nb,-1,-1,donb_flags);
 278         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 279     }
 280
 281     /* If we do foreign lambda and we have soft-core interactions
 282      * we have to recalculate the (non-linear) energies contributions.
 283      */
 284     if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) && fepvals->sc_alpha != 0)
 285     {
 286         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 287         init_enerdata(mtop->groups.grps[egcENER].nr,fepvals->n_lambda,&ed_lam);
 288
 289         for(i=0; i<enerd->n_lambda; i++)
 290         {
 291             for (j=0;j<efptNR;j++)
 292             {
 293                 lam_i[j] = (i==0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 294             }
 295             reset_enerdata(&ir->opts,fr,TRUE,&ed_lam,FALSE);
 296             do_nonbonded(cr,fr,x,f,f_longrange,md,excl,
 297                          &(ed_lam.grpp), box_size,nrnb,
 298                          lam_i,dvdl_dum,-1,-1,
 299                          GMX_NONBONDED_DO_FOREIGNLAMBDA | GMX_NONBONDED_DO_SR);
 300             sum_epot(&ir->opts,&ed_lam);
 301             enerd->enerpart_lambda[i] += ed_lam.term[F_EPOT];
 302         }
 303         destroy_enerdata(&ed_lam);
 304         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 305     }
 306     where();
 307
 308         /* If we are doing GB, calculate bonded forces and apply corrections
 309          * to the solvation forces */
 310     /* MRS: Eventually, many need to include free energy contribution here! */
 311         if (ir->implicit_solvent)
 312     {
 313         wallcycle_sub_start(wcycle, ewcsBONDED);
 314                 calc_gb_forces(cr,md,born,top,atype,x,f,fr,idef,
 315                        ir->gb_algorithm,ir->sa_algorithm,nrnb,bBornRadii,&pbc,graph,enerd);
 316         wallcycle_sub_stop(wcycle, ewcsBONDED);
 317     }
 318
 319 #ifdef GMX_MPI
 320     if (TAKETIME)
 321     {
 322         t1=MPI_Wtime();
 323         fr->t_fnbf += t1-t0;
 324     }
 325 #endif
 326
 327     if (fepvals->sc_alpha!=0)
 328     {
 329         enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
 330     }
 331     else
 332     {
 333         enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
 334     }
 335
 336     if (fepvals->sc_alpha!=0)
 337
 338         /* even though coulomb part is linear, we already added it, beacuse we
 339            need to go through the vdw calculation anyway */
 340     {
 341         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 342     }
 343     else
 344     {
 345         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 346     }
 347
 348     Vsr = 0;
 349     if (bSepDVDL)
 350     {
 351         for(i=0; i<enerd->grpp.nener; i++)
 352         {
 353             Vsr +=
 354                 (fr->bBHAM ?
 355                  enerd->grpp.ener[egBHAMSR][i] :
 356                  enerd->grpp.ener[egLJSR][i])
 357                 + enerd->grpp.ener[egCOULSR][i] + enerd->grpp.ener[egGB][i];
 358         }
 359         dvdlsum = dvdl_nb[efptVDW] + dvdl_nb[efptCOUL];
 360         PRINT_SEPDVDL("VdW and Coulomb SR particle-p.",Vsr,dvdlsum);
 361     }
 362     debug_gmx();
 363
 364
 365     if (debug)
 366     {
 367         pr_rvecs(debug,0,"fshift after SR",fr->fshift,SHIFTS);
 368     }
 369
 370     /* Shift the coordinates. Must be done before bonded forces and PPPM,
 371      * but is also necessary for SHAKE and update, therefore it can NOT
 372      * go when no bonded forces have to be evaluated.
 373      */
 374
 375     /* Here sometimes we would not need to shift with NBFonly,
 376      * but we do so anyhow for consistency of the returned coordinates.
 377      */
 378     if (graph)
 379     {
 380         shift_self(graph,box,x);
 381         if (TRICLINIC(box))
 382         {
 383             inc_nrnb(nrnb,eNR_SHIFTX,2*graph->nnodes);
 384         }
 385         else
 386         {
 387             inc_nrnb(nrnb,eNR_SHIFTX,graph->nnodes);
 388         }
 389     }
 390     /* Check whether we need to do bondeds or correct for exclusions */
 391     if (fr->bMolPBC &&
 392         ((flags & GMX_FORCE_BONDED)
 393          || EEL_RF(fr->eeltype) || EEL_FULL(fr->eeltype)))
 394     {
 395         /* Since all atoms are in the rectangular or triclinic unit-cell,
 396          * only single box vector shifts (2 in x) are required.
