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       bDoLongRange,
  83         gmx_bool       bDoForces,
  84         rvec       *f)
  85 {
  86   char   *ptr;
  87   int    nsearch;
  88
  89
  90   if (!fr->ns.nblist_initialized)
  91   {
  92       init_neighbor_list(fp, fr, md->homenr);
  93   }
  94
  95   if (fr->bTwinRange)
  96     fr->nlr=0;
  97
  98     nsearch = search_neighbours(fp,fr,x,box,top,groups,cr,nrnb,md,
  99                                 lambda,dvdlambda,grppener,
 100                                 bFillGrid,bDoLongRange,
 101                                 bDoForces,f);
 102   if (debug)
 103     fprintf(debug,"nsearch = %d\n",nsearch);
 104
 105   /* Check whether we have to do dynamic load balancing */
 106   /*if ((nsb->nstDlb > 0) && (mod(step,nsb->nstDlb) == 0))
 107     count_nb(cr,nsb,&(top->blocks[ebCGS]),nns,fr->nlr,
 108     &(top->idef),opts->ngener);
 109   */
 110   if (fr->ns.dump_nl > 0)
 111     dump_nblist(fp,cr,fr,fr->ns.dump_nl);
 112 }
 113
 114 static void reduce_thread_forces(int n,rvec *f,
 115                                  tensor vir,
 116                                  real *Vcorr,
 117                                  int efpt_ind,real *dvdl,
 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 += f_t[t].Vcorr;
 134         *dvdl  += f_t[t].dvdl[efpt_ind];
 135         m_add(vir,f_t[t].vir,vir);
 136     }
 137 }
 138
 139 void do_force_lowlevel(FILE       *fplog,   gmx_large_int_t step,
 140                        t_forcerec *fr,      t_inputrec *ir,
 141                        t_idef     *idef,    t_commrec  *cr,
 142                        t_nrnb     *nrnb,    gmx_wallcycle_t wcycle,
 143                        t_mdatoms  *md,
 144                        t_grpopts  *opts,
 145                        rvec       x[],      history_t  *hist,
 146                        rvec       f[],
 147                        gmx_enerdata_t *enerd,
 148                        t_fcdata   *fcd,
 149                        gmx_mtop_t     *mtop,
 150                        gmx_localtop_t *top,
 151                        gmx_genborn_t *born,
 152                        t_atomtypes *atype,
 153                        gmx_bool       bBornRadii,
 154                        matrix     box,
 155                        t_lambda   *fepvals,
 156                        real       *lambda,
 157                        t_graph    *graph,
 158                        t_blocka   *excl,
 159                        rvec       mu_tot[],
 160                        int        flags,
 161                        float      *cycles_pme)
 162 {
 163     int     i,j,status;
 164     int     donb_flags;
 165     gmx_bool    bDoEpot,bSepDVDL,bSB;
 166     int     pme_flags;
 167     matrix  boxs;
 168     rvec    box_size;
 169     real    Vsr,Vlr,Vcorr=0;
 170     t_pbc   pbc;
 171     real    dvdgb;
 172     char    buf[22];
 173     gmx_enerdata_t ed_lam;
 174     double  clam_i,vlam_i;
 175     real    dvdl_dum[efptNR], dvdl, dvdl_nb[efptNR], lam_i[efptNR];
 176     real    dvdlsum;
 177
 178 #ifdef GMX_MPI
 179     double  t0=0.0,t1,t2,t3; /* time measurement for coarse load balancing */
 180 #endif
 181
 182 #define PRINT_SEPDVDL(s,v,dvdlambda) if (bSepDVDL) fprintf(fplog,sepdvdlformat,s,v,dvdlambda);
 183
 184
