src/gromacs/mdlib/sim_util.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, 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 #ifdef HAVE_CONFIG_H
  38 #include <config.h>
  39 #endif
  40
  41 #include <stdio.h>
  42 #ifdef HAVE_SYS_TIME_H
  43 #include <sys/time.h>
  44 #endif
  45 #include <math.h>
  46 #include "typedefs.h"
  47 #include "string2.h"
  48 #include "smalloc.h"
  49 #include "names.h"
  50 #include "mvdata.h"
  51 #include "txtdump.h"
  52 #include "pbc.h"
  53 #include "chargegroup.h"
  54 #include "vec.h"
  55 #include "nrnb.h"
  56 #include "mshift.h"
  57 #include "mdrun.h"
  58 #include "sim_util.h"
  59 #include "update.h"
  60 #include "physics.h"
  61 #include "main.h"
  62 #include "mdatoms.h"
  63 #include "force.h"
  64 #include "bondf.h"
  65 #include "pme.h"
  66 #include "disre.h"
  67 #include "orires.h"
  68 #include "network.h"
  69 #include "calcmu.h"
  70 #include "constr.h"
  71 #include "xvgr.h"
  72 #include "copyrite.h"
  73 #include "pull_rotation.h"
  74 #include "gmx_random.h"
  75 #include "domdec.h"
  76 #include "partdec.h"
  77 #include "genborn.h"
  78 #include "nbnxn_atomdata.h"
  79 #include "nbnxn_search.h"
  80 #include "nbnxn_kernels/nbnxn_kernel_ref.h"
  81 #include "nbnxn_kernels/simd_4xn/nbnxn_kernel_simd_4xn.h"
  82 #include "nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn.h"
  83 #include "nbnxn_kernels/nbnxn_kernel_gpu_ref.h"
  84
  85 #include "gromacs/timing/wallcycle.h"
  86 #include "gromacs/timing/walltime_accounting.h"
  87 #include "gromacs/utility/gmxmpi.h"
  88
  89 #include "adress.h"
  90 #include "qmmm.h"
  91
  92 #include "nbnxn_cuda_data_mgmt.h"
  93 #include "nbnxn_cuda/nbnxn_cuda.h"
  94
  95 void print_time(FILE                     *out,
  96                 gmx_walltime_accounting_t walltime_accounting,
  97                 gmx_large_int_t           step,
  98                 t_inputrec               *ir,
  99                 t_commrec gmx_unused     *cr)
 100 {
 101     time_t finish;
 102     char   timebuf[STRLEN];
 103     double dt, elapsed_seconds, time_per_step;
 104     char   buf[48];
 105
 106 #ifndef GMX_THREAD_MPI
 107     if (!PAR(cr))
 108 #endif
 109     {
 110         fprintf(out, "\r");
 111     }
 112     fprintf(out, "step %s", gmx_step_str(step, buf));
 113     if ((step >= ir->nstlist))
 114     {
 115         double seconds_since_epoch = gmx_gettime();
 116         elapsed_seconds = seconds_since_epoch - walltime_accounting_get_start_time_stamp(walltime_accounting);
 117         time_per_step   = elapsed_seconds/(step - ir->init_step + 1);
 118         dt              = (ir->nsteps + ir->init_step - step) * time_per_step;
 119
 120         if (ir->nsteps >= 0)
 121         {
 122             if (dt >= 300)
 123             {
 124                 finish = (time_t) (seconds_since_epoch + dt);
 125                 gmx_ctime_r(&finish, timebuf, STRLEN);
 126                 sprintf(buf, "%s", timebuf);
 127                 buf[strlen(buf)-1] = '\0';
 128                 fprintf(out, ", will finish %s", buf);
 129             }
 130             else
 131             {
 132                 fprintf(out, ", remaining wall clock time: %5d s          ", (int)dt);
 133             }
 134         }
 135         else
 136         {
 137             fprintf(out, " performance: %.1f ns/day    ",
 138                     ir->delta_t/1000*24*60*60/time_per_step);
 139         }
 140     }
 141 #ifndef GMX_THREAD_MPI
 142     if (PAR(cr))
 143     {
 144         fprintf(out, "\n");
 145     }
 146 #endif
 147
 148     fflush(out);
 149 }
 150
 151 void print_date_and_time(FILE *fplog, int nodeid, const char *title,
 152                          const gmx_walltime_accounting_t walltime_accounting)
 153 {
 154     int    i;
 155     char   timebuf[STRLEN];
 156     char   time_string[STRLEN];
 157     time_t tmptime;
 158
 159     if (fplog)
 160     {
 161         if (walltime_accounting != NULL)
 162         {
 163             tmptime = (time_t) walltime_accounting_get_start_time_stamp(walltime_accounting);
 164             gmx_ctime_r(&tmptime, timebuf, STRLEN);
 165         }
 166         else
 167         {
 168             tmptime = (time_t) gmx_gettime();
 169             gmx_ctime_r(&tmptime, timebuf, STRLEN);
 170         }
 171         for (i = 0; timebuf[i] >= ' '; i++)
 172         {
 173             time_string[i] = timebuf[i];
 174         }
 175         time_string[i] = '\0';
 176
 177         fprintf(fplog, "%s on node %d %s\n", title, nodeid, time_string);
 178     }
 179 }
 180
 181 static void sum_forces(int start, int end, rvec f[], rvec flr[])
 182 {
 183     int i;
 184
 185     if (gmx_debug_at)
 186     {
 187         pr_rvecs(debug, 0, "fsr", f+start, end-start);
 188         pr_rvecs(debug, 0, "flr", flr+start, end-start);
 189     }
 190     for (i = start; (i < end); i++)
 191     {
 192         rvec_inc(f[i], flr[i]);
 193     }
 194 }
 195
 196 /*
 197  * calc_f_el calculates forces due to an electric field.
 198  *
 199  * force is kJ mol^-1 nm^-1 = e * kJ mol^-1 nm^-1 / e
 200  *
 201  * Et[] contains the parameters for the time dependent
 202  * part of the field (not yet used).
 203  * Ex[] contains the parameters for
 204  * the spatial dependent part of the field. You can have cool periodic
 205  * fields in principle, but only a constant field is supported
 206  * now.
 207  * The function should return the energy due to the electric field
 208  * (if any) but for now returns 0.
 209  *
 210  * WARNING:
 211  * There can be problems with the virial.
 212  * Since the field is not self-consistent this is unavoidable.
 213  * For neutral molecules the virial is correct within this approximation.
 214  * For neutral systems with many charged molecules the error is small.
 215  * But for systems with a net charge or a few charged molecules
 216  * the error can be significant when the field is high.
 217  * Solution: implement a self-consitent electric field into PME.
 218  */
 219 static void calc_f_el(FILE *fp, int  start, int homenr,
 220                       real charge[], rvec f[],
 221                       t_cosines Ex[], t_cosines Et[], double t)
 222 {
 223     rvec Ext;
 224     real t0;
 225     int  i, m;
 226
 227     for (m = 0; (m < DIM); m++)
 228     {
 229         if (Et[m].n > 0)
 230         {
 231             if (Et[m].n == 3)
 232             {
 233                 t0     = Et[m].a[1];
 234                 Ext[m] = cos(Et[m].a[0]*(t-t0))*exp(-sqr(t-t0)/(2.0*sqr(Et[m].a[2])));
 235             }
 236             else
 237             {
 238                 Ext[m] = cos(Et[m].a[0]*t);
 239             }
 240         }
 241         else
 242         {
 243             Ext[m] = 1.0;
 244         }
 245         if (Ex[m].n > 0)
 246         {
 247             /* Convert the field strength from V/nm to MD-units */
 248             Ext[m] *= Ex[m].a[0]*FIELDFAC;
 249             for (i = start; (i < start+homenr); i++)
 250             {
 251                 f[i][m] += charge[i]*Ext[m];
 252             }
 253         }
 254         else
 255         {
 256             Ext[m] = 0;
 257         }
 258     }
 259     if (fp != NULL)
 260     {
 261         fprintf(fp, "%10g  %10g  %10g  %10g #FIELD\n", t,
 262                 Ext[XX]/FIELDFAC, Ext[YY]/FIELDFAC, Ext[ZZ]/FIELDFAC);
 263     }
 264 }
 265
 266 static void calc_virial(int start, int homenr, rvec x[], rvec f[],
 267                         tensor vir_part, t_graph *graph, matrix box,
 268                         t_nrnb *nrnb, const t_forcerec *fr, int ePBC)
 269 {
 270     int    i, j;
 271     tensor virtest;
 272
 273     /* The short-range virial from surrounding boxes */
 274     clear_mat(vir_part);
 275     calc_vir(SHIFTS, fr->shift_vec, fr->fshift, vir_part, ePBC == epbcSCREW, box);
 276     inc_nrnb(nrnb, eNR_VIRIAL, SHIFTS);
 277
 278     /* Calculate partial virial, for local atoms only, based on short range.
 279      * Total virial is computed in global_stat, called from do_md
 280      */
 281     f_calc_vir(start, start+homenr, x, f, vir_part, graph, box);
 282     inc_nrnb(nrnb, eNR_VIRIAL, homenr);
 283
 284     /* Add position restraint contribution */
 285     for (i = 0; i < DIM; i++)
 286     {
 287         vir_part[i][i] += fr->vir_diag_posres[i];
 288     }
 289
 290     /* Add wall contribution */
 291     for (i = 0; i < DIM; i++)
 292     {
 293         vir_part[i][ZZ] += fr->vir_wall_z[i];
 294     }
 295
 296     if (debug)
 297     {
 298         pr_rvecs(debug, 0, "vir_part", vir_part, DIM);
 299     }
 300 }
 301
 302 static void posres_wrapper(FILE *fplog,
 303                            int flags,
 304                            gmx_bool bSepDVDL,
 305                            t_inputrec *ir,
 306                            t_nrnb *nrnb,
 307                            gmx_localtop_t *top,
 308                            matrix box, rvec x[],
 309                            gmx_enerdata_t *enerd,
 310                            real *lambda,
 311                            t_forcerec *fr)
 312 {
 313     t_pbc pbc;
 314     real  v, dvdl;
 315     int   i;
 316
 317     /* Position restraints always require full pbc */
 318     set_pbc(&pbc, ir->ePBC, box);
 319     dvdl = 0;
 320     v    = posres(top->idef.il[F_POSRES].nr, top->idef.il[F_POSRES].iatoms,
 321                   top->idef.iparams_posres,
 322                   (const rvec*)x, fr->f_novirsum, fr->vir_diag_posres,
 323                   ir->ePBC == epbcNONE ? NULL : &pbc,
 324                   lambda[efptRESTRAINT], &dvdl,
 325                   fr->rc_scaling, fr->ePBC, fr->posres_com, fr->posres_comB);
 326     if (bSepDVDL)
 327     {
 328         gmx_print_sepdvdl(fplog, interaction_function[F_POSRES].longname, v, dvdl);
 329     }
 330     enerd->term[F_POSRES] += v;
 331     /* If just the force constant changes, the FEP term is linear,
 332      * but if k changes, it is not.
 333      */
 334     enerd->dvdl_nonlin[efptRESTRAINT] += dvdl;
 335     inc_nrnb(nrnb, eNR_POSRES, top->idef.il[F_POSRES].nr/2);
 336
 337     if ((ir->fepvals->n_lambda > 0) && (flags & GMX_FORCE_DHDL))
 338     {
 339         for (i = 0; i < enerd->n_lambda; i++)
 340         {
 341             real dvdl_dum, lambda_dum;
 342
 343             lambda_dum = (i == 0 ? lambda[efptRESTRAINT] : ir->fepvals->all_lambda[efptRESTRAINT][i-1]);
 344             v          = posres(top->idef.il[F_POSRES].nr, top->idef.il[F_POSRES].iatoms,
 345                                 top->idef.iparams_posres,
 346                                 (const rvec*)x, NULL, NULL,
 347                                 ir->ePBC == epbcNONE ? NULL : &pbc, lambda_dum, &dvdl,
 348                                 fr->rc_scaling, fr->ePBC, fr->posres_com, fr->posres_comB);
 349             enerd->enerpart_lambda[i] += v;
 350         }
 351     }
 352 }
 353
 354 static void pull_potential_wrapper(FILE *fplog,
 355                                    gmx_bool bSepDVDL,
 356                                    t_commrec *cr,
 357                                    t_inputrec *ir,
 358                                    matrix box, rvec x[],
 359                                    rvec f[],
 360                                    tensor vir_force,
 361                                    t_mdatoms *mdatoms,
 362                                    gmx_enerdata_t *enerd,
 363                                    real *lambda,
 364                                    double t)
 365 {
 366     t_pbc  pbc;
 367     real   dvdl;
 368
 369     /* Calculate the center of mass forces, this requires communication,
 370      * which is why pull_potential is called close to other communication.
 371      * The virial contribution is calculated directly,
 372      * which is why we call pull_potential after calc_virial.
 373      */
 374     set_pbc(&pbc, ir->ePBC, box);
 375     dvdl                     = 0;
 376     enerd->term[F_COM_PULL] +=
 377         pull_potential(ir->ePull, ir->pull, mdatoms, &pbc,
 378                        cr, t, lambda[efptRESTRAINT], x, f, vir_force, &dvdl);
 379     if (bSepDVDL)
 380     {
 381         gmx_print_sepdvdl(fplog, "Com pull", enerd->term[F_COM_PULL], dvdl);
 382     }
 383     enerd->dvdl_lin[efptRESTRAINT] += dvdl;
 384 }
 385
 386 static void pme_receive_force_ener(FILE           *fplog,
 387                                    gmx_bool        bSepDVDL,
 388                                    t_commrec      *cr,
 389                                    gmx_wallcycle_t wcycle,
 390                                    gmx_enerdata_t *enerd,
 391                                    t_forcerec     *fr)
 392 {
 393     real   e_q, e_lj, v, dvdl_q, dvdl_lj;
 394     float  cycles_ppdpme, cycles_seppme;
 395
 396     cycles_ppdpme = wallcycle_stop(wcycle, ewcPPDURINGPME);
 397     dd_cycles_add(cr->dd, cycles_ppdpme, ddCyclPPduringPME);
 398
 399     /* In case of node-splitting, the PP nodes receive the long-range
 400      * forces, virial and energy from the PME nodes here.
