src/gromacs/mdlib/ns.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * Copyright (c) 2013,2014,2015,2017,2018, by the GROMACS development team, led by
   7  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   8  * and including many others, as listed in the AUTHORS file in the
   9  * top-level source directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37 #include "gmxpre.h"
  38
  39 #include "ns.h"
  40
  41 #include <stdlib.h>
  42 #include <string.h>
  43
  44 #include <cmath>
  45
  46 #include <algorithm>
  47
  48 #include "gromacs/domdec/domdec.h"
  49 #include "gromacs/domdec/domdec_struct.h"
  50 #include "gromacs/gmxlib/network.h"
  51 #include "gromacs/gmxlib/nrnb.h"
  52 #include "gromacs/gmxlib/nonbonded/nonbonded.h"
  53 #include "gromacs/math/functions.h"
  54 #include "gromacs/math/utilities.h"
  55 #include "gromacs/math/vec.h"
  56 #include "gromacs/math/vecdump.h"
  57 #include "gromacs/mdlib/nsgrid.h"
  58 #include "gromacs/mdlib/qmmm.h"
  59 #include "gromacs/mdtypes/commrec.h"
  60 #include "gromacs/mdtypes/forcerec.h"
  61 #include "gromacs/mdtypes/group.h"
  62 #include "gromacs/mdtypes/inputrec.h"
  63 #include "gromacs/mdtypes/md_enums.h"
  64 #include "gromacs/pbcutil/ishift.h"
  65 #include "gromacs/pbcutil/pbc.h"
  66 #include "gromacs/topology/mtop_util.h"
  67 #include "gromacs/utility/fatalerror.h"
  68 #include "gromacs/utility/smalloc.h"
  69
  70 /*
  71  *    E X C L U S I O N   H A N D L I N G
  72  */
  73
  74 #ifdef DEBUG
  75 static void SETEXCL_(t_excl e[], int i, int j)
  76 {
  77     e[j] = e[j] | (1<<i);
  78 }
  79 static void RMEXCL_(t_excl e[], int i, int j)
  80 {
  81     e[j] = e[j] & ~(1<<i);
  82 }
  83 static gmx_bool ISEXCL_(t_excl e[], int i, int j)
  84 {
  85     return (gmx_bool)(e[j] & (1<<i));
  86 }
  87 static gmx_bool NOTEXCL_(t_excl e[], int i, int j)
  88 {
  89     return !(ISEXCL(e, i, j));
  90 }
  91 #else
  92 #define SETEXCL(e, i, j) (e)[((int) (j))] |= (1<<((int) (i)))
  93 #define RMEXCL(e, i, j)  (e)[((int) (j))] &= (~(1<<((int) (i))))
  94 #define ISEXCL(e, i, j)  (gmx_bool) ((e)[((int) (j))] & (1<<((int) (i))))
  95 #define NOTEXCL(e, i, j) !(ISEXCL(e, i, j))
  96 #endif
  97
  98 static int
  99 round_up_to_simd_width(int length, int simd_width)
 100 {
 101     int offset;
 102
 103     offset = (simd_width > 0) ? length % simd_width : 0;
 104
 105     return (offset == 0) ? length : length-offset+simd_width;
 106 }
 107 /************************************************
 108  *
 109  *  U T I L I T I E S    F O R    N S
 110  *
 111  ************************************************/
 112
 113 void reallocate_nblist(t_nblist *nl)
 114 {
 115     if (gmx_debug_at)
 116     {
 117         fprintf(debug, "reallocating neigborlist (ielec=%d, ivdw=%d, igeometry=%d, type=%d), maxnri=%d\n",
 118                 nl->ielec, nl->ivdw, nl->igeometry, nl->type, nl->maxnri);
 119     }
 120     srenew(nl->iinr,   nl->maxnri);
 121     if (nl->igeometry == GMX_NBLIST_GEOMETRY_CG_CG)
 122     {
 123         srenew(nl->iinr_end, nl->maxnri);
 124     }
 125     srenew(nl->gid,    nl->maxnri);
 126     srenew(nl->shift,  nl->maxnri);
 127     srenew(nl->jindex, nl->maxnri+1);
 128 }
 129
 130
 131 static void init_nblist(FILE *log, t_nblist *nl_sr,
 132                         int maxsr,
 133                         int ivdw, int ivdwmod,
 134                         int ielec, int ielecmod,
 135                         int igeometry, int type,
 136                         gmx_bool bElecAndVdwSwitchDiffers)
 137 {
 138     t_nblist *nl;
 139     int       homenr;
 140
 141     {
 142         nl     = nl_sr;
 143         homenr = maxsr;
 144
 145         if (nl == nullptr)
 146         {
 147             return;
 148         }
 149
 150
 151         /* Set coul/vdw in neighborlist, and for the normal loops we determine
 152          * an index of which one to call.
 153          */
 154         nl->ivdw        = ivdw;
 155         nl->ivdwmod     = ivdwmod;
 156         nl->ielec       = ielec;
 157         nl->ielecmod    = ielecmod;
 158         nl->type        = type;
 159         nl->igeometry   = igeometry;
 160
 161         if (nl->type == GMX_NBLIST_INTERACTION_FREE_ENERGY)
 162         {
 163             nl->igeometry  = GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE;
 164         }
 165
 166         /* This will also set the simd_padding_width field */
 167         gmx_nonbonded_set_kernel_pointers(log, nl, bElecAndVdwSwitchDiffers);
 168
 169         /* maxnri is influenced by the number of shifts (maximum is 8)
 170          * and the number of energy groups.
 171          * If it is not enough, nl memory will be reallocated during the run.
 172          * 4 seems to be a reasonable factor, which only causes reallocation
 173          * during runs with tiny and many energygroups.
 174          */
 175         nl->maxnri      = homenr*4;
 176         nl->maxnrj      = 0;
 177         nl->nri         = -1;
 178         nl->nrj         = 0;
 179         nl->iinr        = nullptr;
 180         nl->gid         = nullptr;
 181         nl->shift       = nullptr;
 182         nl->jindex      = nullptr;
 183         nl->jjnr        = nullptr;
 184         nl->excl_fep    = nullptr;
 185         reallocate_nblist(nl);
 186         nl->jindex[0] = 0;
 187
 188         if (debug)
 189         {
 190             fprintf(debug, "Initiating neighbourlist (ielec=%d, ivdw=%d, type=%d) for %s interactions,\nwith %d SR atoms.\n",
 191                     nl->ielec, nl->ivdw, nl->type, gmx_nblist_geometry_names[nl->igeometry], maxsr);
 192         }
 193     }
 194 }
 195
 196 void init_neighbor_list(FILE *log, t_forcerec *fr, int homenr)
 197 {
 198     int        maxsr, maxsr_wat;
 199     int        ielec, ivdw, ielecmod, ivdwmod, type;
 200     int        igeometry_def, igeometry_w, igeometry_ww;
 201     int        i;
 202     gmx_bool   bElecAndVdwSwitchDiffers;
 203     t_nblists *nbl;
 204
 205     /* maxsr     = homenr-fr->nWatMol*3; */
 206     maxsr     = homenr;
 207
 208     if (maxsr < 0)
 209     {
 210         gmx_fatal(FARGS, "%s, %d: Negative number of short range atoms.\n"
 211                   "Call your GROMACS dealer for assistance.", __FILE__, __LINE__);
 212     }
 213     /* This is just for initial allocation, so we do not reallocate
 214      * all the nlist arrays many times in a row.
 215      * The numbers seem very accurate, but they are uncritical.