 185     set_pbc(&pbc,fr->ePBC,box);
 186
 187     /* reset free energy components */
 188     for (i=0;i<efptNR;i++)
 189     {
 190         dvdl_nb[i]  = 0;
 191         dvdl_dum[i] = 0;
 192     }
 193
 194     /* Reset box */
 195     for(i=0; (i<DIM); i++)
 196     {
 197         box_size[i]=box[i][i];
 198     }
 199
 200     bSepDVDL=(fr->bSepDVDL && do_per_step(step,ir->nstlog));
 201     debug_gmx();
 202
 203     /* do QMMM first if requested */
 204     if(fr->bQMMM)
 205     {
 206         enerd->term[F_EQM] = calculate_QMMM(cr,x,f,fr,md);
 207     }
 208
 209     if (bSepDVDL)
 210     {
 211         fprintf(fplog,"Step %s: non-bonded V and dVdl for node %d:\n",
 212                 gmx_step_str(step,buf),cr->nodeid);
 213     }
 214
 215     /* Call the short range functions all in one go. */
 216
 217 #ifdef GMX_MPI
 218     /*#define TAKETIME ((cr->npmenodes) && (fr->timesteps < 12))*/
 219 #define TAKETIME FALSE
 220     if (TAKETIME)
 221     {
 222         MPI_Barrier(cr->mpi_comm_mygroup);
 223         t0=MPI_Wtime();
 224     }
 225 #endif
 226
 227     if (ir->nwall)
 228     {
 229         /* foreign lambda component for walls */
 230         dvdl = do_walls(ir,fr,box,md,x,f,lambda[efptVDW],
 231                         enerd->grpp.ener[egLJSR],nrnb);
 232         PRINT_SEPDVDL("Walls",0.0,dvdl);
 233         enerd->dvdl_lin[efptVDW] += dvdl;
 234     }
 235
 236         /* If doing GB, reset dvda and calculate the Born radii */
 237         if (ir->implicit_solvent)
 238         {
 239         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 240
 241                 for(i=0;i<born->nr;i++)
 242                 {
 243                         fr->dvda[i]=0;
 244                 }
 245
 246                 if(bBornRadii)
 247                 {
 248                         calc_gb_rad(cr,fr,ir,top,atype,x,&(fr->gblist),born,md,nrnb);
 249                 }
 250
 251         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 252         }
 253
 254     where();
 255     if (flags & GMX_FORCE_NONBONDED)
 256     {
 257         donb_flags = 0;
 258         if (flags & GMX_FORCE_FORCES)
 259         {
 260             donb_flags |= GMX_DONB_FORCES;
 261         }
 262
 263         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 264         do_nonbonded(cr,fr,x,f,md,excl,
 265                     fr->bBHAM ?
 266                     enerd->grpp.ener[egBHAMSR] :
 267                     enerd->grpp.ener[egLJSR],
 268                     enerd->grpp.ener[egCOULSR],
 269                     enerd->grpp.ener[egGB],box_size,nrnb,
 270                     lambda,dvdl_nb,-1,-1,donb_flags);
 271         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 272     }
 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         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 280         init_enerdata(mtop->groups.grps[egcENER].nr,fepvals->n_lambda,&ed_lam);
 281
 282         for(i=0; i<enerd->n_lambda; i++)
 283         {
 284             for (j=0;j<efptNR;j++)
 285             {
 286                 lam_i[j] = (i==0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 287             }
 288             reset_enerdata(&ir->opts,fr,TRUE,&ed_lam,FALSE);
 289             do_nonbonded(cr,fr,x,f,md,excl,
 290                          fr->bBHAM ?