 401      */
 402     wallcycle_start(wcycle, ewcPP_PMEWAITRECVF);
 403     dvdl_q  = 0;
 404     dvdl_lj = 0;
 405     gmx_pme_receive_f(cr, fr->f_novirsum, fr->vir_el_recip, &e_q,
 406                       fr->vir_lj_recip, &e_lj, &dvdl_q, &dvdl_lj,
 407                       &cycles_seppme);
 408     if (bSepDVDL)
 409     {
 410         gmx_print_sepdvdl(fplog, "Electrostatic PME mesh", e_q, dvdl_q);
 411         gmx_print_sepdvdl(fplog, "Lennard-Jones PME mesh", e_lj, dvdl_lj);
 412     }
 413     enerd->term[F_COUL_RECIP] += e_q;
 414     enerd->term[F_LJ_RECIP]   += e_lj;
 415     enerd->dvdl_lin[efptCOUL] += dvdl_q;
 416     enerd->dvdl_lin[efptVDW]  += dvdl_lj;
 417
 418     if (wcycle)
 419     {
 420         dd_cycles_add(cr->dd, cycles_seppme, ddCyclPME);
 421     }
 422     wallcycle_stop(wcycle, ewcPP_PMEWAITRECVF);
 423 }
 424
 425 static void print_large_forces(FILE *fp, t_mdatoms *md, t_commrec *cr,
 426                                gmx_large_int_t step, real pforce, rvec *x, rvec *f)
 427 {
 428     int  i;
 429     real pf2, fn2;
 430     char buf[STEPSTRSIZE];
 431
 432     pf2 = sqr(pforce);
 433     for (i = md->start; i < md->start+md->homenr; i++)
 434     {
 435         fn2 = norm2(f[i]);
 436         /* We also catch NAN, if the compiler does not optimize this away. */
 437         if (fn2 >= pf2 || fn2 != fn2)
 438         {
 439             fprintf(fp, "step %s  atom %6d  x %8.3f %8.3f %8.3f  force %12.5e\n",
 440                     gmx_step_str(step, buf),
 441                     ddglatnr(cr->dd, i), x[i][XX], x[i][YY], x[i][ZZ], sqrt(fn2));
 442         }
 443     }
 444 }
 445
 446 static void post_process_forces(t_commrec *cr,
 447                                 gmx_large_int_t step,
 448                                 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
 449                                 gmx_localtop_t *top,
 450                                 matrix box, rvec x[],
 451                                 rvec f[],
 452                                 tensor vir_force,
 453                                 t_mdatoms *mdatoms,
 454                                 t_graph *graph,
 455                                 t_forcerec *fr, gmx_vsite_t *vsite,
 456                                 int flags)
 457 {
 458     if (fr->bF_NoVirSum)
 459     {
 460         if (vsite)
 461         {
 462             /* Spread the mesh force on virtual sites to the other particles...
 463              * This is parallellized. MPI communication is performed
 464              * if the constructing atoms aren't local.
 465              */
 466             wallcycle_start(wcycle, ewcVSITESPREAD);
 467             spread_vsite_f(vsite, x, fr->f_novirsum, NULL,
 468                            (flags & GMX_FORCE_VIRIAL), fr->vir_el_recip,
 469                            nrnb,
 470                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr);
 471             wallcycle_stop(wcycle, ewcVSITESPREAD);
 472         }
 473         if (flags & GMX_FORCE_VIRIAL)
 474         {
 475             /* Now add the forces, this is local */
 476             if (fr->bDomDec)
 477             {
 478                 sum_forces(0, fr->f_novirsum_n, f, fr->f_novirsum);
 479             }
 480             else
 481             {
 482                 sum_forces(mdatoms->start, mdatoms->start+mdatoms->homenr,
 483                            f, fr->f_novirsum);
 484             }
 485             if (EEL_FULL(fr->eeltype))
 486             {
 487                 /* Add the mesh contribution to the virial */
 488                 m_add(vir_force, fr->vir_el_recip, vir_force);
 489             }
 490             if (EVDW_PME(fr->vdwtype))
 491             {
 492                 /* Add the mesh contribution to the virial */
 493                 m_add(vir_force, fr->vir_lj_recip, vir_force);
 494             }
 495             if (debug)
 496             {
 497                 pr_rvecs(debug, 0, "vir_force", vir_force, DIM);
 498             }
 499         }
 500     }
 501
 502     if (fr->print_force >= 0)
 503     {
 504         print_large_forces(stderr, mdatoms, cr, step, fr->print_force, x, f);
 505     }
 506 }
 507
 508 static void do_nb_verlet(t_forcerec *fr,
 509                          interaction_const_t *ic,
 510                          gmx_enerdata_t *enerd,
 511                          int flags, int ilocality,
 512                          int clearF,
 513                          t_nrnb *nrnb,
 514                          gmx_wallcycle_t wcycle)
 515 {
 516     int                        nnbl, kernel_type, enr_nbnxn_kernel_ljc, enr_nbnxn_kernel_lj;
 517     char                      *env;
 518     nonbonded_verlet_group_t  *nbvg;
 519     gmx_bool                   bCUDA;
 520
 521     if (!(flags & GMX_FORCE_NONBONDED))
 522     {
 523         /* skip non-bonded calculation */
 524         return;
 525     }
 526
 527     nbvg = &fr->nbv->grp[ilocality];
 528
 529     /* CUDA kernel launch overhead is already timed separately */
 530     if (fr->cutoff_scheme != ecutsVERLET)
 531     {
 532         gmx_incons("Invalid cut-off scheme passed!");
 533     }
 534
 535     bCUDA = (nbvg->kernel_type == nbnxnk8x8x8_CUDA);
 536
 537     if (!bCUDA)
 538     {
 539         wallcycle_sub_start(wcycle, ewcsNONBONDED);
 540     }
 541     switch (nbvg->kernel_type)
 542     {
 543         case nbnxnk4x4_PlainC:
 544             nbnxn_kernel_ref(&nbvg->nbl_lists,
 545                              nbvg->nbat, ic,
 546                              fr->shift_vec,
 547                              flags,
 548                              clearF,
 549                              fr->fshift[0],
 550                              enerd->grpp.ener[egCOULSR],
 551                              fr->bBHAM ?
 552                              enerd->grpp.ener[egBHAMSR] :
 553                              enerd->grpp.ener[egLJSR]);
 554             break;
 555
 556         case nbnxnk4xN_SIMD_4xN:
 557             nbnxn_kernel_simd_4xn(&nbvg->nbl_lists,
 558                                   nbvg->nbat, ic,
 559                                   nbvg->ewald_excl,
 560                                   fr->shift_vec,
 561                                   flags,
 562                                   clearF,
 563                                   fr->fshift[0],
 564                                   enerd->grpp.ener[egCOULSR],
 565                                   fr->bBHAM ?
 566                                   enerd->grpp.ener[egBHAMSR] :
 567                                   enerd->grpp.ener[egLJSR]);
 568             break;
 569         case nbnxnk4xN_SIMD_2xNN:
 570             nbnxn_kernel_simd_2xnn(&nbvg->nbl_lists,
 571                                    nbvg->nbat, ic,
 572                                    nbvg->ewald_excl,
 573                                    fr->shift_vec,
 574                                    flags,
 575                                    clearF,
 576                                    fr->fshift[0],
 577                                    enerd->grpp.ener[egCOULSR],
 578                                    fr->bBHAM ?
 579                                    enerd->grpp.ener[egBHAMSR] :
 580                                    enerd->grpp.ener[egLJSR]);
 581             break;
 582
 583         case nbnxnk8x8x8_CUDA:
 584             nbnxn_cuda_launch_kernel(fr->nbv->cu_nbv, nbvg->nbat, flags, ilocality);
 585             break;
 586
 587         case nbnxnk8x8x8_PlainC:
 588             nbnxn_kernel_gpu_ref(nbvg->nbl_lists.nbl[0],
 589                                  nbvg->nbat, ic,
 590                                  fr->shift_vec,
 591                                  flags,
 592                                  clearF,
 593                                  nbvg->nbat->out[0].f,
 594                                  fr->fshift[0],
 595                                  enerd->grpp.ener[egCOULSR],
 596                                  fr->bBHAM ?
 597                                  enerd->grpp.ener[egBHAMSR] :
 598                                  enerd->grpp.ener[egLJSR]);
 599             break;
 600
 601         default:
 602             gmx_incons("Invalid nonbonded kernel type passed!");
 603
 604     }
 605     if (!bCUDA)
 606     {
 607         wallcycle_sub_stop(wcycle, ewcsNONBONDED);
 608     }
 609
 610     if (EEL_RF(ic->eeltype) || ic->eeltype == eelCUT)
 611     {
 612         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_RF;
 613     }
 614     else if ((!bCUDA && nbvg->ewald_excl == ewaldexclAnalytical) ||
 615              (bCUDA && nbnxn_cuda_is_kernel_ewald_analytical(fr->nbv->cu_nbv)))
 616     {
 617         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_EWALD;
 618     }
 619     else
 620     {
 621         enr_nbnxn_kernel_ljc = eNR_NBNXN_LJ_TAB;
 622     }
 623     enr_nbnxn_kernel_lj = eNR_NBNXN_LJ;
 624     if (flags & GMX_FORCE_ENERGY)
 625     {
 626         /* In eNR_??? the nbnxn F+E kernels are always the F kernel + 1 */
 627         enr_nbnxn_kernel_ljc += 1;
 628         enr_nbnxn_kernel_lj  += 1;
 629     }
 630
 631     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc,
 632              nbvg->nbl_lists.natpair_ljq);
 633     inc_nrnb(nrnb, enr_nbnxn_kernel_lj,
 634              nbvg->nbl_lists.natpair_lj);
 635     inc_nrnb(nrnb, enr_nbnxn_kernel_ljc-eNR_NBNXN_LJ_RF+eNR_NBNXN_RF,
 636              nbvg->nbl_lists.natpair_q);
 637 }
 638
 639 void do_force_cutsVERLET(FILE *fplog, t_commrec *cr,
 640                          t_inputrec *inputrec,
 641                          gmx_large_int_t step, t_nrnb *nrnb, gmx_wallcycle_t wcycle,
 642                          gmx_localtop_t *top,
 643                          gmx_groups_t gmx_unused *groups,
 644                          matrix box, rvec x[], history_t *hist,
 645                          rvec f[],
 646                          tensor vir_force,
 647                          t_mdatoms *mdatoms,
 648                          gmx_enerdata_t *enerd, t_fcdata *fcd,
 649                          real *lambda, t_graph *graph,
 650                          t_forcerec *fr, interaction_const_t *ic,
 651                          gmx_vsite_t *vsite, rvec mu_tot,
 652                          double t, FILE *field, gmx_edsam_t ed,
 653                          gmx_bool bBornRadii,
 654                          int flags)
 655 {
 656     int                 cg0, cg1, i, j;
 657     int                 start, homenr;
 658     int                 nb_kernel_type;
 659     double              mu[2*DIM];
 660     gmx_bool            bSepDVDL, bStateChanged, bNS, bFillGrid, bCalcCGCM, bBS;
 661     gmx_bool            bDoLongRange, bDoForces, bSepLRF, bUseGPU, bUseOrEmulGPU;
 662     gmx_bool            bDiffKernels = FALSE;
 663     matrix              boxs;
 664     rvec                vzero, box_diag;
 665     real                e, v, dvdl;
 666     float               cycles_pme, cycles_force, cycles_wait_gpu;
 667     nonbonded_verlet_t *nbv;
 668
 669     cycles_force    = 0;
 670     cycles_wait_gpu = 0;
 671     nbv             = fr->nbv;
 672     nb_kernel_type  = fr->nbv->grp[0].kernel_type;
 673
 674     start  = mdatoms->start;
 675     homenr = mdatoms->homenr;
 676
 677     bSepDVDL = (fr->bSepDVDL && do_per_step(step, inputrec->nstlog));
 678
 679     clear_mat(vir_force);
 680
 681     cg0 = 0;
 682     if (DOMAINDECOMP(cr))
 683     {
 684         cg1 = cr->dd->ncg_tot;
 685     }
 686     else
 687     {
 688         cg1 = top->cgs.nr;
 689     }
 690     if (fr->n_tpi > 0)
 691     {
 692         cg1--;
 693     }
 694
 695     bStateChanged = (flags & GMX_FORCE_STATECHANGED);
 696     bNS           = (flags & GMX_FORCE_NS) && (fr->bAllvsAll == FALSE);
 697     bFillGrid     = (bNS && bStateChanged);
 698     bCalcCGCM     = (bFillGrid && !DOMAINDECOMP(cr));
 699     bDoLongRange  = (fr->bTwinRange && bNS && (flags & GMX_FORCE_DO_LR));
 700     bDoForces     = (flags & GMX_FORCE_FORCES);
 701     bSepLRF       = (bDoLongRange && bDoForces && (flags & GMX_FORCE_SEPLRF));
 702     bUseGPU       = fr->nbv->bUseGPU;
 703     bUseOrEmulGPU = bUseGPU || (nbv->grp[0].kernel_type == nbnxnk8x8x8_PlainC);
 704
 705     if (bStateChanged)
 706     {
 707         update_forcerec(fr, box);
 708
 709         if (NEED_MUTOT(*inputrec))
 710         {
 711             /* Calculate total (local) dipole moment in a temporary common array.