 216      */
 217     maxsr_wat = std::min(fr->nWatMol, (homenr+2)/3);
 218
 219     /* Determine the values for ielec/ivdw. */
 220     ielec                    = fr->nbkernel_elec_interaction;
 221     ivdw                     = fr->nbkernel_vdw_interaction;
 222     ielecmod                 = fr->nbkernel_elec_modifier;
 223     ivdwmod                  = fr->nbkernel_vdw_modifier;
 224     type                     = GMX_NBLIST_INTERACTION_STANDARD;
 225     bElecAndVdwSwitchDiffers = ( (fr->ic->rcoulomb_switch != fr->ic->rvdw_switch) || (fr->ic->rcoulomb != fr->ic->rvdw));
 226
 227     fr->ns->bCGlist = (getenv("GMX_NBLISTCG") != nullptr);
 228     if (!fr->ns->bCGlist)
 229     {
 230         igeometry_def = GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE;
 231     }
 232     else
 233     {
 234         igeometry_def = GMX_NBLIST_GEOMETRY_CG_CG;
 235         if (log != nullptr)
 236         {
 237             fprintf(log, "\nUsing charge-group - charge-group neighbor lists and kernels\n\n");
 238         }
 239     }
 240
 241     if (fr->solvent_opt == esolTIP4P)
 242     {
 243         igeometry_w  = GMX_NBLIST_GEOMETRY_WATER4_PARTICLE;
 244         igeometry_ww = GMX_NBLIST_GEOMETRY_WATER4_WATER4;
 245     }
 246     else
 247     {
 248         igeometry_w  = GMX_NBLIST_GEOMETRY_WATER3_PARTICLE;
 249         igeometry_ww = GMX_NBLIST_GEOMETRY_WATER3_WATER3;
 250     }
 251
 252     for (i = 0; i < fr->nnblists; i++)
 253     {
 254         nbl = &(fr->nblists[i]);
 255
 256         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ],
 257                     maxsr, ivdw, ivdwmod, ielec, ielecmod, igeometry_def, type, bElecAndVdwSwitchDiffers);
 258         init_nblist(log, &nbl->nlist_sr[eNL_VDW],
 259                     maxsr, ivdw, ivdwmod, GMX_NBKERNEL_ELEC_NONE, eintmodNONE, igeometry_def, type, bElecAndVdwSwitchDiffers);
 260         init_nblist(log, &nbl->nlist_sr[eNL_QQ],
 261                     maxsr, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_def, type, bElecAndVdwSwitchDiffers);
 262         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_WATER],
 263                     maxsr_wat, ivdw, ivdwmod, ielec, ielecmod, igeometry_w, type, bElecAndVdwSwitchDiffers);
 264         init_nblist(log, &nbl->nlist_sr[eNL_QQ_WATER],
 265                     maxsr_wat, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_w, type, bElecAndVdwSwitchDiffers);
 266         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_WATERWATER],
 267                     maxsr_wat, ivdw, ivdwmod, ielec, ielecmod, igeometry_ww, type, bElecAndVdwSwitchDiffers);
 268         init_nblist(log, &nbl->nlist_sr[eNL_QQ_WATERWATER],
 269                     maxsr_wat, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_ww, type, bElecAndVdwSwitchDiffers);
 270
 271         /* Did we get the solvent loops so we can use optimized water kernels? */
 272         if (nbl->nlist_sr[eNL_VDWQQ_WATER].kernelptr_vf == nullptr
 273             || nbl->nlist_sr[eNL_QQ_WATER].kernelptr_vf == nullptr
 274             || nbl->nlist_sr[eNL_VDWQQ_WATERWATER].kernelptr_vf == nullptr
 275             || nbl->nlist_sr[eNL_QQ_WATERWATER].kernelptr_vf == nullptr)
 276         {
 277             fr->solvent_opt = esolNO;
 278             if (log != nullptr)
 279             {
 280                 fprintf(log, "Note: The available nonbonded kernels do not support water optimization - disabling.\n");
 281             }
 282         }
 283
 284         if (fr->efep != efepNO)
 285         {
 286             init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_FREE],
 287                         maxsr, ivdw, ivdwmod, ielec, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY, bElecAndVdwSwitchDiffers);
 288             init_nblist(log, &nbl->nlist_sr[eNL_VDW_FREE],
 289                         maxsr, ivdw, ivdwmod, GMX_NBKERNEL_ELEC_NONE, eintmodNONE, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY, bElecAndVdwSwitchDiffers);
 290             init_nblist(log, &nbl->nlist_sr[eNL_QQ_FREE],
 291                         maxsr, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY, bElecAndVdwSwitchDiffers);
 292         }
 293     }
 294     /* QMMM MM list */
 295     if (fr->bQMMM && fr->qr->QMMMscheme != eQMMMschemeoniom)
 296     {
 297         if (nullptr == fr->QMMMlist)
 298         {
 299             snew(fr->QMMMlist, 1);
 300         }
 301         init_nblist(log, fr->QMMMlist,
 302                     maxsr, 0, 0, ielec, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_STANDARD, bElecAndVdwSwitchDiffers);
 303     }
 304
 305     if (log != nullptr)
 306     {
 307         fprintf(log, "\n");
 308     }
 309
 310     fr->ns->nblist_initialized = TRUE;
 311 }
 312
 313 static void reset_nblist(t_nblist *nl)
 314 {
 315     nl->nri       = -1;
 316     nl->nrj       = 0;
 317     if (nl->jindex)
 318     {
 319         nl->jindex[0] = 0;
 320     }
 321 }
 322
 323 static void reset_neighbor_lists(t_forcerec *fr)
 324 {
 325     int n, i;
 326
 327     if (fr->bQMMM)
 328     {
 329         /* only reset the short-range nblist */
 330         reset_nblist(fr->QMMMlist);
 331     }
 332
 333     for (n = 0; n < fr->nnblists; n++)
 334     {
 335         for (i = 0; i < eNL_NR; i++)
 336         {
 337             reset_nblist( &(fr->nblists[n].nlist_sr[i]) );
 338         }
 339     }
 340 }
 341
 342
 343
 344
 345 static inline void new_i_nblist(t_nblist *nlist, int i_atom, int shift, int gid)
 346 {
 347     int    nri = nlist->nri;
 348
 349     /* Check whether we have to increase the i counter */
 350     if ((nri == -1) ||
 351         (nlist->iinr[nri]  != i_atom) ||
 352         (nlist->shift[nri] != shift) ||
 353         (nlist->gid[nri]   != gid))
 354     {
 355         /* This is something else. Now see if any entries have
 356          * been added in the list of the previous atom.
 357          */
 358         if ((nri == -1) ||
 359             ((nlist->jindex[nri+1] > nlist->jindex[nri]) &&
 360              (nlist->gid[nri] != -1)))
 361         {
 362             /* If so increase the counter */
 363             nlist->nri++;
 364             nri++;
 365             if (nlist->nri >= nlist->maxnri)
 366             {
 367                 nlist->maxnri += over_alloc_large(nlist->nri);
 368                 reallocate_nblist(nlist);
 369             }
 370         }
 371         /* Set the number of neighbours and the atom number */
 372         nlist->jindex[nri+1] = nlist->jindex[nri];
 373         nlist->iinr[nri]     = i_atom;
 374         nlist->gid[nri]      = gid;
 375         nlist->shift[nri]    = shift;
 376     }
 377     else
 378     {
 379         /* Adding to previous list. First remove possible previous padding */
 380         if (nlist->simd_padding_width > 1)
 381         {
 382             while (nlist->nrj > 0 && nlist->jjnr[nlist->nrj-1] < 0)
 383             {
 384                 nlist->nrj--;
 385             }
 386         }
 387     }
 388 }
 389
 390 static inline void close_i_nblist(t_nblist *nlist)
 391 {
 392     int nri = nlist->nri;
 393     int len;
 394
 395     if (nri >= 0)
 396     {
 397         /* Add elements up to padding. Since we allocate memory in units
 398          * of the simd_padding width, we do not have to check for possible
 399          * list reallocation here.
 400          */
 401         while ((nlist->nrj % nlist->simd_padding_width) != 0)
 402         {
 403             /* Use -4 here, so we can write forces for 4 atoms before real data */
 404             nlist->jjnr[nlist->nrj++] = -4;
 405         }
 406         nlist->jindex[nri+1] = nlist->nrj;
 407
 408         len = nlist->nrj -  nlist->jindex[nri];
 409         /* If there are no j-particles we have to reduce the
 410          * number of i-particles again, to prevent errors in the
 411          * kernel functions.
 412          */
 413         if ((len == 0) && (nlist->nri > 0))
 414         {
 415             nlist->nri--;
 416         }
 417     }
 418 }
 419
 420 static inline void close_nblist(t_nblist *nlist)
 421 {
 422     /* Only close this nblist when it has been initialized.
 423      * Avoid the creation of i-lists with no j-particles.
 424      */
 425     if (nlist->nrj == 0)
 426     {
 427         /* Some assembly kernels do not support empty lists,
 428          * make sure here that we don't generate any empty lists.
 429          * With the current ns code this branch is taken in two cases:
 430          * No i-particles at all: nri=-1 here
 431          * There are i-particles, but no j-particles; nri=0 here
 432          */
 433         nlist->nri = 0;
 434     }
 435     else
 436     {
 437         /* Close list number nri by incrementing the count */
 438         nlist->nri++;
 439     }
 440 }
 441
 442 static inline void close_neighbor_lists(t_forcerec *fr, gmx_bool bMakeQMMMnblist)
 443 {
 444     int n, i;
 445
 446     if (bMakeQMMMnblist)
 447     {
 448         close_nblist(fr->QMMMlist);
 449     }
 450
 451     for (n = 0; n < fr->nnblists; n++)
 452     {
 453         for (i = 0; (i < eNL_NR); i++)
 454         {
 455             close_nblist(&(fr->nblists[n].nlist_sr[i]));
 456         }
 457     }
 458 }
 459
 460
 461 static inline void add_j_to_nblist(t_nblist *nlist, int j_atom)
 462 {
 463     int nrj = nlist->nrj;
 464
 465     if (nlist->nrj >= nlist->maxnrj)
 466     {
 467         nlist->maxnrj = round_up_to_simd_width(over_alloc_small(nlist->nrj + 1), nlist->simd_padding_width);
 468
 469         if (gmx_debug_at)
 470         {
 471             fprintf(debug, "Increasing SR nblist (ielec=%d,ivdw=%d,type=%d,igeometry=%d) j size to %d\n",
 472                     nlist->ielec, nlist->ivdw, nlist->type, nlist->igeometry, nlist->maxnrj);
 473         }
 474
 475         srenew(nlist->jjnr, nlist->maxnrj);
 476     }
 477
 478     nlist->jjnr[nrj] = j_atom;
 479     nlist->nrj++;
 480 }
 481
 482 static inline void add_j_to_nblist_cg(t_nblist *nlist,
 483                                       int j_start, int j_end,
 484                                       const t_excl *bexcl, gmx_bool i_is_j)
 485 {
 486     int nrj = nlist->nrj;
 487     int j;
 488
 489     if (nlist->nrj >= nlist->maxnrj)
 490     {
 491         nlist->maxnrj = over_alloc_small(nlist->nrj + 1);
 492         if (gmx_debug_at)
 493         {
 494             fprintf(debug, "Increasing SR nblist (ielec=%d,ivdw=%d,type=%d,igeometry=%d) j size to %d\n",
 495                     nlist->ielec, nlist->ivdw, nlist->type, nlist->igeometry, nlist->maxnrj);
 496         }
 497
 498         srenew(nlist->jjnr, nlist->maxnrj);
 499         srenew(nlist->jjnr_end, nlist->maxnrj);
 500         srenew(nlist->excl, nlist->maxnrj*MAX_CGCGSIZE);
 501     }
 502
 503     nlist->jjnr[nrj]     = j_start;
 504     nlist->jjnr_end[nrj] = j_end;
 505
 506     if (j_end - j_start > MAX_CGCGSIZE)
 507     {
 508         gmx_fatal(FARGS, "The charge-group - charge-group neighborlist do not support charge groups larger than %d, found a charge group of size %d", MAX_CGCGSIZE, j_end-j_start);
 509     }
 510
 511     /* Set the exclusions */
 512     for (j = j_start; j < j_end; j++)
 513     {
 514         nlist->excl[nrj*MAX_CGCGSIZE + j - j_start] = bexcl[j];
 515     }
 516     if (i_is_j)
 517     {
 518         /* Avoid double counting of intra-cg interactions */
 519         for (j = 1; j < j_end-j_start; j++)
 520         {
 521             nlist->excl[nrj*MAX_CGCGSIZE + j] |= (1<<j) - 1;
 522         }
 523     }
 524
 525     nlist->nrj++;
 526 }
 527
 528 typedef void
 529     put_in_list_t (const gmx_bool        bHaveVdW[],
 530                    int                   ngid,
 531                    const t_mdatoms      *md,
 532                    int                   icg,
 533                    int                   jgid,
 534                    int                   nj,
 535                    const int             jjcg[],
 536                    const int             index[],
 537                    const t_excl          bExcl[],
 538                    int                   shift,
 539                    t_forcerec     *      fr,
 540                    gmx_bool              bDoVdW,
 541                    gmx_bool              bDoCoul,
 542                    int                   solvent_opt);
 543
 544 static void
 545 put_in_list_at(const gmx_bool              bHaveVdW[],
 546                int                         ngid,
 547                const t_mdatoms            *md,
 548                int                         icg,
 549                int                         jgid,
 550                int                         nj,
 551                const int                   jjcg[],
 552                const int                   index[],
 553                const t_excl                bExcl[],
 554                int                         shift,
 555                t_forcerec           *      fr,
 556                gmx_bool                    bDoVdW,
 557                gmx_bool                    bDoCoul,
 558                int                         solvent_opt)
 559 {
 560     /* The a[] index has been removed,
 561      * to put it back in i_atom should be a[i0] and jj should be a[jj].