 291                          ed_lam.grpp.ener[egBHAMSR] :
 292                          ed_lam.grpp.ener[egLJSR],
 293                          ed_lam.grpp.ener[egCOULSR],
 294                          enerd->grpp.ener[egGB], box_size,nrnb,
 295                          lam_i,dvdl_dum,-1,-1,
 296                          GMX_DONB_FOREIGNLAMBDA);
 297             sum_epot(&ir->opts,&ed_lam);
 298             enerd->enerpart_lambda[i] += ed_lam.term[F_EPOT];
 299         }
 300         destroy_enerdata(&ed_lam);
 301         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 302     }
 303     where();
 304
 305         /* If we are doing GB, calculate bonded forces and apply corrections
 306          * to the solvation forces */
 307     /* MRS: Eventually, many need to include free energy contribution here! */
 308         if (ir->implicit_solvent)
 309     {
 310                 calc_gb_forces(cr,md,born,top,atype,x,f,fr,idef,
 311                        ir->gb_algorithm,ir->sa_algorithm,nrnb,bBornRadii,&pbc,graph,enerd);
 312         wallcycle_sub_stop(wcycle, ewcsBONDED);
 313     }
 314
 315 #ifdef GMX_MPI
 316     if (TAKETIME)
 317     {
 318         t1=MPI_Wtime();
 319         fr->t_fnbf += t1-t0;
 320     }
 321 #endif
 322
 323     if (fepvals->sc_alpha!=0)
 324     {
 325         enerd->dvdl_nonlin[efptVDW] += dvdl_nb[efptVDW];
 326     }
 327     else
 328     {
 329         enerd->dvdl_lin[efptVDW] += dvdl_nb[efptVDW];
 330     }
 331
 332     if (fepvals->sc_alpha!=0)
 333
 334         /* even though coulomb part is linear, we already added it, beacuse we
 335            need to go through the vdw calculation anyway */
 336     {
 337         enerd->dvdl_nonlin[efptCOUL] += dvdl_nb[efptCOUL];
 338     }
 339     else
 340     {
 341         enerd->dvdl_lin[efptCOUL] += dvdl_nb[efptCOUL];
 342     }
 343
 344     Vsr = 0;
 345     if (bSepDVDL)
 346     {
 347         for(i=0; i<enerd->grpp.nener; i++)
 348         {
 349             Vsr +=
 350                 (fr->bBHAM ?
 351                  enerd->grpp.ener[egBHAMSR][i] :
 352                  enerd->grpp.ener[egLJSR][i])
 353                 + enerd->grpp.ener[egCOULSR][i] + enerd->grpp.ener[egGB][i];
 354         }
 355         dvdlsum = dvdl_nb[efptVDW] + dvdl_nb[efptCOUL];
 356         PRINT_SEPDVDL("VdW and Coulomb SR particle-p.",Vsr,dvdlsum);
 357     }
 358     debug_gmx();
 359
 360
 361     if (debug)
 362     {
 363         pr_rvecs(debug,0,"fshift after SR",fr->fshift,SHIFTS);
 364     }
 365
 366     /* Shift the coordinates. Must be done before bonded forces and PPPM,
 367      * but is also necessary for SHAKE and update, therefore it can NOT
 368      * go when no bonded forces have to be evaluated.
 369      */
 370
 371     /* Here sometimes we would not need to shift with NBFonly,
 372      * but we do so anyhow for consistency of the returned coordinates.
 373      */
 374     if (graph)
 375     {
 376         shift_self(graph,box,x);
 377         if (TRICLINIC(box))
 378         {
 379             inc_nrnb(nrnb,eNR_SHIFTX,2*graph->nnodes);
 380         }
 381         else
 382         {
 383             inc_nrnb(nrnb,eNR_SHIFTX,graph->nnodes);
 384         }
 385     }
 386     /* Check whether we need to do bondeds or correct for exclusions */
 387     if (fr->bMolPBC &&
 388         ((flags & GMX_FORCE_BONDED)
 389          || EEL_RF(fr->eeltype) || EEL_FULL(fr->eeltype)))
 390     {
 391         /* Since all atoms are in the rectangular or triclinic unit-cell,
 392          * only single box vector shifts (2 in x) are required.
 393          */
 394         set_pbc_dd(&pbc,fr->ePBC,cr->dd,TRUE,box);
 395     }
 396     debug_gmx();
 397
 398     if (flags & GMX_FORCE_BONDED)
 399     {
 400         wallcycle_sub_start(wcycle, ewcsBONDED);
 401         calc_bonds(fplog,cr->ms,
 402                    idef,x,hist,f,fr,&pbc,graph,enerd,nrnb,lambda,md,fcd,
 403                    DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL, atype, born,
 404                    flags,
 405                    fr->bSepDVDL && do_per_step(step,ir->nstlog),step);
 406
 407         /* Check if we have to determine energy differences
 408          * at foreign lambda's.