 712              * This makes it possible to sum them over nodes faster.
 713              */
 714             calc_mu(start, homenr,
 715                     x, mdatoms->chargeA, mdatoms->chargeB, mdatoms->nChargePerturbed,
 716                     mu, mu+DIM);
 717         }
 718     }
 719
 720     if (fr->ePBC != epbcNONE)
 721     {
 722         /* Compute shift vectors every step,
 723          * because of pressure coupling or box deformation!
 724          */
 725         if ((flags & GMX_FORCE_DYNAMICBOX) && bStateChanged)
 726         {
 727             calc_shifts(box, fr->shift_vec);
 728         }
 729
 730         if (bCalcCGCM)
 731         {
 732             put_atoms_in_box_omp(fr->ePBC, box, homenr, x);
 733             inc_nrnb(nrnb, eNR_SHIFTX, homenr);
 734         }
 735         else if (EI_ENERGY_MINIMIZATION(inputrec->eI) && graph)
 736         {
 737             unshift_self(graph, box, x);
 738         }
 739     }
 740
 741     nbnxn_atomdata_copy_shiftvec(flags & GMX_FORCE_DYNAMICBOX,
 742                                  fr->shift_vec, nbv->grp[0].nbat);
 743
 744 #ifdef GMX_MPI
 745     if (!(cr->duty & DUTY_PME))
 746     {
 747         /* Send particle coordinates to the pme nodes.
 748          * Since this is only implemented for domain decomposition
 749          * and domain decomposition does not use the graph,
 750          * we do not need to worry about shifting.
 751          */
 752
 753         int pme_flags = 0;
 754
 755         wallcycle_start(wcycle, ewcPP_PMESENDX);
 756
 757         bBS = (inputrec->nwall == 2);
 758         if (bBS)
 759         {
 760             copy_mat(box, boxs);
 761             svmul(inputrec->wall_ewald_zfac, boxs[ZZ], boxs[ZZ]);
 762         }
 763
 764         if (EEL_PME(fr->eeltype))
 765         {
 766             pme_flags |= GMX_PME_DO_COULOMB;
 767         }
 768
 769         if (EVDW_PME(fr->vdwtype))
 770         {
 771             pme_flags |= GMX_PME_DO_LJ;
 772             if (fr->ljpme_combination_rule == eljpmeLB)
 773             {
 774                 pme_flags |= GMX_PME_LJ_LB;
 775             }
 776         }
 777
 778         gmx_pme_send_coordinates(cr, bBS ? boxs : box, x,
 779                                  mdatoms->nChargePerturbed, mdatoms->nTypePerturbed, lambda[efptCOUL], lambda[efptVDW],
 780                                  (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY)),
 781                                  pme_flags, step);
 782
 783         wallcycle_stop(wcycle, ewcPP_PMESENDX);
 784     }
 785 #endif /* GMX_MPI */
 786
 787     /* do gridding for pair search */
 788     if (bNS)
 789     {
 790         if (graph && bStateChanged)
 791         {
 792             /* Calculate intramolecular shift vectors to make molecules whole */
 793             mk_mshift(fplog, graph, fr->ePBC, box, x);
 794         }
 795
 796         clear_rvec(vzero);
 797         box_diag[XX] = box[XX][XX];
 798         box_diag[YY] = box[YY][YY];
 799         box_diag[ZZ] = box[ZZ][ZZ];
 800
 801         wallcycle_start(wcycle, ewcNS);
 802         if (!fr->bDomDec)
 803         {
 804             wallcycle_sub_start(wcycle, ewcsNBS_GRID_LOCAL);
 805             nbnxn_put_on_grid(nbv->nbs, fr->ePBC, box,
 806                               0, vzero, box_diag,
 807                               0, mdatoms->homenr, -1, fr->cginfo, x,
 808                               0, NULL,
 809                               nbv->grp[eintLocal].kernel_type,
 810                               nbv->grp[eintLocal].nbat);
 811             wallcycle_sub_stop(wcycle, ewcsNBS_GRID_LOCAL);
 812         }
 813         else
 814         {
 815             wallcycle_sub_start(wcycle, ewcsNBS_GRID_NONLOCAL);
 816             nbnxn_put_on_grid_nonlocal(nbv->nbs, domdec_zones(cr->dd),
 817                                        fr->cginfo, x,
 818                                        nbv->grp[eintNonlocal].kernel_type,
 819                                        nbv->grp[eintNonlocal].nbat);
 820             wallcycle_sub_stop(wcycle, ewcsNBS_GRID_NONLOCAL);
 821         }
 822
 823         if (nbv->ngrp == 1 ||
 824             nbv->grp[eintNonlocal].nbat == nbv->grp[eintLocal].nbat)
 825         {
 826             nbnxn_atomdata_set(nbv->grp[eintLocal].nbat, eatAll,
 827                                nbv->nbs, mdatoms, fr->cginfo);
 828         }
 829         else
 830         {
 831             nbnxn_atomdata_set(nbv->grp[eintLocal].nbat, eatLocal,
 832                                nbv->nbs, mdatoms, fr->cginfo);
 833             nbnxn_atomdata_set(nbv->grp[eintNonlocal].nbat, eatAll,
 834                                nbv->nbs, mdatoms, fr->cginfo);
 835         }
 836         wallcycle_stop(wcycle, ewcNS);
 837     }
 838
 839     /* initialize the GPU atom data and copy shift vector */
 840     if (bUseGPU)
 841     {
 842         if (bNS)
 843         {
 844             wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU_NB);
 845             nbnxn_cuda_init_atomdata(nbv->cu_nbv, nbv->grp[eintLocal].nbat);
 846             wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
 847         }
 848
 849         wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU_NB);
 850         nbnxn_cuda_upload_shiftvec(nbv->cu_nbv, nbv->grp[eintLocal].nbat);
 851         wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
 852     }
 853
 854     /* do local pair search */
 855     if (bNS)
 856     {
 857         wallcycle_start_nocount(wcycle, ewcNS);
 858         wallcycle_sub_start(wcycle, ewcsNBS_SEARCH_LOCAL);
 859         nbnxn_make_pairlist(nbv->nbs, nbv->grp[eintLocal].nbat,
 860                             &top->excls,
 861                             ic->rlist,
 862                             nbv->min_ci_balanced,
 863                             &nbv->grp[eintLocal].nbl_lists,
 864                             eintLocal,
 865                             nbv->grp[eintLocal].kernel_type,
 866                             nrnb);
 867         wallcycle_sub_stop(wcycle, ewcsNBS_SEARCH_LOCAL);
 868
 869         if (bUseGPU)
 870         {
 871             /* initialize local pair-list on the GPU */
 872             nbnxn_cuda_init_pairlist(nbv->cu_nbv,
 873                                      nbv->grp[eintLocal].nbl_lists.nbl[0],
 874                                      eintLocal);
 875         }
 876         wallcycle_stop(wcycle, ewcNS);
 877     }
 878     else
 879     {
 880         wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
 881         wallcycle_sub_start(wcycle, ewcsNB_X_BUF_OPS);
 882         nbnxn_atomdata_copy_x_to_nbat_x(nbv->nbs, eatLocal, FALSE, x,
 883                                         nbv->grp[eintLocal].nbat);
 884         wallcycle_sub_stop(wcycle, ewcsNB_X_BUF_OPS);
 885         wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
 886     }
 887
 888     if (bUseGPU)
 889     {
 890         wallcycle_start(wcycle, ewcLAUNCH_GPU_NB);
 891         /* launch local nonbonded F on GPU */
 892         do_nb_verlet(fr, ic, enerd, flags, eintLocal, enbvClearFNo,
 893                      nrnb, wcycle);
 894         wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
 895     }
 896
 897     /* Communicate coordinates and sum dipole if necessary +
 898        do non-local pair search */
 899     if (DOMAINDECOMP(cr))
 900     {
 901         bDiffKernels = (nbv->grp[eintNonlocal].kernel_type !=
 902                         nbv->grp[eintLocal].kernel_type);
 903
 904         if (bDiffKernels)
 905         {
 906             /* With GPU+CPU non-bonded calculations we need to copy
 907              * the local coordinates to the non-local nbat struct
 908              * (in CPU format) as the non-local kernel call also
 909              * calculates the local - non-local interactions.
 910              */
 911             wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
 912             wallcycle_sub_start(wcycle, ewcsNB_X_BUF_OPS);
 913             nbnxn_atomdata_copy_x_to_nbat_x(nbv->nbs, eatLocal, TRUE, x,
 914                                             nbv->grp[eintNonlocal].nbat);
 915             wallcycle_sub_stop(wcycle, ewcsNB_X_BUF_OPS);
 916             wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
 917         }
 918
 919         if (bNS)
 920         {
 921             wallcycle_start_nocount(wcycle, ewcNS);
 922             wallcycle_sub_start(wcycle, ewcsNBS_SEARCH_NONLOCAL);
 923
 924             if (bDiffKernels)
 925             {
 926                 nbnxn_grid_add_simple(nbv->nbs, nbv->grp[eintNonlocal].nbat);
 927             }
 928
 929             nbnxn_make_pairlist(nbv->nbs, nbv->grp[eintNonlocal].nbat,
 930                                 &top->excls,
 931                                 ic->rlist,
 932                                 nbv->min_ci_balanced,
 933                                 &nbv->grp[eintNonlocal].nbl_lists,
 934                                 eintNonlocal,
 935                                 nbv->grp[eintNonlocal].kernel_type,
 936                                 nrnb);
 937
 938             wallcycle_sub_stop(wcycle, ewcsNBS_SEARCH_NONLOCAL);
 939
 940             if (nbv->grp[eintNonlocal].kernel_type == nbnxnk8x8x8_CUDA)
 941             {
 942                 /* initialize non-local pair-list on the GPU */
 943                 nbnxn_cuda_init_pairlist(nbv->cu_nbv,
 944                                          nbv->grp[eintNonlocal].nbl_lists.nbl[0],
 945                                          eintNonlocal);
 946             }
 947             wallcycle_stop(wcycle, ewcNS);
 948         }
 949         else
 950         {
 951             wallcycle_start(wcycle, ewcMOVEX);
 952             dd_move_x(cr->dd, box, x);
 953
 954             /* When we don't need the total dipole we sum it in global_stat */
 955             if (bStateChanged && NEED_MUTOT(*inputrec))
 956             {
 957                 gmx_sumd(2*DIM, mu, cr);
 958             }
 959             wallcycle_stop(wcycle, ewcMOVEX);
 960
 961             wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
 962             wallcycle_sub_start(wcycle, ewcsNB_X_BUF_OPS);
 963             nbnxn_atomdata_copy_x_to_nbat_x(nbv->nbs, eatNonlocal, FALSE, x,
 964                                             nbv->grp[eintNonlocal].nbat);
 965             wallcycle_sub_stop(wcycle, ewcsNB_X_BUF_OPS);
 966             cycles_force += wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
 967         }
 968
 969         if (bUseGPU && !bDiffKernels)
 970         {
 971             wallcycle_start(wcycle, ewcLAUNCH_GPU_NB);
 972             /* launch non-local nonbonded F on GPU */
 973             do_nb_verlet(fr, ic, enerd, flags, eintNonlocal, enbvClearFNo,
 974                          nrnb, wcycle);
 975             cycles_force += wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
 976         }
 977     }
 978
 979     if (bUseGPU)
 980     {
 981         /* launch D2H copy-back F */
 982         wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU_NB);
 983         if (DOMAINDECOMP(cr) && !bDiffKernels)
 984         {
 985             nbnxn_cuda_launch_cpyback(nbv->cu_nbv, nbv->grp[eintNonlocal].nbat,
 986                                       flags, eatNonlocal);
 987         }
 988         nbnxn_cuda_launch_cpyback(nbv->cu_nbv, nbv->grp[eintLocal].nbat,
 989                                   flags, eatLocal);
 990         cycles_force += wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
 991     }
 992
 993     if (bStateChanged && NEED_MUTOT(*inputrec))
 994     {
 995         if (PAR(cr))
 996         {
 997             gmx_sumd(2*DIM, mu, cr);
 998         }
 999
1000         for (i = 0; i < 2; i++)
1001         {
1002             for (j = 0; j < DIM; j++)
1003             {
1004                 fr->mu_tot[i][j] = mu[i*DIM + j];
1005             }
1006         }
1007     }
1008     if (fr->efep == efepNO)
1009     {
1010         copy_rvec(fr->mu_tot[0], mu_tot);
1011     }
1012     else
1013     {
1014         for (j = 0; j < DIM; j++)
1015         {
1016             mu_tot[j] =
1017                 (1.0 - lambda[efptCOUL])*fr->mu_tot[0][j] +
1018                 lambda[efptCOUL]*fr->mu_tot[1][j];
1019         }
1020     }
1021
1022     /* Reset energies */
1023     reset_enerdata(fr, bNS, enerd, MASTER(cr));
1024     clear_rvecs(SHIFTS, fr->fshift);
1025
1026     if (DOMAINDECOMP(cr))
1027     {
1028         if (!(cr->duty & DUTY_PME))
1029         {
1030             wallcycle_start(wcycle, ewcPPDURINGPME);
1031             dd_force_flop_start(cr->dd, nrnb);
1032         }
1033     }
1034
1035     if (inputrec->bRot)
1036     {
1037         /* Enforced rotation has its own cycle counter that starts after the collective
1038          * coordinates have been communicated. It is added to ddCyclF to allow
1039          * for proper load-balancing */
1040         wallcycle_start(wcycle, ewcROT);
1041         do_rotation(cr, inputrec, box, x, t, step, wcycle, bNS);
1042         wallcycle_stop(wcycle, ewcROT);
1043     }
1044
1045     /* Start the force cycle counter.