 397          */
 398         set_pbc_dd(&pbc,fr->ePBC,cr->dd,TRUE,box);
 399     }
 400     debug_gmx();
 401
 402     if (flags & GMX_FORCE_BONDED)
 403     {
 404         wallcycle_sub_start(wcycle, ewcsBONDED);
 405         calc_bonds(fplog,cr->ms,
 406                    idef,x,hist,f,fr,&pbc,graph,enerd,nrnb,lambda,md,fcd,
 407                    DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL, atype, born,
 408                    flags,
 409                    fr->bSepDVDL && do_per_step(step,ir->nstlog),step);
 410
 411         /* Check if we have to determine energy differences
 412          * at foreign lambda's.
 413          */
 414         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) &&
 415             idef->ilsort != ilsortNO_FE)
 416         {
 417             if (idef->ilsort != ilsortFE_SORTED)
 418             {
 419                 gmx_incons("The bonded interactions are not sorted for free energy");
 420             }
 421             init_enerdata(mtop->groups.grps[egcENER].nr,fepvals->n_lambda,&ed_lam);
 422
 423             for(i=0; i<enerd->n_lambda; i++)
 424             {
 425                 reset_enerdata(&ir->opts,fr,TRUE,&ed_lam,FALSE);
 426                 for (j=0;j<efptNR;j++)
 427                 {
 428                     lam_i[j] = (i==0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 429                 }
 430                 calc_bonds_lambda(fplog,idef,x,fr,&pbc,graph,&ed_lam,nrnb,lam_i,md,
 431                                   fcd,DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL);
 432                 sum_epot(&ir->opts,&ed_lam);
 433                 enerd->enerpart_lambda[i] += ed_lam.term[F_EPOT];
 434             }
 435             destroy_enerdata(&ed_lam);
 436         }
 437         debug_gmx();
 438
 439         wallcycle_sub_stop(wcycle, ewcsBONDED);
 440     }
 441
 442     where();
 443
 444     *cycles_pme = 0;
 445     if (EEL_FULL(fr->eeltype))
 446     {
 447         bSB = (ir->nwall == 2);
 448         if (bSB)
 449         {
 450             copy_mat(box,boxs);
 451             svmul(ir->wall_ewald_zfac,boxs[ZZ],boxs[ZZ]);
 452             box_size[ZZ] *= ir->wall_ewald_zfac;
 453         }
 454
 455         clear_mat(fr->vir_el_recip);
 456
 457         if (fr->bEwald)
 458         {
 459             Vcorr = 0;
 460             dvdl  = 0;
 461
 462             /* With the Verlet scheme exclusion forces are calculated
 463              * in the non-bonded kernel.
 464              */
 465             /* The TPI molecule does not have exclusions with the rest
 466              * of the system and no intra-molecular PME grid contributions
 467              * will be calculated in gmx_pme_calc_energy.