 409          */
 410         if (fepvals->n_lambda > 0 && (flags & GMX_FORCE_DHDL) &&
 411             idef->ilsort != ilsortNO_FE)
 412         {
 413             if (idef->ilsort != ilsortFE_SORTED)
 414             {
 415                 gmx_incons("The bonded interactions are not sorted for free energy");
 416             }
 417             init_enerdata(mtop->groups.grps[egcENER].nr,fepvals->n_lambda,&ed_lam);
 418
 419             for(i=0; i<enerd->n_lambda; i++)
 420             {
 421                 reset_enerdata(&ir->opts,fr,TRUE,&ed_lam,FALSE);
 422                 for (j=0;j<efptNR;j++)
 423                 {
 424                     lam_i[j] = (i==0 ? lambda[j] : fepvals->all_lambda[j][i-1]);
 425                 }
 426                 calc_bonds_lambda(fplog,idef,x,fr,&pbc,graph,&ed_lam,nrnb,lam_i,md,
 427                                   fcd,DOMAINDECOMP(cr) ? cr->dd->gatindex : NULL);
 428                 sum_epot(&ir->opts,&ed_lam);
 429                 enerd->enerpart_lambda[i] += ed_lam.term[F_EPOT];
 430             }
 431             destroy_enerdata(&ed_lam);
 432         }
 433         debug_gmx();
 434
 435         wallcycle_sub_stop(wcycle, ewcsBONDED);
 436     }
 437
 438     where();
 439
 440     *cycles_pme = 0;
 441     if (EEL_FULL(fr->eeltype))
 442     {
 443         bSB = (ir->nwall == 2);
 444         if (bSB)
 445         {
 446             copy_mat(box,boxs);
 447             svmul(ir->wall_ewald_zfac,boxs[ZZ],boxs[ZZ]);
 448             box_size[ZZ] *= ir->wall_ewald_zfac;
 449         }
 450
 451         clear_mat(fr->vir_el_recip);
 452
 453         if (fr->bEwald)
 454         {
 455             Vcorr = 0;
 456             dvdl  = 0;
 457
 458             /* With the Verlet scheme exclusion forces are calculated
 459              * in the non-bonded kernel.
 460              */
 461             /* The TPI molecule does not have exclusions with the rest
 462              * of the system and no intra-molecular PME grid contributions
 463              * will be calculated in gmx_pme_calc_energy.
 464              */
 465             if ((ir->cutoff_scheme == ecutsGROUP && fr->n_tpi == 0) ||
 466                 ir->ewald_geometry != eewg3D ||
 467                 ir->epsilon_surface != 0)
 468             {
 469                 int nthreads,t;
 470
 471                 wallcycle_sub_start(wcycle, ewcsEWALD_CORRECTION);
 472
 473                 if (fr->n_tpi > 0)
 474                 {
 475                     gmx_fatal(FARGS,"TPI with PME currently only works in a 3D geometry with tin-foil boundary conditions");
 476                 }
 477
 478                 nthreads = gmx_omp_nthreads_get(emntBonded);
 479 #pragma omp parallel for num_threads(nthreads) schedule(static)
 480                 for(t=0; t<nthreads; t++)
 481                 {
 482                     int s,e,i;
 483                     rvec *fnv;
 484                     tensor *vir;
 485                     real *Vcorrt,*dvdlt;
 486                     if (t == 0)
 487                     {
 488                         fnv    = fr->f_novirsum;
 489                         vir    = &fr->vir_el_recip;
 490                         Vcorrt = &Vcorr;
 491                         dvdlt  = &dvdl;
 492                     }
 493                     else
 494                     {
 495                         fnv    = fr->f_t[t].f;
 496                         vir    = &fr->f_t[t].vir;
 497                         Vcorrt = &fr->f_t[t].Vcorr;
 498                         dvdlt  = &fr->f_t[t].dvdl[efptCOUL];
 499                         for(i=0; i<fr->natoms_force; i++)
 500                         {
 501                             clear_rvec(fnv[i]);
 502                         }
 503                         clear_mat(*vir);
 504                     }
 505                     *dvdlt = 0;
 506                     *Vcorrt =
 507                         ewald_LRcorrection(fplog,