1046      * This counter is stopped in do_forcelow_level.
1047      * No parallel communication should occur while this counter is running,
1048      * since that will interfere with the dynamic load balancing.
1049      */
1050     wallcycle_start(wcycle, ewcFORCE);
1051     if (bDoForces)
1052     {
1053         /* Reset forces for which the virial is calculated separately:
1054          * PME/Ewald forces if necessary */
1055         if (fr->bF_NoVirSum)
1056         {
1057             if (flags & GMX_FORCE_VIRIAL)
1058             {
1059                 fr->f_novirsum = fr->f_novirsum_alloc;
1060                 if (fr->bDomDec)
1061                 {
1062                     clear_rvecs(fr->f_novirsum_n, fr->f_novirsum);
1063                 }
1064                 else
1065                 {
1066                     clear_rvecs(homenr, fr->f_novirsum+start);
1067                 }
1068             }
1069             else
1070             {
1071                 /* We are not calculating the pressure so we do not need
1072                  * a separate array for forces that do not contribute
1073                  * to the pressure.
1074                  */
1075                 fr->f_novirsum = f;
1076             }
1077         }
1078
1079         /* Clear the short- and long-range forces */
1080         clear_rvecs(fr->natoms_force_constr, f);
1081         if (bSepLRF && do_per_step(step, inputrec->nstcalclr))
1082         {
1083             clear_rvecs(fr->natoms_force_constr, fr->f_twin);
1084         }
1085
1086         clear_rvec(fr->vir_diag_posres);
1087     }
1088
1089     if (inputrec->ePull == epullCONSTRAINT)
1090     {
1091         clear_pull_forces(inputrec->pull);
1092     }
1093
1094     /* We calculate the non-bonded forces, when done on the CPU, here.
1095      * We do this before calling do_force_lowlevel, as in there bondeds
1096      * forces are calculated before PME, which does communication.
1097      * With this order, non-bonded and bonded force calculation imbalance
1098      * can be balanced out by the domain decomposition load balancing.
1099      */
1100
1101     if (!bUseOrEmulGPU)
1102     {
1103         /* Maybe we should move this into do_force_lowlevel */
1104         do_nb_verlet(fr, ic, enerd, flags, eintLocal, enbvClearFYes,
1105                      nrnb, wcycle);
1106     }
1107
1108     if (!bUseOrEmulGPU || bDiffKernels)
1109     {
1110         int aloc;
1111
1112         if (DOMAINDECOMP(cr))
1113         {
1114             do_nb_verlet(fr, ic, enerd, flags, eintNonlocal,
1115                          bDiffKernels ? enbvClearFYes : enbvClearFNo,
1116                          nrnb, wcycle);
1117         }
1118
1119         if (!bUseOrEmulGPU)
1120         {
1121             aloc = eintLocal;
1122         }
1123         else
1124         {
1125             aloc = eintNonlocal;
1126         }
1127
1128         /* Add all the non-bonded force to the normal force array.
1129          * This can be split into a local a non-local part when overlapping
1130          * communication with calculation with domain decomposition.
1131          */
1132         cycles_force += wallcycle_stop(wcycle, ewcFORCE);
1133         wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
1134         wallcycle_sub_start(wcycle, ewcsNB_F_BUF_OPS);
1135         nbnxn_atomdata_add_nbat_f_to_f(nbv->nbs, eatAll, nbv->grp[aloc].nbat, f);
1136         wallcycle_sub_stop(wcycle, ewcsNB_F_BUF_OPS);
1137         cycles_force += wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
1138         wallcycle_start_nocount(wcycle, ewcFORCE);
1139
1140         /* if there are multiple fshift output buffers reduce them */
1141         if ((flags & GMX_FORCE_VIRIAL) &&
1142             nbv->grp[aloc].nbl_lists.nnbl > 1)
1143         {
1144             nbnxn_atomdata_add_nbat_fshift_to_fshift(nbv->grp[aloc].nbat,
1145                                                      fr->fshift);
1146         }
1147     }
1148
1149     /* update QMMMrec, if necessary */
1150     if (fr->bQMMM)
1151     {
1152         update_QMMMrec(cr, fr, x, mdatoms, box, top);
1153     }
1154
1155     if ((flags & GMX_FORCE_BONDED) && top->idef.il[F_POSRES].nr > 0)
1156     {
1157         posres_wrapper(fplog, flags, bSepDVDL, inputrec, nrnb, top, box, x,
1158                        enerd, lambda, fr);
1159     }
1160
1161     /* Compute the bonded and non-bonded energies and optionally forces */
1162     do_force_lowlevel(fplog, step, fr, inputrec, &(top->idef),
1163                       cr, nrnb, wcycle, mdatoms,
1164                       x, hist, f, bSepLRF ? fr->f_twin : f, enerd, fcd, top, fr->born,
1165                       &(top->atomtypes), bBornRadii, box,
1166                       inputrec->fepvals, lambda, graph, &(top->excls), fr->mu_tot,
1167                       flags, &cycles_pme);
1168
1169     if (bSepLRF)
1170     {
1171         if (do_per_step(step, inputrec->nstcalclr))
1172         {
1173             /* Add the long range forces to the short range forces */
1174             for (i = 0; i < fr->natoms_force_constr; i++)
1175             {
1176                 rvec_add(fr->f_twin[i], f[i], f[i]);
1177             }
1178         }
1179     }
1180
1181     cycles_force += wallcycle_stop(wcycle, ewcFORCE);
1182
1183     if (ed)
1184     {
1185         do_flood(cr, inputrec, x, f, ed, box, step, bNS);
1186     }
1187
1188     if (bUseOrEmulGPU && !bDiffKernels)
1189     {
1190         /* wait for non-local forces (or calculate in emulation mode) */
1191         if (DOMAINDECOMP(cr))
1192         {
1193             if (bUseGPU)
1194             {
1195                 float cycles_tmp;
1196
1197                 wallcycle_start(wcycle, ewcWAIT_GPU_NB_NL);
1198                 nbnxn_cuda_wait_gpu(nbv->cu_nbv,
1199                                     nbv->grp[eintNonlocal].nbat,
1200                                     flags, eatNonlocal,
1201                                     enerd->grpp.ener[egLJSR], enerd->grpp.ener[egCOULSR],
1202                                     fr->fshift);
1203                 cycles_tmp       = wallcycle_stop(wcycle, ewcWAIT_GPU_NB_NL);
1204                 cycles_wait_gpu += cycles_tmp;
1205                 cycles_force    += cycles_tmp;
1206             }
1207             else
1208             {
1209                 wallcycle_start_nocount(wcycle, ewcFORCE);
1210                 do_nb_verlet(fr, ic, enerd, flags, eintNonlocal, enbvClearFYes,
1211                              nrnb, wcycle);
1212                 cycles_force += wallcycle_stop(wcycle, ewcFORCE);
1213             }
1214             wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
1215             wallcycle_sub_start(wcycle, ewcsNB_F_BUF_OPS);
1216             /* skip the reduction if there was no non-local work to do */
1217             if (nbv->grp[eintLocal].nbl_lists.nbl[0]->nsci > 0)
1218             {
1219                 nbnxn_atomdata_add_nbat_f_to_f(nbv->nbs, eatNonlocal,
1220                                                nbv->grp[eintNonlocal].nbat, f);
1221             }
1222             wallcycle_sub_stop(wcycle, ewcsNB_F_BUF_OPS);
1223             cycles_force += wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
1224         }
1225     }
1226
1227     if (bDoForces)
1228     {
1229         /* Communicate the forces */
1230         if (PAR(cr))
1231         {
1232             wallcycle_start(wcycle, ewcMOVEF);
1233             if (DOMAINDECOMP(cr))
1234             {
1235                 dd_move_f(cr->dd, f, fr->fshift);
1236                 /* Do we need to communicate the separate force array
1237                  * for terms that do not contribute to the single sum virial?
1238                  * Position restraints and electric fields do not introduce
1239                  * inter-cg forces, only full electrostatics methods do.
1240                  * When we do not calculate the virial, fr->f_novirsum = f,
1241                  * so we have already communicated these forces.
1242                  */
1243                 if (EEL_FULL(fr->eeltype) && cr->dd->n_intercg_excl &&
1244                     (flags & GMX_FORCE_VIRIAL))
1245                 {
1246                     dd_move_f(cr->dd, fr->f_novirsum, NULL);
1247                 }
1248                 if (bSepLRF)
1249                 {
1250                     /* We should not update the shift forces here,
1251                      * since f_twin is already included in f.
1252                      */
1253                     dd_move_f(cr->dd, fr->f_twin, NULL);
1254                 }
1255             }
1256             wallcycle_stop(wcycle, ewcMOVEF);
1257         }
1258     }
1259
1260     if (bUseOrEmulGPU)
1261     {
1262         /* wait for local forces (or calculate in emulation mode) */
1263         if (bUseGPU)
1264         {
1265             wallcycle_start(wcycle, ewcWAIT_GPU_NB_L);
1266             nbnxn_cuda_wait_gpu(nbv->cu_nbv,
1267                                 nbv->grp[eintLocal].nbat,
1268                                 flags, eatLocal,
1269                                 enerd->grpp.ener[egLJSR], enerd->grpp.ener[egCOULSR],
1270                                 fr->fshift);
1271             cycles_wait_gpu += wallcycle_stop(wcycle, ewcWAIT_GPU_NB_L);
1272
1273             /* now clear the GPU outputs while we finish the step on the CPU */
1274
1275             wallcycle_start_nocount(wcycle, ewcLAUNCH_GPU_NB);
1276             nbnxn_cuda_clear_outputs(nbv->cu_nbv, flags);
1277             wallcycle_stop(wcycle, ewcLAUNCH_GPU_NB);
1278         }
1279         else
1280         {
1281             wallcycle_start_nocount(wcycle, ewcFORCE);
1282             do_nb_verlet(fr, ic, enerd, flags, eintLocal,
1283                          DOMAINDECOMP(cr) ? enbvClearFNo : enbvClearFYes,
1284                          nrnb, wcycle);
1285             wallcycle_stop(wcycle, ewcFORCE);
1286         }
1287         wallcycle_start(wcycle, ewcNB_XF_BUF_OPS);
1288         wallcycle_sub_start(wcycle, ewcsNB_F_BUF_OPS);
1289         if (nbv->grp[eintLocal].nbl_lists.nbl[0]->nsci > 0)
1290         {
1291             /* skip the reduction if there was no non-local work to do */
1292             nbnxn_atomdata_add_nbat_f_to_f(nbv->nbs, eatLocal,
1293                                            nbv->grp[eintLocal].nbat, f);
1294         }
1295         wallcycle_sub_stop(wcycle, ewcsNB_F_BUF_OPS);
1296         wallcycle_stop(wcycle, ewcNB_XF_BUF_OPS);
1297     }
1298
1299     if (DOMAINDECOMP(cr))
1300     {
1301         dd_force_flop_stop(cr->dd, nrnb);
1302         if (wcycle)
1303         {
1304             dd_cycles_add(cr->dd, cycles_force-cycles_pme, ddCyclF);
1305             if (bUseGPU)
1306             {
1307                 dd_cycles_add(cr->dd, cycles_wait_gpu, ddCyclWaitGPU);
1308             }
1309         }
1310     }
1311
1312     if (bDoForces)
1313     {
1314         if (IR_ELEC_FIELD(*inputrec))
1315         {
1316             /* Compute forces due to electric field */
1317             calc_f_el(MASTER(cr) ? field : NULL,
1318                       start, homenr, mdatoms->chargeA, fr->f_novirsum,
1319                       inputrec->ex, inputrec->et, t);
1320         }
1321
1322         /* If we have NoVirSum forces, but we do not calculate the virial,
1323          * we sum fr->f_novirum=f later.
1324          */
1325         if (vsite && !(fr->bF_NoVirSum && !(flags & GMX_FORCE_VIRIAL)))
1326         {
1327             wallcycle_start(wcycle, ewcVSITESPREAD);
1328             spread_vsite_f(vsite, x, f, fr->fshift, FALSE, NULL, nrnb,
1329                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr);
1330             wallcycle_stop(wcycle, ewcVSITESPREAD);
1331
1332             if (bSepLRF)
1333             {
1334                 wallcycle_start(wcycle, ewcVSITESPREAD);
1335                 spread_vsite_f(vsite, x, fr->f_twin, NULL, FALSE, NULL,
1336                                nrnb,
1337                                &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr);
1338                 wallcycle_stop(wcycle, ewcVSITESPREAD);
1339             }
1340         }
1341
1342         if (flags & GMX_FORCE_VIRIAL)
1343         {
1344             /* Calculation of the virial must be done after vsites! */
1345             calc_virial(mdatoms->start, mdatoms->homenr, x, f,
1346                         vir_force, graph, box, nrnb, fr, inputrec->ePBC);
1347         }
1348     }
1349
1350     if (inputrec->ePull == epullUMBRELLA || inputrec->ePull == epullCONST_F)
1351     {
1352         pull_potential_wrapper(fplog, bSepDVDL, cr, inputrec, box, x,
1353                                f, vir_force, mdatoms, enerd, lambda, t);
1354     }
1355
1356     /* Add the forces from enforced rotation potentials (if any) */
1357     if (inputrec->bRot)
1358     {
1359         wallcycle_start(wcycle, ewcROTadd);
1360         enerd->term[F_COM_PULL] += add_rot_forces(inputrec->rot, f, cr, step, t);
1361         wallcycle_stop(wcycle, ewcROTadd);
1362     }
1363
1364     if (PAR(cr) && !(cr->duty & DUTY_PME))
1365     {
1366         /* In case of node-splitting, the PP nodes receive the long-range
1367          * forces, virial and energy from the PME nodes here.