 468              */
 469             if ((ir->cutoff_scheme == ecutsGROUP && fr->n_tpi == 0) ||
 470                 ir->ewald_geometry != eewg3D ||
 471                 ir->epsilon_surface != 0)
 472             {
 473                 int nthreads,t;
 474
 475                 wallcycle_sub_start(wcycle, ewcsEWALD_CORRECTION);
 476
 477                 if (fr->n_tpi > 0)
 478                 {
 479                     gmx_fatal(FARGS,"TPI with PME currently only works in a 3D geometry with tin-foil boundary conditions");
 480                 }
 481
 482                 nthreads = gmx_omp_nthreads_get(emntBonded);
 483 #pragma omp parallel for num_threads(nthreads) schedule(static)
 484                 for(t=0; t<nthreads; t++)
 485                 {
 486                     int s,e,i;
 487                     rvec *fnv;
 488                     tensor *vir;
 489                     real *Vcorrt,*dvdlt;
 490                     if (t == 0)
 491                     {
 492                         fnv    = fr->f_novirsum;
 493                         vir    = &fr->vir_el_recip;
 494                         Vcorrt = &Vcorr;
 495                         dvdlt  = &dvdl;
 496                     }
 497                     else
 498                     {
 499                         fnv    = fr->f_t[t].f;
 500                         vir    = &fr->f_t[t].vir;
 501                         Vcorrt = &fr->f_t[t].Vcorr;
 502                         dvdlt  = &fr->f_t[t].dvdl[efptCOUL];
 503                         for(i=0; i<fr->natoms_force; i++)
 504                         {
 505                             clear_rvec(fnv[i]);
 506                         }
 507                         clear_mat(*vir);
 508                     }
 509                     *dvdlt = 0;
 510                     *Vcorrt =
 511                         ewald_LRcorrection(fplog,
 512                                            fr->excl_load[t],fr->excl_load[t+1],
 513                                            cr,t,fr,
 514                                            md->chargeA,
 515                                            md->nChargePerturbed ? md->chargeB : NULL,
 516                                            ir->cutoff_scheme != ecutsVERLET,
 517                                            excl,x,bSB ? boxs : box,mu_tot,
 518                                            ir->ewald_geometry,
 519                                            ir->epsilon_surface,
 520                                            fnv,*vir,
 521                                            lambda[efptCOUL],dvdlt);
 522                 }
 523                 if (nthreads > 1)
 524                 {
 525                     reduce_thread_forces(fr->natoms_force,fr->f_novirsum,
 526                                          fr->vir_el_recip,
 527                                          &Vcorr,efptCOUL,&dvdl,
 528                                          nthreads,fr->f_t);
 529                 }
 530
 531                 wallcycle_sub_stop(wcycle, ewcsEWALD_CORRECTION);
 532             }
 533
 534             if (fr->n_tpi == 0)
 535             {
 536                 Vcorr += ewald_charge_correction(cr,fr,lambda[efptCOUL],box,
 537                                                  &dvdl,fr->vir_el_recip);
 538             }
 539
 540             PRINT_SEPDVDL("Ewald excl./charge/dip. corr.",Vcorr,dvdl);
 541             enerd->dvdl_lin[efptCOUL] += dvdl;
 542         }
 543
 544         status = 0;
 545         Vlr  = 0;
 546         dvdl = 0;
 547         switch (fr->eeltype)
 548         {
 549         case eelPME:
 550         case eelPMESWITCH:
 551         case eelPMEUSER:
 552         case eelPMEUSERSWITCH:
 553         case eelP3M_AD:
 554             if (cr->duty & DUTY_PME)
 555             {
 556                 assert(fr->n_tpi >= 0);
 557                 if (fr->n_tpi == 0 || (flags & GMX_FORCE_STATECHANGED))
 558                 {
 559                     pme_flags = GMX_PME_SPREAD_Q | GMX_PME_SOLVE;
 560                     if (flags & GMX_FORCE_FORCES)
 561                     {
 562                         pme_flags |= GMX_PME_CALC_F;
 563                     }
 564                     if (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY))
 565                     {
 566                         pme_flags |= GMX_PME_CALC_ENER_VIR;
 567                     }
 568                     if (fr->n_tpi > 0)
 569                     {
 570                         /* We don't calculate f, but we do want the potential */
 571                         pme_flags |= GMX_PME_CALC_POT;
 572                     }
 573                     wallcycle_start(wcycle,ewcPMEMESH);
 574                     status = gmx_pme_do(fr->pmedata,
 575                                         md->start,md->homenr - fr->n_tpi,
 576                                         x,fr->f_novirsum,
 577                                         md->chargeA,md->chargeB,
 578                                         bSB ? boxs : box,cr,
 579                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_x(cr->dd) : 0,
 580                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_y(cr->dd) : 0,
 581                                         nrnb,wcycle,
 582                                         fr->vir_el_recip,fr->ewaldcoeff,
 583                                         &Vlr,lambda[efptCOUL],&dvdl,
 584                                         pme_flags);
 585                     *cycles_pme = wallcycle_stop(wcycle,ewcPMEMESH);
 586
 587                     /* We should try to do as little computation after
 588                      * this as possible, because parallel PME synchronizes
 589                      * the nodes, so we want all load imbalance of the rest
 590                      * of the force calculation to be before the PME call.
 591                      * DD load balancing is done on the whole time of
 592                      * the force call (without PME).
 593                      */
 594                 }
 595                 if (fr->n_tpi > 0)
 596                 {
 597                     /* Determine the PME grid energy of the test molecule
 598                      * with the PME grid potential of the other charges.