 562      */
 563     t_nblist  *   vdwc;
 564     t_nblist  *   vdw;
 565     t_nblist  *   coul;
 566     t_nblist  *   vdwc_free  = nullptr;
 567     t_nblist  *   vdw_free   = nullptr;
 568     t_nblist  *   coul_free  = nullptr;
 569     t_nblist  *   vdwc_ww    = nullptr;
 570     t_nblist  *   coul_ww    = nullptr;
 571
 572     int           i, j, jcg, igid, gid, nbl_ind;
 573     int           jj, jj0, jj1, i_atom;
 574     int           i0, nicg;
 575
 576     int          *cginfo;
 577     int          *type, *typeB;
 578     real         *charge, *chargeB;
 579     real          qi, qiB;
 580     gmx_bool      bFreeEnergy, bFree, bFreeJ, bNotEx, *bPert;
 581     gmx_bool      bDoVdW_i, bDoCoul_i, bDoCoul_i_sol;
 582     int           iwater, jwater;
 583     t_nblist     *nlist;
 584
 585     /* Copy some pointers */
 586     cginfo  = fr->cginfo;
 587     charge  = md->chargeA;
 588     chargeB = md->chargeB;
 589     type    = md->typeA;
 590     typeB   = md->typeB;
 591     bPert   = md->bPerturbed;
 592
 593     /* Get atom range */
 594     i0     = index[icg];
 595     nicg   = index[icg+1]-i0;
 596
 597     /* Get the i charge group info */
 598     igid   = GET_CGINFO_GID(cginfo[icg]);
 599
 600     iwater = (solvent_opt != esolNO) ? GET_CGINFO_SOLOPT(cginfo[icg]) : esolNO;
 601
 602     bFreeEnergy = FALSE;
 603     if (md->nPerturbed)
 604     {
 605         /* Check if any of the particles involved are perturbed.
 606          * If not we can do the cheaper normal put_in_list
 607          * and use more solvent optimization.
 608          */
 609         for (i = 0; i < nicg; i++)
 610         {
 611             bFreeEnergy |= bPert[i0+i];
 612         }
 613         /* Loop over the j charge groups */
 614         for (j = 0; (j < nj && !bFreeEnergy); j++)
 615         {
 616             jcg = jjcg[j];
 617             jj0 = index[jcg];
 618             jj1 = index[jcg+1];
 619             /* Finally loop over the atoms in the j-charge group */
 620             for (jj = jj0; jj < jj1; jj++)
 621             {
 622                 bFreeEnergy |= bPert[jj];
 623             }
 624         }
 625     }
 626
 627     /* Unpack pointers to neighbourlist structs */
 628     if (fr->nnblists == 1)
 629     {
 630         nbl_ind = 0;
 631     }
 632     else
 633     {
 634         nbl_ind = fr->gid2nblists[GID(igid, jgid, ngid)];
 635     }
 636     nlist = fr->nblists[nbl_ind].nlist_sr;
 637
 638     if (iwater != esolNO)
 639     {
 640         vdwc    = &nlist[eNL_VDWQQ_WATER];
 641         vdw     = &nlist[eNL_VDW];
 642         coul    = &nlist[eNL_QQ_WATER];
 643         vdwc_ww = &nlist[eNL_VDWQQ_WATERWATER];
 644         coul_ww = &nlist[eNL_QQ_WATERWATER];
 645     }
 646     else
 647     {
 648         vdwc = &nlist[eNL_VDWQQ];
 649         vdw  = &nlist[eNL_VDW];
 650         coul = &nlist[eNL_QQ];
 651     }
 652
 653     if (!bFreeEnergy)
 654     {
 655         if (iwater != esolNO)
 656         {
 657             /* Loop over the atoms in the i charge group */
 658             i_atom  = i0;
 659             gid     = GID(igid, jgid, ngid);
 660             /* Create new i_atom for each energy group */
 661             if (bDoCoul && bDoVdW)
 662             {
 663                 new_i_nblist(vdwc, i_atom, shift, gid);
 664                 new_i_nblist(vdwc_ww, i_atom, shift, gid);
 665             }
 666             if (bDoVdW)
 667             {
 668                 new_i_nblist(vdw, i_atom, shift, gid);
 669             }
 670             if (bDoCoul)
 671             {
 672                 new_i_nblist(coul, i_atom, shift, gid);
 673                 new_i_nblist(coul_ww, i_atom, shift, gid);
 674             }
 675             /* Loop over the j charge groups */
 676             for (j = 0; (j < nj); j++)
 677             {
 678                 jcg = jjcg[j];
 679
 680                 if (jcg == icg)
 681                 {
 682                     continue;
 683                 }
 684
 685                 jj0    = index[jcg];
 686                 jwater = GET_CGINFO_SOLOPT(cginfo[jcg]);
 687
 688                 if (iwater == esolSPC && jwater == esolSPC)
 689                 {
 690                     /* Interaction between two SPC molecules */
 691                     if (!bDoCoul)
 692                     {
 693                         /* VdW only - only first atoms in each water interact */
 694                         add_j_to_nblist(vdw, jj0);
 695                     }
 696                     else
 697                     {
 698                         /* One entry for the entire water-water interaction */
 699                         if (!bDoVdW)
 700                         {
 701                             add_j_to_nblist(coul_ww, jj0);
 702                         }
 703                         else
 704                         {
 705                             add_j_to_nblist(vdwc_ww, jj0);
 706                         }
 707                     }
 708                 }
 709                 else if (iwater == esolTIP4P && jwater == esolTIP4P)
 710                 {
 711                     /* Interaction between two TIP4p molecules */
 712                     if (!bDoCoul)
 713                     {
 714                         /* VdW only - only first atoms in each water interact */
 715                         add_j_to_nblist(vdw, jj0);
 716                     }
 717                     else
 718                     {
 719                         /* One entry for the entire water-water interaction */
 720                         if (!bDoVdW)
 721                         {
 722                             add_j_to_nblist(coul_ww, jj0);
 723                         }
 724                         else
 725                         {
 726                             add_j_to_nblist(vdwc_ww, jj0);
 727                         }
 728                     }
 729                 }
 730                 else
 731                 {
 732                     /* j charge group is not water, but i is.
 733                      * Add entries to the water-other_atom lists; the geometry of the water
 734                      * molecule doesn't matter - that is taken care of in the nonbonded kernel,
 735                      * so we don't care if it is SPC or TIP4P...