1368          */
1369         pme_receive_force_ener(fplog, bSepDVDL, cr, wcycle, enerd, fr);
1370     }
1371
1372     if (bDoForces)
1373     {
1374         post_process_forces(cr, step, nrnb, wcycle,
1375                             top, box, x, f, vir_force, mdatoms, graph, fr, vsite,
1376                             flags);
1377     }
1378
1379     /* Sum the potential energy terms from group contributions */
1380     sum_epot(&(enerd->grpp), enerd->term);
1381 }
1382
1383 void do_force_cutsGROUP(FILE *fplog, t_commrec *cr,
1384                         t_inputrec *inputrec,
1385                         gmx_large_int_t step, t_nrnb *nrnb, gmx_wallcycle_t wcycle,
1386                         gmx_localtop_t *top,
1387                         gmx_groups_t *groups,
1388                         matrix box, rvec x[], history_t *hist,
1389                         rvec f[],
1390                         tensor vir_force,
1391                         t_mdatoms *mdatoms,
1392                         gmx_enerdata_t *enerd, t_fcdata *fcd,
1393                         real *lambda, t_graph *graph,
1394                         t_forcerec *fr, gmx_vsite_t *vsite, rvec mu_tot,
1395                         double t, FILE *field, gmx_edsam_t ed,
1396                         gmx_bool bBornRadii,
1397                         int flags)
1398 {
1399     int        cg0, cg1, i, j;
1400     int        start, homenr;
1401     double     mu[2*DIM];
1402     gmx_bool   bSepDVDL, bStateChanged, bNS, bFillGrid, bCalcCGCM, bBS;
1403     gmx_bool   bDoLongRangeNS, bDoForces, bDoPotential, bSepLRF;
1404     gmx_bool   bDoAdressWF;
1405     matrix     boxs;
1406     rvec       vzero, box_diag;
1407     real       e, v, dvdlambda[efptNR];
1408     t_pbc      pbc;
1409     float      cycles_pme, cycles_force;
1410
1411     start  = mdatoms->start;
1412     homenr = mdatoms->homenr;
1413
1414     bSepDVDL = (fr->bSepDVDL && do_per_step(step, inputrec->nstlog));
1415
1416     clear_mat(vir_force);
1417
1418     if (PARTDECOMP(cr))
1419     {
1420         pd_cg_range(cr, &cg0, &cg1);
1421     }
1422     else
1423     {
1424         cg0 = 0;
1425         if (DOMAINDECOMP(cr))
1426         {
1427             cg1 = cr->dd->ncg_tot;
1428         }
1429         else
1430         {
1431             cg1 = top->cgs.nr;
1432         }
1433         if (fr->n_tpi > 0)
1434         {
1435             cg1--;
1436         }
1437     }
1438
1439     bStateChanged  = (flags & GMX_FORCE_STATECHANGED);
1440     bNS            = (flags & GMX_FORCE_NS) && (fr->bAllvsAll == FALSE);
1441     /* Should we update the long-range neighborlists at this step? */
1442     bDoLongRangeNS = fr->bTwinRange && bNS;
1443     /* Should we perform the long-range nonbonded evaluation inside the neighborsearching? */
1444     bFillGrid      = (bNS && bStateChanged);
1445     bCalcCGCM      = (bFillGrid && !DOMAINDECOMP(cr));
1446     bDoForces      = (flags & GMX_FORCE_FORCES);
1447     bDoPotential   = (flags & GMX_FORCE_ENERGY);
1448     bSepLRF        = ((inputrec->nstcalclr > 1) && bDoForces &&
1449                       (flags & GMX_FORCE_SEPLRF) && (flags & GMX_FORCE_DO_LR));
1450
1451     /* should probably move this to the forcerec since it doesn't change */
1452     bDoAdressWF   = ((fr->adress_type != eAdressOff));
1453
1454     if (bStateChanged)
1455     {
1456         update_forcerec(fr, box);
1457
1458         if (NEED_MUTOT(*inputrec))
1459         {
1460             /* Calculate total (local) dipole moment in a temporary common array.
1461              * This makes it possible to sum them over nodes faster.
1462              */
1463             calc_mu(start, homenr,
1464                     x, mdatoms->chargeA, mdatoms->chargeB, mdatoms->nChargePerturbed,
1465                     mu, mu+DIM);
1466         }
1467     }
1468
1469     if (fr->ePBC != epbcNONE)
1470     {
1471         /* Compute shift vectors every step,
1472          * because of pressure coupling or box deformation!
1473          */
1474         if ((flags & GMX_FORCE_DYNAMICBOX) && bStateChanged)
1475         {
1476             calc_shifts(box, fr->shift_vec);
1477         }
1478
1479         if (bCalcCGCM)
1480         {
1481             put_charge_groups_in_box(fplog, cg0, cg1, fr->ePBC, box,
1482                                      &(top->cgs), x, fr->cg_cm);
1483             inc_nrnb(nrnb, eNR_CGCM, homenr);
1484             inc_nrnb(nrnb, eNR_RESETX, cg1-cg0);
1485         }
1486         else if (EI_ENERGY_MINIMIZATION(inputrec->eI) && graph)
1487         {
1488             unshift_self(graph, box, x);
1489         }
1490     }
1491     else if (bCalcCGCM)
1492     {
1493         calc_cgcm(fplog, cg0, cg1, &(top->cgs), x, fr->cg_cm);
1494         inc_nrnb(nrnb, eNR_CGCM, homenr);
1495     }
1496
1497     if (bCalcCGCM)
1498     {
1499         if (PAR(cr))
1500         {
1501             move_cgcm(fplog, cr, fr->cg_cm);
1502         }
1503         if (gmx_debug_at)
1504         {
1505             pr_rvecs(debug, 0, "cgcm", fr->cg_cm, top->cgs.nr);
1506         }
1507     }
1508
1509 #ifdef GMX_MPI
1510     if (!(cr->duty & DUTY_PME))
1511     {
1512         /* Send particle coordinates to the pme nodes.
1513          * Since this is only implemented for domain decomposition
1514          * and domain decomposition does not use the graph,
1515          * we do not need to worry about shifting.
1516          */
1517
1518         int pme_flags = 0;
1519
1520         wallcycle_start(wcycle, ewcPP_PMESENDX);
1521
1522         bBS = (inputrec->nwall == 2);
1523         if (bBS)
1524         {
1525             copy_mat(box, boxs);
1526             svmul(inputrec->wall_ewald_zfac, boxs[ZZ], boxs[ZZ]);
1527         }
1528
1529         if (EEL_PME(fr->eeltype))
1530         {
1531             pme_flags |= GMX_PME_DO_COULOMB;
1532         }
1533
1534         if (EVDW_PME(fr->vdwtype))
1535         {
1536             pme_flags |= GMX_PME_DO_LJ;
1537             if (fr->ljpme_combination_rule == eljpmeLB)
1538             {
1539                 pme_flags |= GMX_PME_LJ_LB;
1540             }
1541         }
1542
1543         gmx_pme_send_coordinates(cr, bBS ? boxs : box, x,
1544                                  mdatoms->nChargePerturbed, mdatoms->nTypePerturbed, lambda[efptCOUL], lambda[efptVDW],
1545                                  (flags & (GMX_FORCE_VIRIAL | GMX_FORCE_ENERGY)),
1546                                  pme_flags, step);
1547
1548         wallcycle_stop(wcycle, ewcPP_PMESENDX);
1549     }
1550 #endif /* GMX_MPI */
1551
1552     /* Communicate coordinates and sum dipole if necessary */
1553     if (PAR(cr))
1554     {
1555         wallcycle_start(wcycle, ewcMOVEX);
1556         if (DOMAINDECOMP(cr))
1557         {
1558             dd_move_x(cr->dd, box, x);
1559         }
1560         else
1561         {
1562             move_x(cr, x, nrnb);
1563         }
1564         wallcycle_stop(wcycle, ewcMOVEX);
1565     }
1566
1567     /* update adress weight beforehand */
1568     if (bStateChanged && bDoAdressWF)
1569     {
1570         /* need pbc for adress weight calculation with pbc_dx */
1571         set_pbc(&pbc, inputrec->ePBC, box);
1572         if (fr->adress_site == eAdressSITEcog)
1573         {
1574             update_adress_weights_cog(top->idef.iparams, top->idef.il, x, fr, mdatoms,
1575                                       inputrec->ePBC == epbcNONE ? NULL : &pbc);
1576         }
1577         else if (fr->adress_site == eAdressSITEcom)
1578         {
1579             update_adress_weights_com(fplog, cg0, cg1, &(top->cgs), x, fr, mdatoms,
1580                                       inputrec->ePBC == epbcNONE ? NULL : &pbc);
1581         }
1582         else if (fr->adress_site == eAdressSITEatomatom)
1583         {
1584             update_adress_weights_atom_per_atom(cg0, cg1, &(top->cgs), x, fr, mdatoms,
1585                                                 inputrec->ePBC == epbcNONE ? NULL : &pbc);
1586         }
1587         else
1588         {
1589             update_adress_weights_atom(cg0, cg1, &(top->cgs), x, fr, mdatoms,
1590                                        inputrec->ePBC == epbcNONE ? NULL : &pbc);
1591         }
1592     }
1593
1594     if (NEED_MUTOT(*inputrec))
1595     {
1596
1597         if (bStateChanged)
1598         {
1599             if (PAR(cr))
1600             {
1601                 gmx_sumd(2*DIM, mu, cr);
1602             }
1603             for (i = 0; i < 2; i++)
1604             {
1605                 for (j = 0; j < DIM; j++)
1606                 {
1607                     fr->mu_tot[i][j] = mu[i*DIM + j];
1608                 }
1609             }
1610         }
1611         if (fr->efep == efepNO)
1612         {
1613             copy_rvec(fr->mu_tot[0], mu_tot);
1614         }
1615         else
1616         {
1617             for (j = 0; j < DIM; j++)
1618             {
1619                 mu_tot[j] =
1620                     (1.0 - lambda[efptCOUL])*fr->mu_tot[0][j] + lambda[efptCOUL]*fr->mu_tot[1][j];
1621             }
1622         }
1623     }
1624
1625     /* Reset energies */
1626     reset_enerdata(fr, bNS, enerd, MASTER(cr));
1627     clear_rvecs(SHIFTS, fr->fshift);
1628
1629     if (bNS)
1630     {
1631         wallcycle_start(wcycle, ewcNS);
1632
1633         if (graph && bStateChanged)
1634         {
1635             /* Calculate intramolecular shift vectors to make molecules whole */
1636             mk_mshift(fplog, graph, fr->ePBC, box, x);
1637         }
1638
1639         /* Do the actual neighbour searching */
1640         ns(fplog, fr, box,
1641            groups, top, mdatoms,
1642            cr, nrnb, bFillGrid,
1643            bDoLongRangeNS);
1644
1645         wallcycle_stop(wcycle, ewcNS);
1646     }
1647
1648     if (inputrec->implicit_solvent && bNS)
1649     {
1650         make_gb_nblist(cr, inputrec->gb_algorithm,
1651                        x, box, fr, &top->idef, graph, fr->born);
1652     }
1653
1654     if (DOMAINDECOMP(cr))
1655     {
1656         if (!(cr->duty & DUTY_PME))
1657         {
1658             wallcycle_start(wcycle, ewcPPDURINGPME);
1659             dd_force_flop_start(cr->dd, nrnb);
1660         }
1661     }
1662
1663     if (inputrec->bRot)
1664     {
1665         /* Enforced rotation has its own cycle counter that starts after the collective
1666          * coordinates have been communicated. It is added to ddCyclF to allow
1667          * for proper load-balancing */
1668         wallcycle_start(wcycle, ewcROT);
1669         do_rotation(cr, inputrec, box, x, t, step, wcycle, bNS);
1670         wallcycle_stop(wcycle, ewcROT);
1671     }
1672
1673     /* Start the force cycle counter.
1674      * This counter is stopped in do_forcelow_level.
1675      * No parallel communication should occur while this counter is running,
1676      * since that will interfere with the dynamic load balancing.
1677      */
1678     wallcycle_start(wcycle, ewcFORCE);
1679
1680     if (bDoForces)
1681     {
1682         /* Reset forces for which the virial is calculated separately:
1683          * PME/Ewald forces if necessary */
1684         if (fr->bF_NoVirSum)
1685         {
1686             if (flags & GMX_FORCE_VIRIAL)
1687             {
1688                 fr->f_novirsum = fr->f_novirsum_alloc;
1689                 if (fr->bDomDec)
1690                 {
1691                     clear_rvecs(fr->f_novirsum_n, fr->f_novirsum);
1692                 }
1693                 else
1694                 {
1695                     clear_rvecs(homenr, fr->f_novirsum+start);
1696                 }
1697             }
1698             else
1699             {
1700                 /* We are not calculating the pressure so we do not need
1701                  * a separate array for forces that do not contribute
1702                  * to the pressure.