 508                                            fr->excl_load[t],fr->excl_load[t+1],
 509                                            cr,t,fr,
 510                                            md->chargeA,
 511                                            md->nChargePerturbed ? md->chargeB : NULL,
 512                                            ir->cutoff_scheme != ecutsVERLET,
 513                                            excl,x,bSB ? boxs : box,mu_tot,
 514                                            ir->ewald_geometry,
 515                                            ir->epsilon_surface,
 516                                            fnv,*vir,
 517                                            lambda[efptCOUL],dvdlt);
 518                 }
 519                 if (nthreads > 1)
 520                 {
 521                     reduce_thread_forces(fr->natoms_force,fr->f_novirsum,
 522                                          fr->vir_el_recip,
 523                                          &Vcorr,efptCOUL,&dvdl,
 524                                          nthreads,fr->f_t);
 525                 }
 526
 527                 wallcycle_sub_stop(wcycle, ewcsEWALD_CORRECTION);
 528             }
 529
 530             if (fr->n_tpi == 0)
 531             {
 532                 Vcorr += ewald_charge_correction(cr,fr,lambda[efptCOUL],box,
 533                                                  &dvdl,fr->vir_el_recip);
 534             }
 535
 536             PRINT_SEPDVDL("Ewald excl./charge/dip. corr.",Vcorr,dvdl);
 537             enerd->dvdl_lin[efptCOUL] += dvdl;
 538         }
 539
 540         status = 0;
 541         Vlr  = 0;
 542         dvdl = 0;
 543         switch (fr->eeltype)
 544         {
 545         case eelPME:
 546         case eelPMESWITCH:
 547         case eelPMEUSER:
 548         case eelPMEUSERSWITCH:
 549         case eelP3M_AD:
 550             if (cr->duty & DUTY_PME)
 551             {
 552                 assert(fr->n_tpi >= 0);
 553                 if (fr->n_tpi == 0 || (flags & GMX_FORCE_STATECHANGED))
 554                 {
 555                     pme_flags = GMX_PME_SPREAD_Q | GMX_PME_SOLVE;
 556                     if (flags & GMX_FORCE_FORCES)
 557                     {
 558                         pme_flags |= GMX_PME_CALC_F;
 559                     }
 560                     if (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY))
 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                                         md->start,md->homenr - fr->n_tpi,
 572                                         x,fr->f_novirsum,
 573                                         md->chargeA,md->chargeB,
 574                                         bSB ? boxs : box,cr,
 575                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_x(cr->dd) : 0,
 576                                         DOMAINDECOMP(cr) ? dd_pme_maxshift_y(cr->dd) : 0,
 577                                         nrnb,wcycle,
 578                                         fr->vir_el_recip,fr->ewaldcoeff,
 579                                         &Vlr,lambda[efptCOUL],&dvdl,
 580                                         pme_flags);
 581                     *cycles_pme = wallcycle_stop(wcycle,ewcPMEMESH);
 582
 583                     /* We should try to do as little computation after
 584                      * this as possible, because parallel PME synchronizes
 585                      * the nodes, so we want all load imbalance of the rest
 586                      * of the force calculation to be before the PME call.
 587                      * DD load balancing is done on the whole time of
 588                      * the force call (without PME).
 589                      */
 590                 }
 591                 if (fr->n_tpi > 0)
 592                 {
 593                     /* Determine the PME grid energy of the test molecule
 594                      * with the PME grid potential of the other charges.