 599                      */
 600                     gmx_pme_calc_energy(fr->pmedata,fr->n_tpi,
 601                                         x + md->homenr - fr->n_tpi,
 602                                         md->chargeA + md->homenr - fr->n_tpi,
 603                                         &Vlr);
 604                 }
 605                 PRINT_SEPDVDL("PME mesh",Vlr,dvdl);
 606             }
 607             break;
 608         case eelEWALD:
 609             Vlr = do_ewald(fplog,FALSE,ir,x,fr->f_novirsum,
 610                            md->chargeA,md->chargeB,
 611                            box_size,cr,md->homenr,
 612                            fr->vir_el_recip,fr->ewaldcoeff,
 613                            lambda[efptCOUL],&dvdl,fr->ewald_table);
 614             PRINT_SEPDVDL("Ewald long-range",Vlr,dvdl);
 615             break;
 616         default:
 617             gmx_fatal(FARGS,"No such electrostatics method implemented %s",
 618                       eel_names[fr->eeltype]);
 619         }
 620         if (status != 0)
 621         {
 622             gmx_fatal(FARGS,"Error %d in long range electrostatics routine %s",
 623                       status,EELTYPE(fr->eeltype));
 624                 }
 625         /* Note that with separate PME nodes we get the real energies later */
 626         enerd->dvdl_lin[efptCOUL] += dvdl;
 627         enerd->term[F_COUL_RECIP] = Vlr + Vcorr;
 628         if (debug)
 629         {
 630             fprintf(debug,"Vlr = %g, Vcorr = %g, Vlr_corr = %g\n",
 631                     Vlr,Vcorr,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         }
 635     }
 636     else
 637     {
 638         if (EEL_RF(fr->eeltype))
 639         {
 640             /* With the Verlet scheme exclusion forces are calculated
 641              * in the non-bonded kernel.
 642              */
 643             if (ir->cutoff_scheme != ecutsVERLET && fr->eeltype != eelRF_NEC)
 644             {
 645                 dvdl = 0;
 646                 enerd->term[F_RF_EXCL] =
 647                     RF_excl_correction(fplog,fr,graph,md,excl,x,f,
 648                                        fr->fshift,&pbc,lambda[efptCOUL],&dvdl);
 649             }
 650
 651             enerd->dvdl_lin[efptCOUL] += dvdl;
 652             PRINT_SEPDVDL("RF exclusion correction",
 653                           enerd->term[F_RF_EXCL],dvdl);
 654         }
 655     }
 656     where();
 657     debug_gmx();
 658
 659     if (debug)
 660     {
 661         print_nrnb(debug,nrnb);
 662     }
 663     debug_gmx();
 664
 665 #ifdef GMX_MPI
 666     if (TAKETIME)
 667     {
 668         t2=MPI_Wtime();
 669         MPI_Barrier(cr->mpi_comm_mygroup);
 670         t3=MPI_Wtime();
 671         fr->t_wait += t3-t2;
 672         if (fr->timesteps == 11)
 673         {
 674             fprintf(stderr,"* PP load balancing info: node %d, step %s, rel wait time=%3.0f%% , load string value: %7.2f\n",
 675                     cr->nodeid, gmx_step_str(fr->timesteps,buf),
 676                     100*fr->t_wait/(fr->t_wait+fr->t_fnbf),
 677                     (fr->t_fnbf+fr->t_wait)/fr->t_fnbf);
 678         }
 679         fr->timesteps++;
 680     }
 681 #endif
 682
 683     if (debug)
 684     {
 685         pr_rvecs(debug,0,"fshift after bondeds",fr->fshift,SHIFTS);
 686     }
 687
 688 }
 689
 690 void init_enerdata(int ngener,int n_lambda,gmx_enerdata_t *enerd)
 691 {
 692     int i,n2;
 693
 694     for(i=0; i<F_NRE; i++)
 695     {
 696         enerd->term[i] = 0;
 697     }
 698
 699
 700     for(i=0; i<efptNR; i++) {
 701         enerd->dvdl_lin[i]  = 0;
 702         enerd->dvdl_nonlin[i]  = 0;
 703     }
 704
 705     n2=ngener*ngener;
 706     if (debug)
 707     {
 708         fprintf(debug,"Creating %d sized group matrix for energies\n",n2);
 709     }
 710     enerd->grpp.nener = n2;
 711     for(i=0; (i<egNR); i++)
 712     {
 713         snew(enerd->grpp.ener[i],n2);
 714     }
 715
 716     if (n_lambda)