 736                      */
 737
 738                     jj1 = index[jcg+1];
 739
 740                     if (!bDoVdW)
 741                     {
 742                         for (jj = jj0; (jj < jj1); jj++)
 743                         {
 744                             if (charge[jj] != 0)
 745                             {
 746                                 add_j_to_nblist(coul, jj);
 747                             }
 748                         }
 749                     }
 750                     else if (!bDoCoul)
 751                     {
 752                         for (jj = jj0; (jj < jj1); jj++)
 753                         {
 754                             if (bHaveVdW[type[jj]])
 755                             {
 756                                 add_j_to_nblist(vdw, jj);
 757                             }
 758                         }
 759                     }
 760                     else
 761                     {
 762                         /* _charge_ _groups_ interact with both coulomb and LJ */
 763                         /* Check which atoms we should add to the lists!       */
 764                         for (jj = jj0; (jj < jj1); jj++)
 765                         {
 766                             if (bHaveVdW[type[jj]])
 767                             {
 768                                 if (charge[jj] != 0)
 769                                 {
 770                                     add_j_to_nblist(vdwc, jj);
 771                                 }
 772                                 else
 773                                 {
 774                                     add_j_to_nblist(vdw, jj);
 775                                 }
 776                             }
 777                             else if (charge[jj] != 0)
 778                             {
 779                                 add_j_to_nblist(coul, jj);
 780                             }
 781                         }
 782                     }
 783                 }
 784             }
 785             close_i_nblist(vdw);
 786             close_i_nblist(coul);
 787             close_i_nblist(vdwc);
 788             close_i_nblist(coul_ww);
 789             close_i_nblist(vdwc_ww);
 790         }
 791         else
 792         {
 793             /* no solvent as i charge group */
 794             /* Loop over the atoms in the i charge group */
 795             for (i = 0; i < nicg; i++)
 796             {
 797                 i_atom  = i0+i;
 798                 gid     = GID(igid, jgid, ngid);
 799                 qi      = charge[i_atom];
 800
 801                 /* Create new i_atom for each energy group */
 802                 if (bDoVdW && bDoCoul)
 803                 {
 804                     new_i_nblist(vdwc, i_atom, shift, gid);
 805                 }
 806                 if (bDoVdW)
 807                 {
 808                     new_i_nblist(vdw, i_atom, shift, gid);
 809                 }
 810                 if (bDoCoul)
 811                 {
 812                     new_i_nblist(coul, i_atom, shift, gid);
 813                 }
 814                 bDoVdW_i  = (bDoVdW  && bHaveVdW[type[i_atom]]);
 815                 bDoCoul_i = (bDoCoul && qi != 0);
 816
 817                 if (bDoVdW_i || bDoCoul_i)
 818                 {
 819                     /* Loop over the j charge groups */
 820                     for (j = 0; (j < nj); j++)
 821                     {
 822                         jcg = jjcg[j];
 823
 824                         /* Check for large charge groups */
 825                         if (jcg == icg)
 826                         {
 827                             jj0 = i0 + i + 1;
 828                         }
 829                         else
 830                         {
 831                             jj0 = index[jcg];
 832                         }
 833
 834                         jj1 = index[jcg+1];
 835                         /* Finally loop over the atoms in the j-charge group */
 836                         for (jj = jj0; jj < jj1; jj++)
 837                         {
 838                             bNotEx = NOTEXCL(bExcl, i, jj);
 839
 840                             if (bNotEx)
 841                             {
 842                                 if (!bDoVdW_i)
 843                                 {
 844                                     if (charge[jj] != 0)
 845                                     {
 846                                         add_j_to_nblist(coul, jj);
 847                                     }
 848                                 }
 849                                 else if (!bDoCoul_i)
 850                                 {
 851                                     if (bHaveVdW[type[jj]])
 852                                     {
 853                                         add_j_to_nblist(vdw, jj);
 854                                     }
 855                                 }
 856                                 else
 857                                 {
 858                                     if (bHaveVdW[type[jj]])
 859                                     {
 860                                         if (charge[jj] != 0)
 861                                         {
 862                                             add_j_to_nblist(vdwc, jj);
 863                                         }
 864                                         else
 865                                         {
 866                                             add_j_to_nblist(vdw, jj);
 867                                         }
 868                                     }
 869                                     else if (charge[jj] != 0)
 870                                     {
 871                                         add_j_to_nblist(coul, jj);
 872                                     }
 873                                 }
 874                             }
 875                         }
 876                     }
 877                 }
 878                 close_i_nblist(vdw);
 879                 close_i_nblist(coul);
 880                 close_i_nblist(vdwc);
 881             }
 882         }
 883     }
 884     else
 885     {
 886         /* we are doing free energy */
 887         vdwc_free = &nlist[eNL_VDWQQ_FREE];
 888         vdw_free  = &nlist[eNL_VDW_FREE];
 889         coul_free = &nlist[eNL_QQ_FREE];
 890         /* Loop over the atoms in the i charge group */
 891         for (i = 0; i < nicg; i++)
 892         {
 893             i_atom  = i0+i;
 894             gid     = GID(igid, jgid, ngid);
 895             qi      = charge[i_atom];
 896             qiB     = chargeB[i_atom];
 897
 898             /* Create new i_atom for each energy group */
 899             if (bDoVdW && bDoCoul)
 900             {
 901                 new_i_nblist(vdwc, i_atom, shift, gid);
 902             }
 903             if (bDoVdW)
 904             {
 905                 new_i_nblist(vdw, i_atom, shift, gid);
 906             }
 907             if (bDoCoul)
 908             {
 909                 new_i_nblist(coul, i_atom, shift, gid);
 910             }
 911
 912             new_i_nblist(vdw_free, i_atom, shift, gid);
 913             new_i_nblist(coul_free, i_atom, shift, gid);
 914             new_i_nblist(vdwc_free, i_atom, shift, gid);
 915
 916             bDoVdW_i  = (bDoVdW  &&
 917                          (bHaveVdW[type[i_atom]] || bHaveVdW[typeB[i_atom]]));
 918             bDoCoul_i = (bDoCoul && (qi != 0 || qiB != 0));
 919             /* For TIP4P the first atom does not have a charge,
 920              * but the last three do. So we should still put an atom
 921              * without LJ but with charge in the water-atom neighborlist
 922              * for a TIP4p i charge group.
 923              * For SPC type water the first atom has LJ and charge,
 924              * so there is no such problem.
 925              */
 926             if (iwater == esolNO)
 927             {
 928                 bDoCoul_i_sol = bDoCoul_i;
 929             }
 930             else
 931             {
 932                 bDoCoul_i_sol = bDoCoul;
 933             }
 934
 935             if (bDoVdW_i || bDoCoul_i_sol)
 936             {
 937                 /* Loop over the j charge groups */
 938                 for (j = 0; (j < nj); j++)
 939                 {
 940                     jcg = jjcg[j];
 941
 942                     /* Check for large charge groups */
 943                     if (jcg == icg)
 944                     {
 945                         jj0 = i0 + i + 1;
 946                     }
 947                     else
 948                     {
 949                         jj0 = index[jcg];
 950                     }
 951
 952                     jj1 = index[jcg+1];
 953                     /* Finally loop over the atoms in the j-charge group */
 954                     bFree = bPert[i_atom];
 955                     for (jj = jj0; (jj < jj1); jj++)
 956                     {
 957                         bFreeJ = bFree || bPert[jj];
 958                         /* Complicated if, because the water H's should also
 959                          * see perturbed j-particles
 960                          */
 961                         if (iwater == esolNO || i == 0 || bFreeJ)
 962                         {
 963                             bNotEx = NOTEXCL(bExcl, i, jj);
 964
 965                             if (bNotEx)
 966                             {
 967                                 if (bFreeJ)
 968                                 {
 969                                     if (!bDoVdW_i)
 970                                     {
 971                                         if (charge[jj] != 0 || chargeB[jj] != 0)
 972                                         {
 973                                             add_j_to_nblist(coul_free, jj);
 974                                         }
 975                                     }
 976                                     else if (!bDoCoul_i)
 977                                     {
 978                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
 979                                         {
 980                                             add_j_to_nblist(vdw_free, jj);
 981                                         }
 982                                     }
 983                                     else
 984                                     {
 985                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
 986                                         {
 987                                             if (charge[jj] != 0 || chargeB[jj] != 0)
 988                                             {
 989                                                 add_j_to_nblist(vdwc_free, jj);
 990                                             }
 991                                             else
 992                                             {
 993                                                 add_j_to_nblist(vdw_free, jj);
 994                                             }
 995                                         }
 996                                         else if (charge[jj] != 0 || chargeB[jj] != 0)
 997                                         {
 998                                             add_j_to_nblist(coul_free, jj);
 999                                         }
1000                                     }
1001                                 }
1002                                 else if (!bDoVdW_i)
1003                                 {
1004                                     /* This is done whether or not bWater is set */
1005                                     if (charge[jj] != 0)
1006                                     {
1007                                         add_j_to_nblist(coul, jj);
1008                                     }
1009                                 }
1010                                 else if (!bDoCoul_i_sol)
1011                                 {
1012                                     if (bHaveVdW[type[jj]])
1013                                     {
1014                                         add_j_to_nblist(vdw, jj);
1015                                     }
1016                                 }
1017                                 else
1018                                 {
1019                                     if (bHaveVdW[type[jj]])
1020                                     {
1021                                         if (charge[jj] != 0)
1022                                         {
1023                                             add_j_to_nblist(vdwc, jj);
1024                                         }
1025                                         else
1026                                         {
1027                                             add_j_to_nblist(vdw, jj);
1028                                         }
1029                                     }
1030                                     else if (charge[jj] != 0)
1031                                     {
1032                                         add_j_to_nblist(coul, jj);
1033                                     }
1034                                 }
1035                             }
1036                         }
1037                     }
1038                 }
1039             }
1040             close_i_nblist(vdw);
1041             close_i_nblist(coul);
1042             close_i_nblist(vdwc);
1043             close_i_nblist(vdw_free);
1044             close_i_nblist(coul_free);
1045             close_i_nblist(vdwc_free);
1046         }
1047     }
1048 }
1049
1050 static void
1051 put_in_list_qmmm(const gmx_bool gmx_unused        bHaveVdW[],
1052                  int                              ngid,
1053                  const t_mdatoms                  * /* md */,
1054                  int                              icg,
1055                  int                              jgid,
1056                  int                              nj,
1057                  const int                        jjcg[],
1058                  const int                        index[],
1059                  const t_excl                     bExcl[],
1060                  int                              shift,
1061                  t_forcerec                *      fr,
1062                  gmx_bool  gmx_unused             bDoVdW,
1063                  gmx_bool  gmx_unused             bDoCoul,
1064                  int       gmx_unused             solvent_opt)
1065 {
1066     t_nblist  *   coul;
1067     int           i, j, jcg, igid, gid;
1068     int           jj, jj0, jj1, i_atom;
1069     int           i0, nicg;
1070     gmx_bool      bNotEx;
1071
1072     /* Get atom range */
1073     i0     = index[icg];
1074     nicg   = index[icg+1]-i0;
1075
1076     /* Get the i charge group info */
1077     igid   = GET_CGINFO_GID(fr->cginfo[icg]);
1078
1079     coul = fr->QMMMlist;
1080
1081     /* Loop over atoms in the ith charge group */
1082     for (i = 0; i < nicg; i++)
1083     {
1084         i_atom = i0+i;
1085         gid    = GID(igid, jgid, ngid);
1086         /* Create new i_atom for each energy group */
1087         new_i_nblist(coul, i_atom, shift, gid);
1088
1089         /* Loop over the j charge groups */
1090         for (j = 0; j < nj; j++)
1091         {
1092             jcg = jjcg[j];
1093
1094             /* Charge groups cannot have QM and MM atoms simultaneously */
1095             if (jcg != icg)
1096             {
1097                 jj0 = index[jcg];
1098                 jj1 = index[jcg+1];
1099                 /* Finally loop over the atoms in the j-charge group */
1100                 for (jj = jj0; jj < jj1; jj++)
1101                 {
1102                     bNotEx = NOTEXCL(bExcl, i, jj);
1103                     if (bNotEx)
1104                     {
1105                         add_j_to_nblist(coul, jj);
1106                     }
1107                 }
1108             }
1109         }
1110         close_i_nblist(coul);
1111     }
1112 }
1113
1114 static void
1115 put_in_list_cg(const gmx_bool  gmx_unused       bHaveVdW[],
1116                int                              ngid,
1117                const t_mdatoms                  * /* md */,
1118                int                              icg,
1119                int                              jgid,
1120                int                              nj,
1121                const int                        jjcg[],
1122                const int                        index[],
1123                const t_excl                     bExcl[],
1124                int                              shift,
1125                t_forcerec                *      fr,
1126                gmx_bool   gmx_unused            bDoVdW,
1127                gmx_bool   gmx_unused            bDoCoul,
1128                int        gmx_unused            solvent_opt)
1129 {
1130     int          cginfo;
1131     int          igid, gid, nbl_ind;
1132     t_nblist *   vdwc;
1133     int          j, jcg;
1134
1135     cginfo = fr->cginfo[icg];
1136
1137     igid = GET_CGINFO_GID(cginfo);
1138     gid  = GID(igid, jgid, ngid);
1139
1140     /* Unpack pointers to neighbourlist structs */
1141     if (fr->nnblists == 1)
1142     {
1143         nbl_ind = 0;
1144     }
1145     else
1146     {
1147         nbl_ind = fr->gid2nblists[gid];
1148     }
1149     vdwc = &fr->nblists[nbl_ind].nlist_sr[eNL_VDWQQ];
1150
1151     /* Make a new neighbor list for charge group icg.