 595                      */
 596                     gmx_pme_calc_energy(fr->pmedata,fr->n_tpi,
 597                                         x + md->homenr - fr->n_tpi,
 598                                         md->chargeA + md->homenr - fr->n_tpi,
 599                                         &Vlr);
 600                 }
 601                 PRINT_SEPDVDL("PME mesh",Vlr,dvdl);
 602             }
 603             break;
 604         case eelEWALD:
 605             Vlr = do_ewald(fplog,FALSE,ir,x,fr->f_novirsum,
 606                            md->chargeA,md->chargeB,
 607                            box_size,cr,md->homenr,
 608                            fr->vir_el_recip,fr->ewaldcoeff,
 609                            lambda[efptCOUL],&dvdl,fr->ewald_table);
 610             PRINT_SEPDVDL("Ewald long-range",Vlr,dvdl);
 611             break;
 612         default:
 613             gmx_fatal(FARGS,"No such electrostatics method implemented %s",
 614                       eel_names[fr->eeltype]);
 615         }
 616         if (status != 0)
 617         {
 618             gmx_fatal(FARGS,"Error %d in long range electrostatics routine %s",
 619                       status,EELTYPE(fr->eeltype));
 620                 }
 621         /* Note that with separate PME nodes we get the real energies later */
 622         enerd->dvdl_lin[efptCOUL] += dvdl;
 623         enerd->term[F_COUL_RECIP] = Vlr + Vcorr;
 624         if (debug)
 625         {
 626             fprintf(debug,"Vlr = %g, Vcorr = %g, Vlr_corr = %g\n",
 627                     Vlr,Vcorr,enerd->term[F_COUL_RECIP]);
 628             pr_rvecs(debug,0,"vir_el_recip after corr",fr->vir_el_recip,DIM);
 629             pr_rvecs(debug,0,"fshift after LR Corrections",fr->fshift,SHIFTS);
 630         }
 631     }
 632     else
 633     {
 634         if (EEL_RF(fr->eeltype))
 635         {
 636             /* With the Verlet scheme exclusion forces are calculated
 637              * in the non-bonded kernel.
 638              */
 639             if (ir->cutoff_scheme != ecutsVERLET && fr->eeltype != eelRF_NEC)
 640             {
 641                 dvdl = 0;
 642                 enerd->term[F_RF_EXCL] =
 643                     RF_excl_correction(fplog,fr,graph,md,excl,x,f,
 644                                        fr->fshift,&pbc,lambda[efptCOUL],&dvdl);
 645             }
 646
 647             enerd->dvdl_lin[efptCOUL] += dvdl;
 648             PRINT_SEPDVDL("RF exclusion correction",
 649                           enerd->term[F_RF_EXCL],dvdl);
 650         }
 651     }
 652     where();
 653     debug_gmx();
 654
 655     if (debug)
 656     {
 657         print_nrnb(debug,nrnb);
 658     }
 659     debug_gmx();
 660
 661 #ifdef GMX_MPI
 662     if (TAKETIME)
 663     {
 664         t2=MPI_Wtime();
 665         MPI_Barrier(cr->mpi_comm_mygroup);
 666         t3=MPI_Wtime();
 667         fr->t_wait += t3-t2;
 668         if (fr->timesteps == 11)
 669         {
 670             fprintf(stderr,"* PP load balancing info: node %d, step %s, rel wait time=%3.0f%% , load string value: %7.2f\n",
 671                     cr->nodeid, gmx_step_str(fr->timesteps,buf),
 672                     100*fr->t_wait/(fr->t_wait+fr->t_fnbf),
 673                     (fr->t_fnbf+fr->t_wait)/fr->t_fnbf);
 674         }
 675         fr->timesteps++;
 676     }
 677 #endif
 678
 679     if (debug)
 680     {
 681         pr_rvecs(debug,0,"fshift after bondeds",fr->fshift,SHIFTS);
 682     }
 683
 684 }
 685
 686 void init_enerdata(int ngener,int n_lambda,gmx_enerdata_t *enerd)
 687 {
 688     int i,n2;
 689
 690     for(i=0; i<F_NRE; i++)
 691     {
 692         enerd->term[i] = 0;
 693     }
 694
 695
 696     for(i=0; i<efptNR; i++) {
 697         enerd->dvdl_lin[i]  = 0;
 698         enerd->dvdl_nonlin[i]  = 0;
 699     }
 700
 701     n2=ngener*ngener;
 702     if (debug)
 703     {
 704         fprintf(debug,"Creating %d sized group matrix for energies\n",n2);
 705     }
 706     enerd->grpp.nener = n2;
 707     for(i=0; (i<egNR); i++)
 708     {