 717     {
 718         enerd->n_lambda = 1 + n_lambda;
 719         snew(enerd->enerpart_lambda,enerd->n_lambda);
 720     }
 721     else
 722     {
 723         enerd->n_lambda = 0;
 724     }
 725 }
 726
 727 void destroy_enerdata(gmx_enerdata_t *enerd)
 728 {
 729     int i;
 730
 731     for(i=0; (i<egNR); i++)
 732     {
 733         sfree(enerd->grpp.ener[i]);
 734     }
 735
 736     if (enerd->n_lambda)
 737     {
 738         sfree(enerd->enerpart_lambda);
 739     }
 740 }
 741
 742 static real sum_v(int n,real v[])
 743 {
 744   real t;
 745   int  i;
 746
 747   t = 0.0;
 748   for(i=0; (i<n); i++)
 749     t = t + v[i];
 750
 751   return t;
 752 }
 753
 754 void sum_epot(t_grpopts *opts,gmx_enerdata_t *enerd)
 755 {
 756   gmx_grppairener_t *grpp;
 757   real *epot;
 758   int i;
 759
 760   grpp = &enerd->grpp;
 761   epot = enerd->term;
 762
 763   /* Accumulate energies */
 764   epot[F_COUL_SR]  = sum_v(grpp->nener,grpp->ener[egCOULSR]);
 765   epot[F_LJ]       = sum_v(grpp->nener,grpp->ener[egLJSR]);
 766   epot[F_LJ14]     = sum_v(grpp->nener,grpp->ener[egLJ14]);
 767   epot[F_COUL14]   = sum_v(grpp->nener,grpp->ener[egCOUL14]);
 768   epot[F_COUL_LR]  = sum_v(grpp->nener,grpp->ener[egCOULLR]);
 769   epot[F_LJ_LR]    = sum_v(grpp->nener,grpp->ener[egLJLR]);
 770   /* We have already added 1-2,1-3, and 1-4 terms to F_GBPOL */
 771   epot[F_GBPOL]   += sum_v(grpp->nener,grpp->ener[egGB]);
 772
 773 /* lattice part of LR doesnt belong to any group
 774  * and has been added earlier
 775  */
 776   epot[F_BHAM]     = sum_v(grpp->nener,grpp->ener[egBHAMSR]);
 777   epot[F_BHAM_LR]  = sum_v(grpp->nener,grpp->ener[egBHAMLR]);
 778
 779   epot[F_EPOT] = 0;
 780   for(i=0; (i<F_EPOT); i++)
 781   {
 782       if (i != F_DISRESVIOL && i != F_ORIRESDEV)
 783       {
 784           epot[F_EPOT] += epot[i];
 785       }
 786   }
 787 }
 788
 789 void sum_dhdl(gmx_enerdata_t *enerd, real *lambda, t_lambda *fepvals)
 790 {
 791     int i,j,index;
 792     double dlam;
 793
 794     enerd->dvdl_lin[efptVDW] += enerd->term[F_DVDL_VDW];  /* include dispersion correction */
 795     enerd->term[F_DVDL] = 0.0;
 796     for (i=0;i<efptNR;i++)
 797     {
 798         if (fepvals->separate_dvdl[i])
 799         {
 800             /* could this be done more readably/compactly? */
 801             switch (i) {
 802             case (efptCOUL):
 803                 index = F_DVDL_COUL;
 804                 break;
 805             case (efptVDW):
 806                 index = F_DVDL_VDW;
 807                 break;
 808             case (efptBONDED):
 809                 index = F_DVDL_BONDED;
 810                 break;
 811             case (efptRESTRAINT):
 812                 index = F_DVDL_RESTRAINT;
 813                 break;
 814             case (efptMASS):
 815                 index = F_DKDL;
 816                 break;
 817             default:
 818                 index = F_DVDL;
 819                 break;
 820             }
 821             enerd->term[index] = enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 822             if (debug)
 823             {
 824                 fprintf(debug,"dvdl-%s[%2d]: %f: non-linear %f + linear %f\n",
 825                         efpt_names[i],i,enerd->term[index],enerd->dvdl_nonlin[i],enerd->dvdl_lin[i]);
 826             }
 827         }
 828         else
 829         {
 830             enerd->term[F_DVDL] += enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 831             if (debug)
 832             {
 833                 fprintf(debug,"dvd-%sl[%2d]: %f: non-linear %f + linear %f\n",
 834                         efpt_names[0],i,enerd->term[F_DVDL],enerd->dvdl_nonlin[i],enerd->dvdl_lin[i]);
 835             }
 836         }
 837     }
 838
 839     /* Notes on the foreign lambda free energy difference evaluation:
 840      * Adding the potential and ekin terms that depend linearly on lambda
 841      * as delta lam * dvdl to the energy differences is exact.