1152      * Currently simply one neighbor list is made with LJ and Coulomb.
1153      * If required, zero interactions could be removed here
1154      * or in the force loop.
1155      */
1156     new_i_nblist(vdwc, index[icg], shift, gid);
1157     vdwc->iinr_end[vdwc->nri] = index[icg+1];
1158
1159     for (j = 0; (j < nj); j++)
1160     {
1161         jcg = jjcg[j];
1162         /* Skip the icg-icg pairs if all self interactions are excluded */
1163         if (!(jcg == icg && GET_CGINFO_EXCL_INTRA(cginfo)))
1164         {
1165             /* Here we add the j charge group jcg to the list,
1166              * exclusions are also added to the list.
1167              */
1168             add_j_to_nblist_cg(vdwc, index[jcg], index[jcg+1], bExcl, icg == jcg);
1169         }
1170     }
1171
1172     close_i_nblist(vdwc);
1173 }
1174
1175 static void setexcl(int start, int end, t_blocka *excl, gmx_bool b,
1176                     t_excl bexcl[])
1177 {
1178     int i, k;
1179
1180     if (b)
1181     {
1182         for (i = start; i < end; i++)
1183         {
1184             for (k = excl->index[i]; k < excl->index[i+1]; k++)
1185             {
1186                 SETEXCL(bexcl, i-start, excl->a[k]);
1187             }
1188         }
1189     }
1190     else
1191     {
1192         for (i = start; i < end; i++)
1193         {
1194             for (k = excl->index[i]; k < excl->index[i+1]; k++)
1195             {
1196                 RMEXCL(bexcl, i-start, excl->a[k]);
1197             }
1198         }
1199     }
1200 }
1201
1202 int calc_naaj(int icg, int cgtot)
1203 {
1204     int naaj;
1205
1206     if ((cgtot % 2) == 1)
1207     {
1208         /* Odd number of charge groups, easy */
1209         naaj = 1 + (cgtot/2);
1210     }
1211     else if ((cgtot % 4) == 0)
1212     {
1213         /* Multiple of four is hard */
1214         if (icg < cgtot/2)
1215         {
1216             if ((icg % 2) == 0)
1217             {
1218                 naaj = 1+(cgtot/2);
1219             }
1220             else
1221             {
1222                 naaj = cgtot/2;
1223             }
1224         }
1225         else
1226         {
1227             if ((icg % 2) == 1)
1228             {
1229                 naaj = 1+(cgtot/2);
1230             }
1231             else
1232             {
1233                 naaj = cgtot/2;
1234             }
1235         }
1236     }
1237     else
1238     {
1239         /* cgtot/2 = odd */
1240         if ((icg % 2) == 0)
1241         {
1242             naaj = 1+(cgtot/2);
1243         }
1244         else
1245         {
1246             naaj = cgtot/2;
1247         }
1248     }
1249 #ifdef DEBUG
1250     fprintf(log, "naaj=%d\n", naaj);
1251 #endif
1252
1253     return naaj;
1254 }
1255
1256 /************************************************
1257  *
1258  *  S I M P L E      C O R E     S T U F F
1259  *
1260  ************************************************/
1261
1262 static real calc_image_tric(const rvec xi, const rvec xj, matrix box,
1263                             const rvec b_inv, int *shift)
1264 {
1265     /* This code assumes that the cut-off is smaller than
1266      * a half times the smallest diagonal element of the box.
1267      */
1268     const real h25 = 2.5;
1269     real       dx, dy, dz;
1270     real       r2;
1271     int        tx, ty, tz;
1272
1273     /* Compute diff vector */
1274     dz = xj[ZZ] - xi[ZZ];
1275     dy = xj[YY] - xi[YY];
1276     dx = xj[XX] - xi[XX];
1277
1278     /* Perform NINT operation, using trunc operation, therefore
1279      * we first add 2.5 then subtract 2 again
1280      */
1281     tz  = static_cast<int>(dz*b_inv[ZZ] + h25);
1282     tz -= 2;
1283     dz -= tz*box[ZZ][ZZ];
1284     dy -= tz*box[ZZ][YY];
1285     dx -= tz*box[ZZ][XX];
1286
1287     ty  = static_cast<int>(dy*b_inv[YY] + h25);
1288     ty -= 2;
1289     dy -= ty*box[YY][YY];
1290     dx -= ty*box[YY][XX];
1291
1292     tx  = static_cast<int>(dx*b_inv[XX]+h25);
1293     tx -= 2;
1294     dx -= tx*box[XX][XX];
1295
1296     /* Distance squared */
1297     r2 = (dx*dx) + (dy*dy) + (dz*dz);
1298
1299     *shift = XYZ2IS(tx, ty, tz);
1300
1301     return r2;
1302 }
1303
1304 static real calc_image_rect(const rvec xi, const rvec xj, const rvec box_size,
1305                             const rvec b_inv, int *shift)
1306 {
1307     const real h15 = 1.5;
1308     real       ddx, ddy, ddz;
1309     real       dx, dy, dz;
1310     real       r2;
1311     int        tx, ty, tz;
1312
1313     /* Compute diff vector */
1314     dx = xj[XX] - xi[XX];
1315     dy = xj[YY] - xi[YY];
1316     dz = xj[ZZ] - xi[ZZ];
1317
1318     /* Perform NINT operation, using trunc operation, therefore
1319      * we first add 1.5 then subtract 1 again
1320      */
1321     tx = static_cast<int>(dx*b_inv[XX] + h15);
1322     ty = static_cast<int>(dy*b_inv[YY] + h15);
1323     tz = static_cast<int>(dz*b_inv[ZZ] + h15);
1324     tx--;
1325     ty--;
1326     tz--;
1327
1328     /* Correct diff vector for translation */
1329     ddx = tx*box_size[XX] - dx;
1330     ddy = ty*box_size[YY] - dy;
1331     ddz = tz*box_size[ZZ] - dz;
1332
1333     /* Distance squared */
1334     r2 = (ddx*ddx) + (ddy*ddy) + (ddz*ddz);
1335
1336     *shift = XYZ2IS(tx, ty, tz);
1337
1338     return r2;
1339 }
1340
1341 static void add_simple(t_ns_buf       * nsbuf,
1342                        int              nrj,
1343                        int              cg_j,
1344                        gmx_bool         bHaveVdW[],
1345                        int              ngid,
1346                        const t_mdatoms *md,
1347                        int              icg,
1348                        int              jgid,
1349                        t_block         *cgs,
1350                        t_excl           bexcl[],
1351                        int              shift,
1352                        t_forcerec      *fr,
1353                        put_in_list_t   *put_in_list)
1354 {
1355     if (nsbuf->nj + nrj > MAX_CG)
1356     {
1357         put_in_list(bHaveVdW, ngid, md, icg, jgid, nsbuf->ncg, nsbuf->jcg,
1358                     cgs->index, bexcl, shift, fr, TRUE, TRUE, fr->solvent_opt);
1359         /* Reset buffer contents */
1360         nsbuf->ncg = nsbuf->nj = 0;
1361     }
1362     nsbuf->jcg[nsbuf->ncg++] = cg_j;
1363     nsbuf->nj               += nrj;
1364 }
1365
1366 static void ns_inner_tric(rvec                   x[],
1367                           int                    icg,
1368                           const int             *i_egp_flags,
1369                           int                    njcg,
1370                           const int              jcg[],
1371                           matrix                 box,
1372                           rvec                   b_inv,
1373                           real                   rcut2,
1374                           t_block               *cgs,
1375                           t_ns_buf             **ns_buf,
1376                           gmx_bool               bHaveVdW[],
1377                           int                    ngid,
1378                           const t_mdatoms       *md,
1379                           t_excl                 bexcl[],
1380                           t_forcerec            *fr,
1381                           put_in_list_t         *put_in_list)
1382 {
1383     int       shift;
1384     int       j, nrj, jgid;
1385     int      *cginfo = fr->cginfo;
1386     int       cg_j, *cgindex;
1387
1388     cgindex = cgs->index;
1389     shift   = CENTRAL;
1390     for (j = 0; (j < njcg); j++)
1391     {
1392         cg_j   = jcg[j];
1393         nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1394         if (calc_image_tric(x[icg], x[cg_j], box, b_inv, &shift) < rcut2)
1395         {
1396             jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1397             if (!(i_egp_flags[jgid] & EGP_EXCL))
1398             {
1399                 add_simple(&ns_buf[jgid][shift], nrj, cg_j,
1400                            bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, shift, fr,
1401                            put_in_list);
1402             }
1403         }
1404     }
1405 }
1406
1407 static void ns_inner_rect(rvec                   x[],
1408                           int                    icg,
1409                           const int             *i_egp_flags,
1410                           int                    njcg,
1411                           const int              jcg[],
1412                           gmx_bool               bBox,
1413                           rvec                   box_size,
1414                           rvec                   b_inv,
1415                           real                   rcut2,
1416                           t_block               *cgs,
1417                           t_ns_buf             **ns_buf,
1418                           gmx_bool               bHaveVdW[],
1419                           int                    ngid,
1420                           const t_mdatoms       *md,
1421                           t_excl                 bexcl[],