 709         snew(enerd->grpp.ener[i],n2);
 710     }
 711
 712     if (n_lambda)
 713     {
 714         enerd->n_lambda = 1 + n_lambda;
 715         snew(enerd->enerpart_lambda,enerd->n_lambda);
 716     }
 717     else
 718     {
 719         enerd->n_lambda = 0;
 720     }
 721 }
 722
 723 void destroy_enerdata(gmx_enerdata_t *enerd)
 724 {
 725     int i;
 726
 727     for(i=0; (i<egNR); i++)
 728     {
 729         sfree(enerd->grpp.ener[i]);
 730     }
 731
 732     if (enerd->n_lambda)
 733     {
 734         sfree(enerd->enerpart_lambda);
 735     }
 736 }
 737
 738 static real sum_v(int n,real v[])
 739 {
 740   real t;
 741   int  i;
 742
 743   t = 0.0;
 744   for(i=0; (i<n); i++)
 745     t = t + v[i];
 746
 747   return t;
 748 }
 749
 750 void sum_epot(t_grpopts *opts,gmx_enerdata_t *enerd)
 751 {
 752   gmx_grppairener_t *grpp;
 753   real *epot;
 754   int i;
 755
 756   grpp = &enerd->grpp;
 757   epot = enerd->term;
 758
 759   /* Accumulate energies */
 760   epot[F_COUL_SR]  = sum_v(grpp->nener,grpp->ener[egCOULSR]);
 761   epot[F_LJ]       = sum_v(grpp->nener,grpp->ener[egLJSR]);
 762   epot[F_LJ14]     = sum_v(grpp->nener,grpp->ener[egLJ14]);
 763   epot[F_COUL14]   = sum_v(grpp->nener,grpp->ener[egCOUL14]);
 764   epot[F_COUL_LR]  = sum_v(grpp->nener,grpp->ener[egCOULLR]);
 765   epot[F_LJ_LR]    = sum_v(grpp->nener,grpp->ener[egLJLR]);
 766   /* We have already added 1-2,1-3, and 1-4 terms to F_GBPOL */
 767   epot[F_GBPOL]   += sum_v(grpp->nener,grpp->ener[egGB]);
 768
 769 /* lattice part of LR doesnt belong to any group
 770  * and has been added earlier
 771  */
 772   epot[F_BHAM]     = sum_v(grpp->nener,grpp->ener[egBHAMSR]);
 773   epot[F_BHAM_LR]  = sum_v(grpp->nener,grpp->ener[egBHAMLR]);
 774
 775   epot[F_EPOT] = 0;
 776   for(i=0; (i<F_EPOT); i++)
 777   {
 778       if (i != F_DISRESVIOL && i != F_ORIRESDEV)
 779       {
 780           epot[F_EPOT] += epot[i];
 781       }
 782   }
 783 }
 784
 785 void sum_dhdl(gmx_enerdata_t *enerd, real *lambda, t_lambda *fepvals)
 786 {
 787     int i,j,index;
 788     double dlam;
 789
 790     enerd->dvdl_lin[efptVDW] += enerd->term[F_DVDL_VDW];  /* include dispersion correction */
 791     enerd->term[F_DVDL] = 0.0;
 792     for (i=0;i<efptNR;i++)
 793     {
 794         if (fepvals->separate_dvdl[i])
 795         {
 796             /* could this be done more readably/compactly? */
 797             switch (i) {
 798             case (efptCOUL):
 799                 index = F_DVDL_COUL;
 800                 break;
 801             case (efptVDW):
 802                 index = F_DVDL_VDW;
 803                 break;
 804             case (efptBONDED):
 805                 index = F_DVDL_BONDED;
 806                 break;
 807             case (efptRESTRAINT):
 808                 index = F_DVDL_RESTRAINT;
 809                 break;
 810             case (efptMASS):
 811                 index = F_DKDL;
 812                 break;
 813             default:
 814                 index = F_DVDL;
 815                 break;
 816             }
 817             enerd->term[index] = enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 818             if (debug)
 819             {
 820                 fprintf(debug,"dvdl-%s[%2d]: %f: non-linear %f + linear %f\n",
 821                         efpt_names[i],i,enerd->term[index],enerd->dvdl_nonlin[i],enerd->dvdl_lin[i]);
 822             }
 823         }
 824         else
 825         {
 826             enerd->term[F_DVDL] += enerd->dvdl_lin[i] + enerd->dvdl_nonlin[i];
 827             if (debug)
 828             {
 829                 fprintf(debug,"dvd-%sl[%2d]: %f: non-linear %f + linear %f\n",
 830                         efpt_names[0],i,enerd->term[F_DVDL],enerd->dvdl_nonlin[i],enerd->dvdl_lin[i]);
 831             }
 832         }
 833     }
 834
 835     /* Notes on the foreign lambda free energy difference evaluation:
 836      * Adding the potential and ekin terms that depend linearly on lambda
 837      * as delta lam * dvdl to the energy differences is exact.