 842      * For the constraints this is not exact, but we have no other option
 843      * without literally changing the lengths and reevaluating the energies at each step.
 844      * (try to remedy this post 4.6 - MRS)
 845      * For the non-bonded LR term we assume that the soft-core (if present)
 846      * no longer affects the energy beyond the short-range cut-off,
 847      * which is a very good approximation (except for exotic settings).
 848      * (investigate how to overcome this post 4.6 - MRS)
 849      */
 850
 851     for(i=0; i<fepvals->n_lambda; i++)
 852     {                                         /* note we are iterating over fepvals here!
 853                                                  For the current lam, dlam = 0 automatically,
 854                                                  so we don't need to add anything to the
 855                                                  enerd->enerpart_lambda[0] */
 856
 857         /* we don't need to worry about dvdl contributions to the current lambda, because
 858            it's automatically zero */
 859
 860         /* first kinetic energy term */
 861         dlam = (fepvals->all_lambda[efptMASS][i] - lambda[efptMASS]);
 862
 863         enerd->enerpart_lambda[i+1] += enerd->term[F_DKDL]*dlam;
 864
 865         for (j=0;j<efptNR;j++)
 866         {
 867             if (j==efptMASS) {continue;} /* no other mass term to worry about */
 868
 869             dlam = (fepvals->all_lambda[j][i]-lambda[j]);
 870             enerd->enerpart_lambda[i+1] += dlam*enerd->dvdl_lin[j];
 871             if (debug)
 872             {
 873                 fprintf(debug,"enerdiff lam %g: (%15s), non-linear %f linear %f*%f\n",
 874                         fepvals->all_lambda[j][i],efpt_names[j],
 875                         (enerd->enerpart_lambda[i+1] - enerd->enerpart_lambda[0]),
 876                         dlam,enerd->dvdl_lin[j]);
 877             }
 878         }
 879     }
 880 }
 881
 882 void reset_enerdata(t_grpopts *opts,
 883                     t_forcerec *fr,gmx_bool bNS,
 884                     gmx_enerdata_t *enerd,
 885                     gmx_bool bMaster)
 886 {
 887     gmx_bool bKeepLR;
 888     int  i,j;
 889
 890     /* First reset all energy components, except for the long range terms
 891      * on the master at non neighbor search steps, since the long range
 892      * terms have already been summed at the last neighbor search step.
 893      */
 894     bKeepLR = (fr->bTwinRange && !bNS);
 895     for(i=0; (i<egNR); i++) {
 896         if (!(bKeepLR && bMaster && (i == egCOULLR || i == egLJLR))) {
 897             for(j=0; (j<enerd->grpp.nener); j++)
 898                 enerd->grpp.ener[i][j] = 0.0;
 899         }
 900     }
 901     for (i=0;i<efptNR;i++)
 902     {
 903         enerd->dvdl_lin[i]    = 0.0;
 904         enerd->dvdl_nonlin[i] = 0.0;
 905     }
 906
 907     /* Normal potential energy components */
 908     for(i=0; (i<=F_EPOT); i++) {
 909         enerd->term[i] = 0.0;
 910     }
 911     /* Initialize the dVdlambda term with the long range contribution */
 912     /* Initialize the dvdl term with the long range contribution */
 913     enerd->term[F_DVDL]            = 0.0;
 914     enerd->term[F_DVDL_COUL]       = 0.0;
 915     enerd->term[F_DVDL_VDW]        = 0.0;
 916     enerd->term[F_DVDL_BONDED]     = 0.0;
 917     enerd->term[F_DVDL_RESTRAINT]  = 0.0;
 918     enerd->term[F_DKDL]            = 0.0;
 919     if (enerd->n_lambda > 0)
 920     {
 921         for(i=0; i<enerd->n_lambda; i++)
 922         {
 923             enerd->enerpart_lambda[i] = 0.0;
 924         }
 925     }
 926 }