1422                           t_forcerec            *fr,
1423                           put_in_list_t         *put_in_list)
1424 {
1425     int       shift;
1426     int       j, nrj, jgid;
1427     int      *cginfo = fr->cginfo;
1428     int       cg_j, *cgindex;
1429
1430     cgindex = cgs->index;
1431     if (bBox)
1432     {
1433         shift = CENTRAL;
1434         for (j = 0; (j < njcg); j++)
1435         {
1436             cg_j   = jcg[j];
1437             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1438             if (calc_image_rect(x[icg], x[cg_j], box_size, b_inv, &shift) < rcut2)
1439             {
1440                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1441                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1442                 {
1443                     add_simple(&ns_buf[jgid][shift], nrj, cg_j,
1444                                bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, shift, fr,
1445                                put_in_list);
1446                 }
1447             }
1448         }
1449     }
1450     else
1451     {
1452         for (j = 0; (j < njcg); j++)
1453         {
1454             cg_j   = jcg[j];
1455             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1456             if ((rcut2 == 0) || (distance2(x[icg], x[cg_j]) < rcut2))
1457             {
1458                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1459                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1460                 {
1461                     add_simple(&ns_buf[jgid][CENTRAL], nrj, cg_j,
1462                                bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, CENTRAL, fr,
1463                                put_in_list);
1464                 }
1465             }
1466         }
1467     }
1468 }
1469
1470 /* ns_simple_core needs to be adapted for QMMM still 2005 */
1471
1472 static int ns_simple_core(t_forcerec      *fr,
1473                           gmx_localtop_t  *top,
1474                           const t_mdatoms *md,
1475                           matrix           box,
1476                           rvec             box_size,
1477                           t_excl           bexcl[],
1478                           int             *aaj,
1479                           int              ngid,
1480                           t_ns_buf       **ns_buf,
1481                           put_in_list_t   *put_in_list,
1482                           gmx_bool         bHaveVdW[])
1483 {
1484     int          naaj, k;
1485     real         rlist2;
1486     int          nsearch, icg, igid, nn;
1487     int         *cginfo;
1488     t_ns_buf    *nsbuf;
1489     /* int  *i_atoms; */
1490     t_block     *cgs  = &(top->cgs);
1491     t_blocka    *excl = &(top->excls);
1492     rvec         b_inv;
1493     int          m;
1494     gmx_bool     bBox, bTriclinic;
1495     int         *i_egp_flags;
1496
1497     rlist2 = gmx::square(fr->rlist);
1498
1499     bBox = (fr->ePBC != epbcNONE);
1500     if (bBox)
1501     {
1502         for (m = 0; (m < DIM); m++)
1503         {
1504             if (gmx_numzero(box_size[m]))
1505             {
1506                 gmx_fatal(FARGS, "Dividing by zero box size!");
1507             }
1508             b_inv[m] = 1.0/box_size[m];
1509         }
1510         bTriclinic = TRICLINIC(box);
1511     }
1512     else
1513     {
1514         bTriclinic = FALSE;
1515     }
1516
1517     cginfo = fr->cginfo;
1518
1519     nsearch = 0;
1520     for (icg = fr->cg0; (icg < fr->hcg); icg++)
1521     {
1522         /*
1523            i0        = cgs->index[icg];
1524            nri       = cgs->index[icg+1]-i0;
1525            i_atoms   = &(cgs->a[i0]);
1526            i_eg_excl = fr->eg_excl + ngid*md->cENER[*i_atoms];
1527            setexcl(nri,i_atoms,excl,TRUE,bexcl);
1528          */
1529         igid        = GET_CGINFO_GID(cginfo[icg]);
1530         i_egp_flags = fr->egp_flags + ngid*igid;
1531         setexcl(cgs->index[icg], cgs->index[icg+1], excl, TRUE, bexcl);
1532
1533         naaj = calc_naaj(icg, cgs->nr);
1534         if (bTriclinic)
1535         {
1536             ns_inner_tric(fr->cg_cm, icg, i_egp_flags, naaj, &(aaj[icg]),
1537                           box, b_inv, rlist2, cgs, ns_buf,
1538                           bHaveVdW, ngid, md, bexcl, fr, put_in_list);
1539         }
1540         else
1541         {
1542             ns_inner_rect(fr->cg_cm, icg, i_egp_flags, naaj, &(aaj[icg]),
1543                           bBox, box_size, b_inv, rlist2, cgs, ns_buf,
1544                           bHaveVdW, ngid, md, bexcl, fr, put_in_list);
1545         }
1546         nsearch += naaj;
1547
1548         for (nn = 0; (nn < ngid); nn++)
1549         {
1550             for (k = 0; (k < SHIFTS); k++)
1551             {
1552                 nsbuf = &(ns_buf[nn][k]);
1553                 if (nsbuf->ncg > 0)
1554                 {
1555                     put_in_list(bHaveVdW, ngid, md, icg, nn, nsbuf->ncg, nsbuf->jcg,
1556                                 cgs->index, bexcl, k, fr, TRUE, TRUE, fr->solvent_opt);
1557                     nsbuf->ncg = nsbuf->nj = 0;
1558                 }
1559             }
1560         }
1561         /* setexcl(nri,i_atoms,excl,FALSE,bexcl); */
1562         setexcl(cgs->index[icg], cgs->index[icg+1], excl, FALSE, bexcl);
1563     }
1564     close_neighbor_lists(fr, FALSE);
1565
1566     return nsearch;
1567 }
1568
1569 /************************************************
1570  *
1571  *    N S 5     G R I D     S T U F F
1572  *
1573  ************************************************/
1574
1575 static inline void get_dx_dd(int Nx, real gridx, real rc2, int xgi, real x,
1576                              int ncpddc, int shift_min, int shift_max,
1577                              int *g0, int *g1, real *dcx2)
1578 {
1579     real dcx, tmp;
1580     int  g_min, g_max, shift_home;
1581
1582     if (xgi < 0)
1583     {
1584         g_min = 0;
1585         g_max = Nx - 1;
1586         *g0   = 0;
1587         *g1   = -1;
1588     }
1589     else if (xgi >= Nx)
1590     {
1591         g_min = 0;
1592         g_max = Nx - 1;
1593         *g0   = Nx;
1594         *g1   = Nx - 1;
1595     }
1596     else
1597     {
1598         if (ncpddc == 0)
1599         {
1600             g_min = 0;
1601             g_max = Nx - 1;
1602         }
1603         else
1604         {
1605             if (xgi < ncpddc)
1606             {
1607                 shift_home = 0;
1608             }
1609             else
1610             {
1611                 shift_home = -1;
1612             }
1613             g_min = (shift_min == shift_home ? 0          : ncpddc);
1614             g_max = (shift_max == shift_home ? ncpddc - 1 : Nx - 1);
1615         }
1616         if (shift_min > 0)
1617         {
1618             *g0 = g_min;
1619             *g1 = g_min - 1;
1620         }
1621         else if (shift_max < 0)
1622         {
1623             *g0 = g_max + 1;
1624             *g1 = g_max;
1625         }
1626         else
1627         {
1628             *g0       = xgi;
1629             *g1       = xgi;
1630             dcx2[xgi] = 0;
1631         }
1632     }
1633
1634     while (*g0 > g_min)
1635     {
1636         /* Check one grid cell down */
1637         dcx = ((*g0 - 1) + 1)*gridx - x;
1638         tmp = dcx*dcx;
1639         if (tmp >= rc2)
1640         {
1641             break;
1642         }
1643         (*g0)--;
1644         dcx2[*g0] = tmp;
1645     }
1646
1647     while (*g1 < g_max)
1648     {
1649         /* Check one grid cell up */
1650         dcx = (*g1 + 1)*gridx - x;
1651         tmp = dcx*dcx;
1652         if (tmp >= rc2)
1653         {
1654             break;
1655         }
1656         (*g1)++;
1657         dcx2[*g1] = tmp;
1658     }
1659 }
1660
1661
1662 #define calc_dx2(XI, YI, ZI, y) (gmx::square((XI)-(y)[XX]) + gmx::square((YI)-(y)[YY]) + gmx::square((ZI)-(y)[ZZ]))
1663 #define calc_cyl_dx2(XI, YI, y) (gmx::square((XI)-(y)[XX]) + gmx::square((YI)-(y)[YY]))
1664 /****************************************************
1665  *
1666  *    F A S T   N E I G H B O R  S E A R C H I N G
1667  *
1668  *    Optimized neighboursearching routine using grid
1669  *    at least 1x1x1, see GROMACS manual
1670  *
1671  ****************************************************/
1672
1673
1674 static void get_cutoff2(t_forcerec *fr, real *rs2)
1675 {
1676     *rs2 = gmx::square(fr->rlist);
1677 }
1678
1679 static void init_nsgrid_lists(t_forcerec *fr, int ngid, gmx_ns_t *ns)
1680 {
1681     real rs2;
1682     int  j;
1683
1684     get_cutoff2(fr, &rs2);
1685
1686     /* Short range buffers */
1687     snew(ns->nl_sr, ngid);
1688     /* Counters */
1689     snew(ns->nsr, ngid);
1690
1691     for (j = 0; (j < ngid); j++)
1692     {
1693         snew(ns->nl_sr[j], MAX_CG);
1694     }
1695     if (debug)
1696     {
1697         fprintf(debug,
1698                 "ns5_core: rs2 = %g (nm^2)\n",
1699                 rs2);
1700     }
1701 }
1702
1703 static int nsgrid_core(const t_commrec *cr,