1703                  */
1704                 fr->f_novirsum = f;
1705             }
1706         }
1707
1708         /* Clear the short- and long-range forces */
1709         clear_rvecs(fr->natoms_force_constr, f);
1710         if (bSepLRF && do_per_step(step, inputrec->nstcalclr))
1711         {
1712             clear_rvecs(fr->natoms_force_constr, fr->f_twin);
1713         }
1714
1715         clear_rvec(fr->vir_diag_posres);
1716     }
1717     if (inputrec->ePull == epullCONSTRAINT)
1718     {
1719         clear_pull_forces(inputrec->pull);
1720     }
1721
1722     /* update QMMMrec, if necessary */
1723     if (fr->bQMMM)
1724     {
1725         update_QMMMrec(cr, fr, x, mdatoms, box, top);
1726     }
1727
1728     if ((flags & GMX_FORCE_BONDED) && top->idef.il[F_POSRES].nr > 0)
1729     {
1730         posres_wrapper(fplog, flags, bSepDVDL, inputrec, nrnb, top, box, x,
1731                        enerd, lambda, fr);
1732     }
1733
1734     if ((flags & GMX_FORCE_BONDED) && top->idef.il[F_FBPOSRES].nr > 0)
1735     {
1736         /* Flat-bottomed position restraints always require full pbc */
1737         if (!(bStateChanged && bDoAdressWF))
1738         {
1739             set_pbc(&pbc, inputrec->ePBC, box);
1740         }
1741         v = fbposres(top->idef.il[F_FBPOSRES].nr, top->idef.il[F_FBPOSRES].iatoms,
1742                      top->idef.iparams_fbposres,
1743                      (const rvec*)x, fr->f_novirsum, fr->vir_diag_posres,
1744                      inputrec->ePBC == epbcNONE ? NULL : &pbc,
1745                      fr->rc_scaling, fr->ePBC, fr->posres_com);
1746         enerd->term[F_FBPOSRES] += v;
1747         inc_nrnb(nrnb, eNR_FBPOSRES, top->idef.il[F_FBPOSRES].nr/2);
1748     }
1749
1750     /* Compute the bonded and non-bonded energies and optionally forces */
1751     do_force_lowlevel(fplog, step, fr, inputrec, &(top->idef),
1752                       cr, nrnb, wcycle, mdatoms,
1753                       x, hist, f, bSepLRF ? fr->f_twin : f, enerd, fcd, top, fr->born,
1754                       &(top->atomtypes), bBornRadii, box,
1755                       inputrec->fepvals, lambda,
1756                       graph, &(top->excls), fr->mu_tot,
1757                       flags,
1758                       &cycles_pme);
1759
1760     if (bSepLRF)
1761     {
1762         if (do_per_step(step, inputrec->nstcalclr))
1763         {
1764             /* Add the long range forces to the short range forces */
1765             for (i = 0; i < fr->natoms_force_constr; i++)
1766             {
1767                 rvec_add(fr->f_twin[i], f[i], f[i]);
1768             }
1769         }
1770     }
1771
1772     cycles_force = wallcycle_stop(wcycle, ewcFORCE);
1773
1774     if (ed)
1775     {
1776         do_flood(cr, inputrec, x, f, ed, box, step, bNS);
1777     }
1778
1779     if (DOMAINDECOMP(cr))
1780     {
1781         dd_force_flop_stop(cr->dd, nrnb);
1782         if (wcycle)
1783         {
1784             dd_cycles_add(cr->dd, cycles_force-cycles_pme, ddCyclF);
1785         }
1786     }
1787
1788     if (bDoForces)
1789     {
1790         if (IR_ELEC_FIELD(*inputrec))
1791         {
1792             /* Compute forces due to electric field */
1793             calc_f_el(MASTER(cr) ? field : NULL,
1794                       start, homenr, mdatoms->chargeA, fr->f_novirsum,
1795                       inputrec->ex, inputrec->et, t);
1796         }
1797
1798         if (bDoAdressWF && fr->adress_icor == eAdressICThermoForce)
1799         {
1800             /* Compute thermodynamic force in hybrid AdResS region */
1801             adress_thermo_force(start, homenr, &(top->cgs), x, fr->f_novirsum, fr, mdatoms,
1802                                 inputrec->ePBC == epbcNONE ? NULL : &pbc);
1803         }
1804
1805         /* Communicate the forces */
1806         if (PAR(cr))
1807         {
1808             wallcycle_start(wcycle, ewcMOVEF);
1809             if (DOMAINDECOMP(cr))
1810             {
1811                 dd_move_f(cr->dd, f, fr->fshift);
1812                 /* Do we need to communicate the separate force array
1813                  * for terms that do not contribute to the single sum virial?
1814                  * Position restraints and electric fields do not introduce
1815                  * inter-cg forces, only full electrostatics methods do.
1816                  * When we do not calculate the virial, fr->f_novirsum = f,
1817                  * so we have already communicated these forces.
1818                  */
1819                 if (EEL_FULL(fr->eeltype) && cr->dd->n_intercg_excl &&
1820                     (flags & GMX_FORCE_VIRIAL))
1821                 {
1822                     dd_move_f(cr->dd, fr->f_novirsum, NULL);
1823                 }
1824                 if (bSepLRF)
1825                 {
1826                     /* We should not update the shift forces here,
1827                      * since f_twin is already included in f.
1828                      */
1829                     dd_move_f(cr->dd, fr->f_twin, NULL);
1830                 }
1831             }
1832             else
1833             {
1834                 pd_move_f(cr, f, nrnb);
1835                 if (bSepLRF)
1836                 {
1837                     pd_move_f(cr, fr->f_twin, nrnb);
1838                 }
1839             }
1840             wallcycle_stop(wcycle, ewcMOVEF);
1841         }
1842
1843         /* If we have NoVirSum forces, but we do not calculate the virial,
1844          * we sum fr->f_novirum=f later.
1845          */
1846         if (vsite && !(fr->bF_NoVirSum && !(flags & GMX_FORCE_VIRIAL)))
1847         {
1848             wallcycle_start(wcycle, ewcVSITESPREAD);
1849             spread_vsite_f(vsite, x, f, fr->fshift, FALSE, NULL, nrnb,
1850                            &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr);
1851             wallcycle_stop(wcycle, ewcVSITESPREAD);
1852
1853             if (bSepLRF)
1854             {
1855                 wallcycle_start(wcycle, ewcVSITESPREAD);
1856                 spread_vsite_f(vsite, x, fr->f_twin, NULL, FALSE, NULL,
1857                                nrnb,
1858                                &top->idef, fr->ePBC, fr->bMolPBC, graph, box, cr);
1859                 wallcycle_stop(wcycle, ewcVSITESPREAD);
1860             }
1861         }
1862
1863         if (flags & GMX_FORCE_VIRIAL)
1864         {
1865             /* Calculation of the virial must be done after vsites! */
1866             calc_virial(mdatoms->start, mdatoms->homenr, x, f,
1867                         vir_force, graph, box, nrnb, fr, inputrec->ePBC);
1868         }
1869     }
1870
1871     if (inputrec->ePull == epullUMBRELLA || inputrec->ePull == epullCONST_F)
1872     {
1873         pull_potential_wrapper(fplog, bSepDVDL, cr, inputrec, box, x,
1874                                f, vir_force, mdatoms, enerd, lambda, t);
1875     }
1876
1877     /* Add the forces from enforced rotation potentials (if any) */
1878     if (inputrec->bRot)
1879     {
1880         wallcycle_start(wcycle, ewcROTadd);
1881         enerd->term[F_COM_PULL] += add_rot_forces(inputrec->rot, f, cr, step, t);
1882         wallcycle_stop(wcycle, ewcROTadd);
1883     }
1884
1885     if (PAR(cr) && !(cr->duty & DUTY_PME))
1886     {
1887         /* In case of node-splitting, the PP nodes receive the long-range
1888          * forces, virial and energy from the PME nodes here.
1889          */
1890         pme_receive_force_ener(fplog, bSepDVDL, cr, wcycle, enerd, fr);
1891     }
1892
1893     if (bDoForces)
1894     {
1895         post_process_forces(cr, step, nrnb, wcycle,
1896                             top, box, x, f, vir_force, mdatoms, graph, fr, vsite,
1897                             flags);
1898     }
1899
1900     /* Sum the potential energy terms from group contributions */
1901     sum_epot(&(enerd->grpp), enerd->term);
1902 }
1903
1904 void do_force(FILE *fplog, t_commrec *cr,
1905               t_inputrec *inputrec,
1906               gmx_large_int_t step, t_nrnb *nrnb, gmx_wallcycle_t wcycle,
1907               gmx_localtop_t *top,
1908               gmx_groups_t *groups,
1909               matrix box, rvec x[], history_t *hist,
1910               rvec f[],
1911               tensor vir_force,
1912               t_mdatoms *mdatoms,
1913               gmx_enerdata_t *enerd, t_fcdata *fcd,
1914               real *lambda, t_graph *graph,
1915               t_forcerec *fr,
1916               gmx_vsite_t *vsite, rvec mu_tot,
1917               double t, FILE *field, gmx_edsam_t ed,
1918               gmx_bool bBornRadii,
1919               int flags)
1920 {
1921     /* modify force flag if not doing nonbonded */
1922     if (!fr->bNonbonded)
1923     {
1924         flags &= ~GMX_FORCE_NONBONDED;
1925     }
1926
1927     switch (inputrec->cutoff_scheme)
1928     {
1929         case ecutsVERLET:
1930             do_force_cutsVERLET(fplog, cr, inputrec,
1931                                 step, nrnb, wcycle,
1932                                 top,
1933                                 groups,
1934                                 box, x, hist,
1935                                 f, vir_force,
1936                                 mdatoms,
1937                                 enerd, fcd,
1938                                 lambda, graph,
1939                                 fr, fr->ic,
1940                                 vsite, mu_tot,
1941                                 t, field, ed,
1942                                 bBornRadii,
1943                                 flags);
1944             break;
1945         case ecutsGROUP:
1946             do_force_cutsGROUP(fplog, cr, inputrec,
1947                                step, nrnb, wcycle,
1948                                top,
1949                                groups,
1950                                box, x, hist,
1951                                f, vir_force,
1952                                mdatoms,
1953                                enerd, fcd,
1954                                lambda, graph,
1955                                fr, vsite, mu_tot,
1956                                t, field, ed,
1957                                bBornRadii,
1958                                flags);
1959             break;
1960         default:
1961             gmx_incons("Invalid cut-off scheme passed!");
1962     }
1963 }
1964
1965
1966 void do_constrain_first(FILE *fplog, gmx_constr_t constr,
1967                         t_inputrec *ir, t_mdatoms *md,
1968                         t_state *state, t_commrec *cr, t_nrnb *nrnb,
1969                         t_forcerec *fr, gmx_localtop_t *top)
1970 {
1971     int             i, m, start, end;
1972     gmx_large_int_t step;
1973     real            dt = ir->delta_t;
1974     real            dvdl_dum;
1975     rvec           *savex;
1976
1977     snew(savex, state->natoms);
1978
1979     start = md->start;
1980     end   = md->homenr + start;
1981
1982     if (debug)
1983     {
1984         fprintf(debug, "vcm: start=%d, homenr=%d, end=%d\n",
1985                 start, md->homenr, end);
1986     }
1987     /* Do a first constrain to reset particles... */
1988     step = ir->init_step;
1989     if (fplog)
1990     {
1991         char buf[STEPSTRSIZE];
1992         fprintf(fplog, "\nConstraining the starting coordinates (step %s)\n",
1993                 gmx_step_str(step, buf));
1994     }
1995     dvdl_dum = 0;
1996
1997     /* constrain the current position */
1998     constrain(NULL, TRUE, FALSE, constr, &(top->idef),
1999               ir, NULL, cr, step, 0, md,
2000               state->x, state->x, NULL,
2001               fr->bMolPBC, state->box,
2002               state->lambda[efptBONDED], &dvdl_dum,
2003               NULL, NULL, nrnb, econqCoord,
2004               ir->epc == epcMTTK, state->veta, state->veta);
2005     if (EI_VV(ir->eI))
2006     {
2007         /* constrain the inital velocity, and save it */
2008         /* also may be useful if we need the ekin from the halfstep for velocity verlet */
2009         /* might not yet treat veta correctly */
2010         constrain(NULL, TRUE, FALSE, constr, &(top->idef),
2011                   ir, NULL, cr, step, 0, md,
2012                   state->x, state->v, state->v,
2013                   fr->bMolPBC, state->box,
2014                   state->lambda[efptBONDED], &dvdl_dum,
2015                   NULL, NULL, nrnb, econqVeloc,
2016                   ir->epc == epcMTTK, state->veta, state->veta);
2017     }
2018     /* constrain the inital velocities at t-dt/2 */
2019     if (EI_STATE_VELOCITY(ir->eI) && ir->eI != eiVV)
2020     {
2021         for (i = start; (i < end); i++)
2022         {
2023             for (m = 0; (m < DIM); m++)
2024             {
2025                 /* Reverse the velocity */
2026                 state->v[i][m] = -state->v[i][m];
2027                 /* Store the position at t-dt in buf */
2028                 savex[i][m] = state->x[i][m] + dt*state->v[i][m];
2029             }
2030         }
2031         /* Shake the positions at t=-dt with the positions at t=0
2032          * as reference coordinates.
2033          */
2034         if (fplog)
2035         {
2036             char buf[STEPSTRSIZE];
2037             fprintf(fplog, "\nConstraining the coordinates at t0-dt (step %s)\n",
2038                     gmx_step_str(step, buf));
2039         }
2040         dvdl_dum = 0;
2041         constrain(NULL, TRUE, FALSE, constr, &(top->idef),
2042                   ir, NULL, cr, step, -1, md,
2043                   state->x, savex, NULL,
2044                   fr->bMolPBC, state->box,
2045                   state->lambda[efptBONDED], &dvdl_dum,
2046                   state->v, NULL, nrnb, econqCoord,
2047                   ir->epc == epcMTTK, state->veta, state->veta);
2048
2049         for (i = start; i < end; i++)
2050         {
2051             for (m = 0; m < DIM; m++)
2052             {
2053                 /* Re-reverse the velocities */
2054                 state->v[i][m] = -state->v[i][m];
2055             }
2056         }
2057     }
2058     sfree(savex);
2059 }
2060
2061
2062 static void
2063 integrate_table(real vdwtab[], real scale, int offstart, int rstart, int rend,
2064                 double *enerout, double *virout)
2065 {
2066     double enersum, virsum;
2067     double invscale, invscale2, invscale3;
2068     double r, ea, eb, ec, pa, pb, pc, pd;
2069     double y0, f, g, h;
2070     int    ri, offset, tabfactor;
2071
2072     invscale  = 1.0/scale;
2073     invscale2 = invscale*invscale;
2074     invscale3 = invscale*invscale2;
2075
2076     /* Following summation derived from cubic spline definition,
2077      * Numerical Recipies in C, second edition, p. 113-116.  Exact for
2078      * the cubic spline.  We first calculate the negative of the
2079      * energy from rvdw to rvdw_switch, assuming that g(r)=1, and then
2080      * add the more standard, abrupt cutoff correction to that result,
2081      * yielding the long-range correction for a switched function.  We
2082      * perform both the pressure and energy loops at the same time for
2083      * simplicity, as the computational cost is low. */
2084
2085     if (offstart == 0)
2086     {
2087         /* Since the dispersion table has been scaled down a factor
2088          * 6.0 and the repulsion a factor 12.0 to compensate for the
2089          * c6/c12 parameters inside nbfp[] being scaled up (to save
2090          * flops in kernels), we need to correct for this.