 838      * For the constraints this is not exact, but we have no other option
 839      * without literally changing the lengths and reevaluating the energies at each step.
 840      * (try to remedy this post 4.6 - MRS)
 841      * For the non-bonded LR term we assume that the soft-core (if present)
 842      * no longer affects the energy beyond the short-range cut-off,
 843      * which is a very good approximation (except for exotic settings).
 844      * (investigate how to overcome this post 4.6 - MRS)
 845      */
 846
 847     for(i=0; i<fepvals->n_lambda; i++)
 848     {                                         /* note we are iterating over fepvals here!
 849                                                  For the current lam, dlam = 0 automatically,
 850                                                  so we don't need to add anything to the
 851                                                  enerd->enerpart_lambda[0] */
 852
 853         /* we don't need to worry about dvdl contributions to the current lambda, because
 854            it's automatically zero */
 855
 856         /* first kinetic energy term */
 857         dlam = (fepvals->all_lambda[efptMASS][i] - lambda[efptMASS]);
 858
 859         enerd->enerpart_lambda[i+1] += enerd->term[F_DKDL]*dlam;
 860
 861         for (j=0;j<efptNR;j++)
 862         {
 863             if (j==efptMASS) {continue;} /* no other mass term to worry about */
 864
 865             dlam = (fepvals->all_lambda[j][i]-lambda[j]);
 866             enerd->enerpart_lambda[i+1] += dlam*enerd->dvdl_lin[j];
 867             if (debug)
 868             {
 869                 fprintf(debug,"enerdiff lam %g: (%15s), non-linear %f linear %f*%f\n",
 870                         fepvals->all_lambda[j][i],efpt_names[j],
 871                         (enerd->enerpart_lambda[i+1] - enerd->enerpart_lambda[0]),
 872                         dlam,enerd->dvdl_lin[j]);
 873             }
 874         }
 875     }
 876 }
 877
 878 void reset_enerdata(t_grpopts *opts,
 879                     t_forcerec *fr,gmx_bool bNS,
 880                     gmx_enerdata_t *enerd,
 881                     gmx_bool bMaster)
 882 {
 883     gmx_bool bKeepLR;
 884     int  i,j;
 885
 886     /* First reset all energy components, except for the long range terms
 887      * on the master at non neighbor search steps, since the long range
 888      * terms have already been summed at the last neighbor search step.
 889      */
 890     bKeepLR = (fr->bTwinRange && !bNS);
 891     for(i=0; (i<egNR); i++) {
 892         if (!(bKeepLR && bMaster && (i == egCOULLR || i == egLJLR))) {
 893             for(j=0; (j<enerd->grpp.nener); j++)
 894                 enerd->grpp.ener[i][j] = 0.0;
 895         }
 896     }
 897     for (i=0;i<efptNR;i++)
 898     {
 899         enerd->dvdl_lin[i]    = 0.0;
 900         enerd->dvdl_nonlin[i] = 0.0;
 901     }
 902
 903     /* Normal potential energy components */
 904     for(i=0; (i<=F_EPOT); i++) {
 905         enerd->term[i] = 0.0;
 906     }
 907     /* Initialize the dVdlambda term with the long range contribution */
 908     /* Initialize the dvdl term with the long range contribution */
 909     enerd->term[F_DVDL]            = 0.0;
 910     enerd->term[F_DVDL_COUL]       = 0.0;
 911     enerd->term[F_DVDL_VDW]        = 0.0;
 912     enerd->term[F_DVDL_BONDED]     = 0.0;
 913     enerd->term[F_DVDL_RESTRAINT]  = 0.0;
 914     enerd->term[F_DKDL]            = 0.0;
 915     if (enerd->n_lambda > 0)
 916     {
 917         for(i=0; i<enerd->n_lambda; i++)
 918         {
 919             enerd->enerpart_lambda[i] = 0.0;
 920         }
 921     }
 922 }