1704                        t_forcerec      *fr,
1705                        matrix           box,
1706                        int              ngid,
1707                        gmx_localtop_t  *top,
1708                        t_grid          *grid,
1709                        t_excl           bexcl[],
1710                        const gmx_bool  *bExcludeAlleg,
1711                        const t_mdatoms *md,
1712                        put_in_list_t   *put_in_list,
1713                        gmx_bool         bHaveVdW[],
1714                        gmx_bool         bMakeQMMMnblist)
1715 {
1716     gmx_ns_t      *ns;
1717     int          **nl_sr;
1718     int           *nsr;
1719     gmx_domdec_t  *dd;
1720     const t_block *cgs    = &(top->cgs);
1721     int           *cginfo = fr->cginfo;
1722     /* int *i_atoms,*cgsindex=cgs->index; */
1723     ivec           sh0, sh1, shp;
1724     int            cell_x, cell_y, cell_z;
1725     int            d, tx, ty, tz, dx, dy, dz, cj;
1726 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
1727     int            zsh_ty, zsh_tx, ysh_tx;
1728 #endif
1729     int            dx0, dx1, dy0, dy1, dz0, dz1;
1730     int            Nx, Ny, Nz, shift = -1, j, nrj, nns, nn = -1;
1731     real           gridx, gridy, gridz, grid_x, grid_y;
1732     real          *dcx2, *dcy2, *dcz2;
1733     int            zgi, ygi, xgi;
1734     int            cg0, cg1, icg = -1, cgsnr, i0, igid, naaj, max_jcg;
1735     int            jcg0, jcg1, jjcg, cgj0, jgid;
1736     int           *grida, *gridnra, *gridind;
1737     rvec          *cgcm, grid_offset;
1738     real           r2, rs2, XI, YI, ZI, tmp1, tmp2;
1739     int           *i_egp_flags;
1740     gmx_bool       bDomDec, bTriclinicX, bTriclinicY;
1741     ivec           ncpddc;
1742
1743     ns = fr->ns;
1744
1745     bDomDec = DOMAINDECOMP(cr);
1746     dd      = cr->dd;
1747
1748     bTriclinicX = ((YY < grid->npbcdim &&
1749                     (!bDomDec || dd->nc[YY] == 1) && box[YY][XX] != 0) ||
1750                    (ZZ < grid->npbcdim &&
1751                     (!bDomDec || dd->nc[ZZ] == 1) && box[ZZ][XX] != 0));
1752     bTriclinicY =  (ZZ < grid->npbcdim &&
1753                     (!bDomDec || dd->nc[ZZ] == 1) && box[ZZ][YY] != 0);
1754
1755     cgsnr    = cgs->nr;
1756
1757     get_cutoff2(fr, &rs2);
1758
1759     nl_sr     = ns->nl_sr;
1760     nsr       = ns->nsr;
1761
1762     /* Unpack arrays */
1763     cgcm    = fr->cg_cm;
1764     Nx      = grid->n[XX];
1765     Ny      = grid->n[YY];
1766     Nz      = grid->n[ZZ];
1767     grida   = grid->a;
1768     gridind = grid->index;
1769     gridnra = grid->nra;
1770     nns     = 0;
1771
1772     gridx      = grid->cell_size[XX];
1773     gridy      = grid->cell_size[YY];
1774     gridz      = grid->cell_size[ZZ];
1775     grid_x     = 1/gridx;
1776     grid_y     = 1/gridy;
1777     copy_rvec(grid->cell_offset, grid_offset);
1778     copy_ivec(grid->ncpddc, ncpddc);
1779     dcx2       = grid->dcx2;
1780     dcy2       = grid->dcy2;
1781     dcz2       = grid->dcz2;
1782
1783 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
1784     zsh_ty = floor(-box[ZZ][YY]/box[YY][YY]+0.5);
1785     zsh_tx = floor(-box[ZZ][XX]/box[XX][XX]+0.5);
1786     ysh_tx = floor(-box[YY][XX]/box[XX][XX]+0.5);
1787     if (zsh_tx != 0 && ysh_tx != 0)
1788     {
1789         /* This could happen due to rounding, when both ratios are 0.5 */
1790         ysh_tx = 0;
1791     }
1792 #endif
1793
1794     if (fr->n_tpi)
1795     {
1796         /* We only want a list for the test particle */
1797         cg0 = cgsnr - 1;
1798     }
1799     else
1800     {
1801         cg0 = grid->icg0;
1802     }
1803     cg1 = grid->icg1;
1804
1805     /* Set the shift range */
1806     for (d = 0; d < DIM; d++)
1807     {
1808         sh0[d] = -1;
1809         sh1[d] = 1;
1810         /* Check if we need periodicity shifts.
1811          * Without PBC or with domain decomposition we don't need them.
1812          */
1813         if (d >= ePBC2npbcdim(fr->ePBC) || (bDomDec && dd->nc[d] > 1))
1814         {
1815             shp[d] = 0;
1816         }
1817         else
1818         {
1819             if (d == XX &&
1820                 box[XX][XX] - std::abs(box[YY][XX]) - std::abs(box[ZZ][XX]) < std::sqrt(rs2))
1821             {
1822                 shp[d] = 2;
1823             }
1824             else
1825             {
1826                 shp[d] = 1;
1827             }
1828         }
1829     }
1830
1831     /* Loop over charge groups */
1832     for (icg = cg0; (icg < cg1); icg++)
1833     {
1834         igid = GET_CGINFO_GID(cginfo[icg]);
1835         /* Skip this charge group if all energy groups are excluded! */
1836         if (bExcludeAlleg[igid])
1837         {
1838             continue;
1839         }
1840
1841         i0   = cgs->index[icg];
1842
1843         if (bMakeQMMMnblist)
1844         {
1845             /* Skip this charge group if it is not a QM atom while making a
1846              * QM/MM neighbourlist
1847              */
1848             if (md->bQM[i0] == FALSE)
1849             {
1850                 continue; /* MM particle, go to next particle */
1851             }
1852
1853             /* Compute the number of charge groups that fall within the control
1854              * of this one (icg)
1855              */
1856             naaj    = calc_naaj(icg, cgsnr);
1857             jcg0    = icg;
1858             jcg1    = icg + naaj;
1859             max_jcg = cgsnr;
1860         }
1861         else
1862         {
1863             /* make a normal neighbourlist */
1864
1865             if (bDomDec)
1866             {
1867                 /* Get the j charge-group and dd cell shift ranges */
1868                 dd_get_ns_ranges(cr->dd, icg, &jcg0, &jcg1, sh0, sh1);
1869                 max_jcg = 0;
1870             }
1871             else
1872             {
1873                 /* Compute the number of charge groups that fall within the control
1874                  * of this one (icg)
1875                  */
1876                 naaj = calc_naaj(icg, cgsnr);
1877                 jcg0 = icg;
1878                 jcg1 = icg + naaj;
1879
1880                 if (fr->n_tpi)
1881                 {
1882                     /* The i-particle is awlways the test particle,
1883                      * so we want all j-particles
1884                      */
1885                     max_jcg = cgsnr - 1;
1886                 }
1887                 else
1888                 {
1889                     max_jcg  = jcg1 - cgsnr;
1890                 }
1891             }
1892         }
1893
1894         i_egp_flags = fr->egp_flags + igid*ngid;
1895
1896         /* Set the exclusions for the atoms in charge group icg using a bitmask */
1897         setexcl(i0, cgs->index[icg+1], &top->excls, TRUE, bexcl);
1898
1899         ci2xyz(grid, icg, &cell_x, &cell_y, &cell_z);
1900
1901         /* Changed iicg to icg, DvdS 990115
1902          * (but see consistency check above, DvdS 990330)
1903          */
1904 #ifdef NS5DB
1905         fprintf(log, "icg=%5d, naaj=%5d, cell %d %d %d\n",
1906                 icg, naaj, cell_x, cell_y, cell_z);
1907 #endif
1908         /* Loop over shift vectors in three dimensions */
1909         for (tz = -shp[ZZ]; tz <= shp[ZZ]; tz++)
1910         {
1911             ZI = cgcm[icg][ZZ]+tz*box[ZZ][ZZ];
1912             /* Calculate range of cells in Z direction that have the shift tz */
1913             zgi = cell_z + tz*Nz;
1914             get_dx_dd(Nz, gridz, rs2, zgi, ZI-grid_offset[ZZ],
1915                       ncpddc[ZZ], sh0[ZZ], sh1[ZZ], &dz0, &dz1, dcz2);
1916             if (dz0 > dz1)
1917             {
1918                 continue;
1919             }
1920             for (ty = -shp[YY]; ty <= shp[YY]; ty++)
1921             {
1922                 YI = cgcm[icg][YY]+ty*box[YY][YY]+tz*box[ZZ][YY];
1923                 /* Calculate range of cells in Y direction that have the shift ty */
1924                 if (bTriclinicY)
1925                 {
1926                     ygi = static_cast<int>(Ny + (YI - grid_offset[YY])*grid_y) - Ny;
1927                 }
1928                 else
1929                 {
1930                     ygi = cell_y + ty*Ny;
1931                 }
1932                 get_dx_dd(Ny, gridy, rs2, ygi, YI-grid_offset[YY],
1933                           ncpddc[YY], sh0[YY], sh1[YY], &dy0, &dy1, dcy2);
1934                 if (dy0 > dy1)
1935                 {
1936                     continue;
1937                 }
1938                 for (tx = -shp[XX]; tx <= shp[XX]; tx++)
1939                 {
1940                     XI = cgcm[icg][XX]+tx*box[XX][XX]+ty*box[YY][XX]+tz*box[ZZ][XX];
1941                     /* Calculate range of cells in X direction that have the shift tx */
1942                     if (bTriclinicX)
1943                     {
1944                         xgi = static_cast<int>(Nx + (XI - grid_offset[XX])*grid_x) - Nx;
1945                     }
1946                     else
1947                     {
1948                         xgi = cell_x + tx*Nx;