2091          */
2092         tabfactor = 6.0;
2093     }
2094     else
2095     {
2096         tabfactor = 12.0;
2097     }
2098
2099     enersum = 0.0;
2100     virsum  = 0.0;
2101     for (ri = rstart; ri < rend; ++ri)
2102     {
2103         r  = ri*invscale;
2104         ea = invscale3;
2105         eb = 2.0*invscale2*r;
2106         ec = invscale*r*r;
2107
2108         pa = invscale3;
2109         pb = 3.0*invscale2*r;
2110         pc = 3.0*invscale*r*r;
2111         pd = r*r*r;
2112
2113         /* this "8" is from the packing in the vdwtab array - perhaps
2114            should be defined? */
2115
2116         offset = 8*ri + offstart;
2117         y0     = vdwtab[offset];
2118         f      = vdwtab[offset+1];
2119         g      = vdwtab[offset+2];
2120         h      = vdwtab[offset+3];
2121
2122         enersum += y0*(ea/3 + eb/2 + ec) + f*(ea/4 + eb/3 + ec/2) + g*(ea/5 + eb/4 + ec/3) + h*(ea/6 + eb/5 + ec/4);
2123         virsum  +=  f*(pa/4 + pb/3 + pc/2 + pd) + 2*g*(pa/5 + pb/4 + pc/3 + pd/2) + 3*h*(pa/6 + pb/5 + pc/4 + pd/3);
2124     }
2125     *enerout = 4.0*M_PI*enersum*tabfactor;
2126     *virout  = 4.0*M_PI*virsum*tabfactor;
2127 }
2128
2129 void calc_enervirdiff(FILE *fplog, int eDispCorr, t_forcerec *fr)
2130 {
2131     double eners[2], virs[2], enersum, virsum, y0, f, g, h;
2132     double r0, r1, r, rc3, rc9, ea, eb, ec, pa, pb, pc, pd;
2133     double invscale, invscale2, invscale3;
2134     int    ri0, ri1, ri, i, offstart, offset;
2135     real   scale, *vdwtab, tabfactor, tmp;
2136
2137     fr->enershiftsix    = 0;
2138     fr->enershifttwelve = 0;
2139     fr->enerdiffsix     = 0;
2140     fr->enerdifftwelve  = 0;
2141     fr->virdiffsix      = 0;
2142     fr->virdifftwelve   = 0;
2143
2144     if (eDispCorr != edispcNO)
2145     {
2146         for (i = 0; i < 2; i++)
2147         {
2148             eners[i] = 0;
2149             virs[i]  = 0;
2150         }
2151         if ((fr->vdwtype == evdwSWITCH) || (fr->vdwtype == evdwSHIFT))
2152         {
2153             if (fr->rvdw_switch == 0)
2154             {
2155                 gmx_fatal(FARGS,
2156                           "With dispersion correction rvdw-switch can not be zero "
2157                           "for vdw-type = %s", evdw_names[fr->vdwtype]);
2158             }
2159
2160             scale  = fr->nblists[0].table_elec_vdw.scale;
2161             vdwtab = fr->nblists[0].table_vdw.data;
2162
2163             /* Round the cut-offs to exact table values for precision */
2164             ri0  = floor(fr->rvdw_switch*scale);
2165             ri1  = ceil(fr->rvdw*scale);
2166             r0   = ri0/scale;
2167             r1   = ri1/scale;
2168             rc3  = r0*r0*r0;
2169             rc9  = rc3*rc3*rc3;
2170
2171             if (fr->vdwtype == evdwSHIFT)
2172             {
2173                 /* Determine the constant energy shift below rvdw_switch.
2174                  * Table has a scale factor since we have scaled it down to compensate
2175                  * for scaling-up c6/c12 with the derivative factors to save flops in analytical kernels.
2176                  */
2177                 fr->enershiftsix    = (real)(-1.0/(rc3*rc3)) - 6.0*vdwtab[8*ri0];
2178                 fr->enershifttwelve = (real)( 1.0/(rc9*rc3)) - 12.0*vdwtab[8*ri0 + 4];
2179             }
2180             /* Add the constant part from 0 to rvdw_switch.
2181              * This integration from 0 to rvdw_switch overcounts the number
2182              * of interactions by 1, as it also counts the self interaction.
2183              * We will correct for this later.
2184              */
2185             eners[0] += 4.0*M_PI*fr->enershiftsix*rc3/3.0;
2186             eners[1] += 4.0*M_PI*fr->enershifttwelve*rc3/3.0;
2187             for (i = 0; i < 2; i++)
2188             {
2189                 enersum = 0;
2190                 virsum  = 0;
2191                 integrate_table(vdwtab, scale, (i == 0 ? 0 : 4), ri0, ri1, &enersum, &virsum);
2192                 eners[i] -= enersum;
2193                 virs[i]  -= virsum;
2194             }
2195
2196             /* now add the correction for rvdw_switch to infinity */
2197             eners[0] += -4.0*M_PI/(3.0*rc3);
2198             eners[1] +=  4.0*M_PI/(9.0*rc9);
2199             virs[0]  +=  8.0*M_PI/rc3;
2200             virs[1]  += -16.0*M_PI/(3.0*rc9);
2201         }
2202         else if (EVDW_PME(fr->vdwtype))
2203         {
2204             if (EVDW_SWITCHED(fr->vdwtype) && fr->rvdw_switch == 0)
2205             {
2206                 gmx_fatal(FARGS,
2207                           "With dispersion correction rvdw-switch can not be zero "
2208                           "for vdw-type = %s", evdw_names[fr->vdwtype]);
2209             }
2210
2211             scale  = fr->nblists[0].table_vdw.scale;
2212             vdwtab = fr->nblists[0].table_vdw.data;
2213
2214             ri0  = floor(fr->rvdw_switch*scale);
2215             ri1  = ceil(fr->rvdw*scale);
2216             r0   = ri0/scale;
2217             r1   = ri1/scale;
2218             rc3  = r0*r0*r0;
2219             rc9  = rc3*rc3*rc3;
2220
2221             /* Calculate self-interaction coefficient (assuming that
2222              * the reciprocal-space contribution is constant in the
2223              * region that contributes to the self-interaction).
2224              */
2225             fr->enershiftsix = pow(fr->ewaldcoeff_lj, 6) / 6.0;
2226
2227             /* Calculate C12 values as without PME. */
2228             if (EVDW_SWITCHED(fr->vdwtype))
2229             {
2230                 enersum = 0;
2231                 virsum  = 0;
2232                 integrate_table(vdwtab, scale, 4, ri0, ri1, &enersum, &virsum);
2233                 eners[1] -= enersum;
2234                 virs[1]  -= virsum;
2235             }
2236             /* Add analytical corrections, C6 for the whole range, C12
2237              * from rvdw_switch to infinity.
2238              */
2239
2240             eners[0] += -pow(sqrt(M_PI)*fr->ewaldcoeff_lj, 3)/3.0;
2241             eners[1] +=  4.0*M_PI/(9.0*rc9);
2242             virs[0]  +=  pow(sqrt(M_PI)*fr->ewaldcoeff_lj, 3);
2243             virs[1]  += -16.0*M_PI/(3.0*rc9);
2244         }
2245         else if ((fr->vdwtype == evdwCUT) || (fr->vdwtype == evdwUSER))
2246         {
2247             if (fr->vdwtype == evdwUSER && fplog)
2248             {
2249                 fprintf(fplog,
2250                         "WARNING: using dispersion correction with user tables\n");
2251             }
2252             rc3  = fr->rvdw*fr->rvdw*fr->rvdw;
2253             rc9  = rc3*rc3*rc3;
2254             /* Contribution beyond the cut-off */
2255             eners[0] += -4.0*M_PI/(3.0*rc3);
2256             eners[1] +=  4.0*M_PI/(9.0*rc9);
2257             if (fr->vdw_modifier == eintmodPOTSHIFT)
2258             {
2259                 /* Contribution within the cut-off */
2260                 eners[0] += -4.0*M_PI/(3.0*rc3);
2261                 eners[1] +=  4.0*M_PI/(3.0*rc9);
2262             }
2263             /* Contribution beyond the cut-off */
2264             virs[0]  +=  8.0*M_PI/rc3;
2265             virs[1]  += -16.0*M_PI/(3.0*rc9);
2266         }
2267         else
2268         {
2269             gmx_fatal(FARGS,
2270                       "Dispersion correction is not implemented for vdw-type = %s",
2271                       evdw_names[fr->vdwtype]);
2272         }
2273         fr->enerdiffsix    = eners[0];
2274         fr->enerdifftwelve = eners[1];
2275         /* The 0.5 is due to the Gromacs definition of the virial */
2276         fr->virdiffsix     = 0.5*virs[0];
2277         fr->virdifftwelve  = 0.5*virs[1];
2278     }
2279 }
2280
2281 void calc_dispcorr(FILE *fplog, t_inputrec *ir, t_forcerec *fr,
2282                    gmx_large_int_t step, int natoms,
2283                    matrix box, real lambda, tensor pres, tensor virial,
2284                    real *prescorr, real *enercorr, real *dvdlcorr)
2285 {
2286     gmx_bool bCorrAll, bCorrPres;
2287     real     dvdlambda, invvol, dens, ninter, avcsix, avctwelve, enerdiff, svir = 0, spres = 0;
2288     int      m;
2289
2290     *prescorr = 0;
2291     *enercorr = 0;
2292     *dvdlcorr = 0;
2293
2294     clear_mat(virial);
2295     clear_mat(pres);
2296
2297     if (ir->eDispCorr != edispcNO)
2298     {
2299         bCorrAll  = (ir->eDispCorr == edispcAllEner ||
2300                      ir->eDispCorr == edispcAllEnerPres);
2301         bCorrPres = (ir->eDispCorr == edispcEnerPres ||
2302                      ir->eDispCorr == edispcAllEnerPres);
2303
2304         invvol = 1/det(box);
2305         if (fr->n_tpi)
2306         {
2307             /* Only correct for the interactions with the inserted molecule */
2308             dens   = (natoms - fr->n_tpi)*invvol;
2309             ninter = fr->n_tpi;
2310         }
2311         else
2312         {
2313             dens   = natoms*invvol;
2314             ninter = 0.5*natoms;
2315         }
2316
2317         if (ir->efep == efepNO)
2318         {
2319             avcsix    = fr->avcsix[0];
2320             avctwelve = fr->avctwelve[0];
2321         }
2322         else
2323         {
2324             avcsix    = (1 - lambda)*fr->avcsix[0]    + lambda*fr->avcsix[1];
2325             avctwelve = (1 - lambda)*fr->avctwelve[0] + lambda*fr->avctwelve[1];
2326         }
2327
2328         enerdiff   = ninter*(dens*fr->enerdiffsix - fr->enershiftsix);
2329         *enercorr += avcsix*enerdiff;
2330         dvdlambda  = 0.0;
2331         if (ir->efep != efepNO)
2332         {
2333             dvdlambda += (fr->avcsix[1] - fr->avcsix[0])*enerdiff;
2334         }
2335         if (bCorrAll)
2336         {
2337             enerdiff   = ninter*(dens*fr->enerdifftwelve - fr->enershifttwelve);
2338             *enercorr += avctwelve*enerdiff;
2339             if (fr->efep != efepNO)
2340             {
2341                 dvdlambda += (fr->avctwelve[1] - fr->avctwelve[0])*enerdiff;
2342             }
2343         }
2344
2345         if (bCorrPres)
2346         {
2347             svir = ninter*dens*avcsix*fr->virdiffsix/3.0;
2348             if (ir->eDispCorr == edispcAllEnerPres)
2349             {
2350                 svir += ninter*dens*avctwelve*fr->virdifftwelve/3.0;
2351             }
2352             /* The factor 2 is because of the Gromacs virial definition */
2353             spres = -2.0*invvol*svir*PRESFAC;
2354
2355             for (m = 0; m < DIM; m++)
2356             {
2357                 virial[m][m] += svir;
2358                 pres[m][m]   += spres;
2359             }
2360             *prescorr += spres;
2361         }
2362
2363         /* Can't currently control when it prints, for now, just print when degugging */
2364         if (debug)
2365         {
2366             if (bCorrAll)
2367             {
2368                 fprintf(debug, "Long Range LJ corr.: <C6> %10.4e, <C12> %10.4e\n",
2369                         avcsix, avctwelve);
2370             }
2371             if (bCorrPres)
2372             {
2373                 fprintf(debug,
2374                         "Long Range LJ corr.: Epot %10g, Pres: %10g, Vir: %10g\n",
2375                         *enercorr, spres, svir);
2376             }
2377             else
2378             {
2379                 fprintf(debug, "Long Range LJ corr.: Epot %10g\n", *enercorr);
2380             }
2381         }
2382
2383         if (fr->bSepDVDL && do_per_step(step, ir->nstlog))
2384         {
2385             gmx_print_sepdvdl(fplog, "Dispersion correction", *enercorr, dvdlambda);
2386         }
2387         if (fr->efep != efepNO)
2388         {
2389             *dvdlcorr += dvdlambda;
2390         }
2391     }
2392 }
2393
2394 void do_pbc_first(FILE *fplog, matrix box, t_forcerec *fr,
2395                   t_graph *graph, rvec x[])
2396 {
2397     if (fplog)
2398     {
2399         fprintf(fplog, "Removing pbc first time\n");
2400     }
2401     calc_shifts(box, fr->shift_vec);
2402     if (graph)
2403     {
2404         mk_mshift(fplog, graph, fr->ePBC, box, x);
2405         if (gmx_debug_at)
2406         {
2407             p_graph(debug, "do_pbc_first 1", graph);
2408         }
2409         shift_self(graph, box, x);
2410         /* By doing an extra mk_mshift the molecules that are broken
2411          * because they were e.g. imported from another software
2412          * will be made whole again. Such are the healing powers
2413          * of GROMACS.
2414          */
2415         mk_mshift(fplog, graph, fr->ePBC, box, x);
2416         if (gmx_debug_at)
2417         {
2418             p_graph(debug, "do_pbc_first 2", graph);
2419         }
2420     }
2421     if (fplog)
2422     {
2423         fprintf(fplog, "Done rmpbc\n");
2424     }
2425 }
2426
2427 static void low_do_pbc_mtop(FILE *fplog, int ePBC, matrix box,
2428                             gmx_mtop_t *mtop, rvec x[],
2429                             gmx_bool bFirst)
2430 {
2431     t_graph        *graph;
2432     int             mb, as, mol;
2433     gmx_molblock_t *molb;
2434
2435     if (bFirst && fplog)
2436     {
2437         fprintf(fplog, "Removing pbc first time\n");
2438     }
2439
2440     snew(graph, 1);
2441     as = 0;
2442     for (mb = 0; mb < mtop->nmolblock; mb++)
2443     {
2444         molb = &mtop->molblock[mb];
2445         if (molb->natoms_mol == 1 ||
2446             (!bFirst && mtop->moltype[molb->type].cgs.nr == 1))
2447         {
2448             /* Just one atom or charge group in the molecule, no PBC required */
2449             as += molb->nmol*molb->natoms_mol;
2450         }
2451         else
2452         {
2453             /* Pass NULL iso fplog to avoid graph prints for each molecule type */
2454             mk_graph_ilist(NULL, mtop->moltype[molb->type].ilist,
2455                            0, molb->natoms_mol, FALSE, FALSE, graph);
2456
2457             for (mol = 0; mol < molb->nmol; mol++)
2458             {
2459                 mk_mshift(fplog, graph, ePBC, box, x+as);
2460
2461                 shift_self(graph, box, x+as);
2462                 /* The molecule is whole now.
2463                  * We don't need the second mk_mshift call as in do_pbc_first,
2464                  * since we no longer need this graph.
2465                  */
2466
2467                 as += molb->natoms_mol;
2468             }
2469             done_graph(graph);
2470         }
2471     }
2472     sfree(graph);
2473 }
2474
2475 void do_pbc_first_mtop(FILE *fplog, int ePBC, matrix box,
2476                        gmx_mtop_t *mtop, rvec x[])
2477 {
2478     low_do_pbc_mtop(fplog, ePBC, box, mtop, x, TRUE);
2479 }
2480
2481 void do_pbc_mtop(FILE *fplog, int ePBC, matrix box,
2482                  gmx_mtop_t *mtop, rvec x[])
2483 {
2484     low_do_pbc_mtop(fplog, ePBC, box, mtop, x, FALSE);
2485 }
2486
2487 void finish_run(FILE *fplog, t_commrec *cr,
2488                 t_inputrec *inputrec,
2489                 t_nrnb nrnb[], gmx_wallcycle_t wcycle,
2490                 gmx_walltime_accounting_t walltime_accounting,
2491                 wallclock_gpu_t *gputimes,
2492                 gmx_bool bWriteStat)
2493 {
2494     int     i, j;
2495     t_nrnb *nrnb_tot = NULL;
2496     real    delta_t;
2497     double  nbfs, mflop;
2498     double  elapsed_time,
2499             elapsed_time_over_all_ranks,
2500             elapsed_time_over_all_threads,
2501             elapsed_time_over_all_threads_over_all_ranks;
2502     wallcycle_sum(cr, wcycle);
2503
2504     if (cr->nnodes > 1)
2505     {
2506         snew(nrnb_tot, 1);
2507 #ifdef GMX_MPI
2508         MPI_Allreduce(nrnb->n, nrnb_tot->n, eNRNB, MPI_DOUBLE, MPI_SUM,
2509                       cr->mpi_comm_mysim);
2510 #endif
2511     }
2512     else
2513     {
2514         nrnb_tot = nrnb;
2515     }
2516
2517     elapsed_time                                 = walltime_accounting_get_elapsed_time(walltime_accounting);
2518     elapsed_time_over_all_ranks                  = elapsed_time;
2519     elapsed_time_over_all_threads                = walltime_accounting_get_elapsed_time_over_all_threads(walltime_accounting);
2520     elapsed_time_over_all_threads_over_all_ranks = elapsed_time_over_all_threads;
2521 #ifdef GMX_MPI
2522     if (cr->nnodes > 1)
2523     {
2524         /* reduce elapsed_time over all MPI ranks in the current simulation */
2525         MPI_Allreduce(&elapsed_time,
2526                       &elapsed_time_over_all_ranks,
2527                       1, MPI_DOUBLE, MPI_SUM,
2528                       cr->mpi_comm_mysim);
2529         elapsed_time_over_all_ranks /= cr->nnodes;
2530         /* Reduce elapsed_time_over_all_threads over all MPI ranks in the
2531          * current simulation. */
2532         MPI_Allreduce(&elapsed_time_over_all_threads,
2533                       &elapsed_time_over_all_threads_over_all_ranks,
2534                       1, MPI_DOUBLE, MPI_SUM,
2535                       cr->mpi_comm_mysim);
2536     }
2537 #endif
2538
2539     if (SIMMASTER(cr))
2540     {
2541         print_flop(fplog, nrnb_tot, &nbfs, &mflop);
2542     }
2543     if (cr->nnodes > 1)
2544     {
2545         sfree(nrnb_tot);
2546     }
2547
2548     if ((cr->duty & DUTY_PP) && DOMAINDECOMP(cr))
2549     {
2550         print_dd_statistics(cr, inputrec, fplog);
2551     }
2552
2553 #ifdef GMX_MPI
2554     if (PARTDECOMP(cr))
2555     {
2556         if (MASTER(cr))
2557         {
2558             t_nrnb     *nrnb_all;
2559             int         s;
2560             MPI_Status  stat;
2561
2562             snew(nrnb_all, cr->nnodes);
2563             nrnb_all[0] = *nrnb;
2564             for (s = 1; s < cr->nnodes; s++)
2565             {
2566                 MPI_Recv(nrnb_all[s].n, eNRNB, MPI_DOUBLE, s, 0,
2567                          cr->mpi_comm_mysim, &stat);
2568             }
2569             pr_load(fplog, cr, nrnb_all);
2570             sfree(nrnb_all);
2571         }
2572         else
2573         {
2574             MPI_Send(nrnb->n, eNRNB, MPI_DOUBLE, MASTERRANK(cr), 0,
2575                      cr->mpi_comm_mysim);
2576         }
2577     }
2578 #endif
2579
2580     if (SIMMASTER(cr))
2581     {
2582         wallcycle_print(fplog, cr->nnodes, cr->npmenodes,
2583                         elapsed_time_over_all_ranks,
2584                         wcycle, gputimes);
2585
2586         if (EI_DYNAMICS(inputrec->eI))
2587         {
2588             delta_t = inputrec->delta_t;
2589         }
2590         else
2591         {
2592             delta_t = 0;
2593         }
2594
2595         if (fplog)
2596         {
2597             print_perf(fplog, elapsed_time_over_all_threads_over_all_ranks,
2598                        elapsed_time_over_all_ranks,
2599                        walltime_accounting_get_nsteps_done(walltime_accounting),
2600                        delta_t, nbfs, mflop);
2601         }
2602         if (bWriteStat)
2603         {
2604             print_perf(stderr, elapsed_time_over_all_threads_over_all_ranks,
2605                        elapsed_time_over_all_ranks,
2606                        walltime_accounting_get_nsteps_done(walltime_accounting),
2607                        delta_t, nbfs, mflop);
2608         }
2609     }
2610 }
2611
2612 extern void initialize_lambdas(FILE *fplog, t_inputrec *ir, int *fep_state, real *lambda, double *lam0)
2613 {
2614     /* this function works, but could probably use a logic rewrite to keep all the different
2615        types of efep straight. */
2616
2617     int       i;
2618     t_lambda *fep = ir->fepvals;
2619
2620     if ((ir->efep == efepNO) && (ir->bSimTemp == FALSE))
2621     {
2622         for (i = 0; i < efptNR; i++)
2623         {
2624             lambda[i] = 0.0;
2625             if (lam0)
2626             {
2627                 lam0[i] = 0.0;
2628             }
2629         }
2630         return;
2631     }
2632     else
2633     {
2634         *fep_state = fep->init_fep_state; /* this might overwrite the checkpoint
2635                                              if checkpoint is set -- a kludge is in for now
2636                                              to prevent this.*/
2637         for (i = 0; i < efptNR; i++)
2638         {
2639             /* overwrite lambda state with init_lambda for now for backwards compatibility */
2640             if (fep->init_lambda >= 0) /* if it's -1, it was never initializd */
2641             {
2642                 lambda[i] = fep->init_lambda;
2643                 if (lam0)
2644                 {
2645                     lam0[i] = lambda[i];
2646                 }
2647             }
2648             else
2649             {
2650                 lambda[i] = fep->all_lambda[i][*fep_state];
2651                 if (lam0)
2652                 {
2653                     lam0[i] = lambda[i];
2654                 }
2655             }
2656         }
2657         if (ir->bSimTemp)
2658         {
2659             /* need to rescale control temperatures to match current state */
2660             for (i = 0; i < ir->opts.ngtc; i++)
2661             {
2662                 if (ir->opts.ref_t[i] > 0)
2663                 {
2664                     ir->opts.ref_t[i] = ir->simtempvals->temperatures[*fep_state];
2665                 }
2666             }
2667         }
2668     }
2669
2670     /* Send to the log the information on the current lambdas */
2671     if (fplog != NULL)
2672     {
2673         fprintf(fplog, "Initial vector of lambda components:[ ");
2674         for (i = 0; i < efptNR; i++)
2675         {
2676             fprintf(fplog, "%10.4f ", lambda[i]);
2677         }
2678         fprintf(fplog, "]\n");
2679     }
2680     return;
2681 }
2682
2683
2684 void init_md(FILE *fplog,
2685              t_commrec *cr, t_inputrec *ir, const output_env_t oenv,
2686              double *t, double *t0,
2687              real *lambda, int *fep_state, double *lam0,
2688              t_nrnb *nrnb, gmx_mtop_t *mtop,
2689              gmx_update_t *upd,
2690              int nfile, const t_filenm fnm[],
2691              gmx_mdoutf_t **outf, t_mdebin **mdebin,
2692              tensor force_vir, tensor shake_vir, rvec mu_tot,
2693              gmx_bool *bSimAnn, t_vcm **vcm, unsigned long Flags)
2694 {
2695     int  i, j, n;
2696     real tmpt, mod;
2697
2698     /* Initial values */
2699     *t = *t0       = ir->init_t;
2700
2701     *bSimAnn = FALSE;
2702     for (i = 0; i < ir->opts.ngtc; i++)
2703     {
2704         /* set bSimAnn if any group is being annealed */
2705         if (ir->opts.annealing[i] != eannNO)
2706         {
2707             *bSimAnn = TRUE;
2708         }
2709     }
2710     if (*bSimAnn)
2711     {
2712         update_annealing_target_temp(&(ir->opts), ir->init_t);
2713     }
2714
2715     /* Initialize lambda variables */
2716     initialize_lambdas(fplog, ir, fep_state, lambda, lam0);
2717
2718     if (upd)
2719     {
2720         *upd = init_update(ir);
2721     }
2722
2723
2724     if (vcm != NULL)
2725     {
2726         *vcm = init_vcm(fplog, &mtop->groups, ir);
2727     }
2728
2729     if (EI_DYNAMICS(ir->eI) && !(Flags & MD_APPENDFILES))
2730     {
2731         if (ir->etc == etcBERENDSEN)
2732         {
2733             please_cite(fplog, "Berendsen84a");
2734         }
2735         if (ir->etc == etcVRESCALE)
2736         {
2737             please_cite(fplog, "Bussi2007a");
2738         }
2739     }
2740
2741     init_nrnb(nrnb);
2742
2743     if (nfile != -1)
2744     {
2745         *outf = init_mdoutf(nfile, fnm, Flags, cr, ir, oenv);
2746
2747         *mdebin = init_mdebin((Flags & MD_APPENDFILES) ? NULL : (*outf)->fp_ene,
2748                               mtop, ir, (*outf)->fp_dhdl);
2749     }
2750
2751     if (ir->bAdress)
2752     {
2753         please_cite(fplog, "Fritsch12");
2754         please_cite(fplog, "Junghans10");
2755     }
2756     /* Initiate variables */
2757     clear_mat(force_vir);
2758     clear_mat(shake_vir);
2759     clear_rvec(mu_tot);
2760
2761     debug_gmx();
2762 }