1949                     }
1950                     get_dx_dd(Nx, gridx, rs2, xgi, XI-grid_offset[XX],
1951                               ncpddc[XX], sh0[XX], sh1[XX], &dx0, &dx1, dcx2);
1952                     if (dx0 > dx1)
1953                     {
1954                         continue;
1955                     }
1956                     /* Get shift vector */
1957                     shift = XYZ2IS(tx, ty, tz);
1958 #ifdef NS5DB
1959                     range_check(shift, 0, SHIFTS);
1960 #endif
1961                     for (nn = 0; (nn < ngid); nn++)
1962                     {
1963                         nsr[nn]      = 0;
1964                     }
1965 #ifdef NS5DB
1966                     fprintf(log, "shift: %2d, dx0,1: %2d,%2d, dy0,1: %2d,%2d, dz0,1: %2d,%2d\n",
1967                             shift, dx0, dx1, dy0, dy1, dz0, dz1);
1968                     fprintf(log, "cgcm: %8.3f  %8.3f  %8.3f\n", cgcm[icg][XX],
1969                             cgcm[icg][YY], cgcm[icg][ZZ]);
1970                     fprintf(log, "xi:   %8.3f  %8.3f  %8.3f\n", XI, YI, ZI);
1971 #endif
1972                     for (dx = dx0; (dx <= dx1); dx++)
1973                     {
1974                         tmp1 = rs2 - dcx2[dx];
1975                         for (dy = dy0; (dy <= dy1); dy++)
1976                         {
1977                             tmp2 = tmp1 - dcy2[dy];
1978                             if (tmp2 > 0)
1979                             {
1980                                 for (dz = dz0; (dz <= dz1); dz++)
1981                                 {
1982                                     if (tmp2 > dcz2[dz])
1983                                     {
1984                                         /* Find grid-cell cj in which possible neighbours are */
1985                                         cj   = xyz2ci(Ny, Nz, dx, dy, dz);
1986
1987                                         /* Check out how many cgs (nrj) there in this cell */
1988                                         nrj  = gridnra[cj];
1989
1990                                         /* Find the offset in the cg list */
1991                                         cgj0 = gridind[cj];
1992
1993                                         /* Check if all j's are out of range so we
1994                                          * can skip the whole cell.
1995                                          * Should save some time, especially with DD.
1996                                          */
1997                                         if (nrj == 0 ||
1998                                             (grida[cgj0] >= max_jcg &&
1999                                              (grida[cgj0] >= jcg1 || grida[cgj0+nrj-1] < jcg0)))
2000                                         {
2001                                             continue;
2002                                         }
2003
2004                                         /* Loop over cgs */
2005                                         for (j = 0; (j < nrj); j++)
2006                                         {
2007                                             jjcg = grida[cgj0+j];
2008
2009                                             /* check whether this guy is in range! */
2010                                             if ((jjcg >= jcg0 && jjcg < jcg1) ||
2011                                                 (jjcg < max_jcg))
2012                                             {
2013                                                 r2 = calc_dx2(XI, YI, ZI, cgcm[jjcg]);
2014                                                 if (r2 < rs2)
2015                                                 {
2016                                                     /* jgid = gid[cgsatoms[cgsindex[jjcg]]]; */
2017                                                     jgid = GET_CGINFO_GID(cginfo[jjcg]);
2018                                                     /* check energy group exclusions */
2019                                                     if (!(i_egp_flags[jgid] & EGP_EXCL))
2020                                                     {
2021                                                         if (nsr[jgid] >= MAX_CG)
2022                                                         {
2023                                                             /* Add to short-range list */
2024                                                             put_in_list(bHaveVdW, ngid, md, icg, jgid,
2025                                                                         nsr[jgid], nl_sr[jgid],
2026                                                                         cgs->index, /* cgsatoms, */ bexcl,
2027                                                                         shift, fr, TRUE, TRUE, fr->solvent_opt);
2028                                                             nsr[jgid] = 0;
2029                                                         }
2030                                                         nl_sr[jgid][nsr[jgid]++] = jjcg;
2031                                                     }
2032                                                 }
2033                                                 nns++;
2034                                             }
2035                                         }
2036                                     }
2037                                 }
2038                             }
2039                         }
2040                     }
2041                     /* CHECK whether there is anything left in the buffers */
2042                     for (nn = 0; (nn < ngid); nn++)
2043                     {
2044                         if (nsr[nn] > 0)
2045                         {
2046                             put_in_list(bHaveVdW, ngid, md, icg, nn, nsr[nn], nl_sr[nn],
2047                                         cgs->index, /* cgsatoms, */ bexcl,
2048                                         shift, fr, TRUE, TRUE, fr->solvent_opt);
2049                         }
2050                     }
2051                 }
2052             }
2053         }
2054         setexcl(cgs->index[icg], cgs->index[icg+1], &top->excls, FALSE, bexcl);
2055     }
2056     /* No need to perform any left-over force calculations anymore (as we used to do here)
2057      * since we now save the proper long-range lists for later evaluation.
2058      */
2059
2060     /* Close neighbourlists */
2061     close_neighbor_lists(fr, bMakeQMMMnblist);
2062
2063     return nns;
2064 }
2065
2066 static void ns_realloc_natoms(gmx_ns_t *ns, int natoms)
2067 {
2068     int i;
2069
2070     if (natoms > ns->nra_alloc)
2071     {
2072         ns->nra_alloc = over_alloc_dd(natoms);
2073         srenew(ns->bexcl, ns->nra_alloc);
2074         for (i = 0; i < ns->nra_alloc; i++)
2075         {
2076             ns->bexcl[i] = 0;
2077         }
2078     }
2079 }
2080
2081 void init_ns(FILE *fplog, const t_commrec *cr,
2082              gmx_ns_t *ns, t_forcerec *fr,
2083              const gmx_mtop_t *mtop)
2084 {
2085     int      icg, nr_in_cg, maxcg, i, j, jcg, ngid, ncg;
2086
2087     /* Compute largest charge groups size (# atoms) */
2088     nr_in_cg = 1;
2089     for (const gmx_moltype_t &molt : mtop->moltype)
2090     {
2091         const t_block *cgs = &molt.cgs;
2092         for (icg = 0; (icg < cgs->nr); icg++)
2093         {
2094             nr_in_cg = std::max(nr_in_cg, (cgs->index[icg+1]-cgs->index[icg]));
2095         }
2096     }
2097
2098     /* Verify whether largest charge group is <= max cg.
2099      * This is determined by the type of the local exclusion type
2100      * Exclusions are stored in bits. (If the type is not large
2101      * enough, enlarge it, unsigned char -> unsigned short -> unsigned long)
2102      */
2103     maxcg = sizeof(t_excl)*8;
2104     if (nr_in_cg > maxcg)
2105     {
2106         gmx_fatal(FARGS, "Max #atoms in a charge group: %d > %d\n",
2107                   nr_in_cg, maxcg);
2108     }
2109
2110     ngid = mtop->groups.grps[egcENER].nr;
2111     snew(ns->bExcludeAlleg, ngid);
2112     for (i = 0; i < ngid; i++)
2113     {
2114         ns->bExcludeAlleg[i] = TRUE;
2115         for (j = 0; j < ngid; j++)
2116         {
2117             if (!(fr->egp_flags[i*ngid+j] & EGP_EXCL))
2118             {
2119                 ns->bExcludeAlleg[i] = FALSE;
2120             }
2121         }
2122     }
2123
2124     if (fr->bGrid)
2125     {
2126         /* Grid search */
2127         ns->grid = init_grid(fplog, fr);
2128         init_nsgrid_lists(fr, ngid, ns);
2129     }
2130     else
2131     {
2132         /* Simple search */
2133         snew(ns->ns_buf, ngid);
2134         for (i = 0; (i < ngid); i++)
2135         {
2136             snew(ns->ns_buf[i], SHIFTS);
2137         }
2138         ncg = ncg_mtop(mtop);
2139         snew(ns->simple_aaj, 2*ncg);
2140         for (jcg = 0; (jcg < ncg); jcg++)
2141         {
2142             ns->simple_aaj[jcg]     = jcg;
2143             ns->simple_aaj[jcg+ncg] = jcg;
2144         }
2145     }
2146
2147     /* Create array that determines whether or not atoms have VdW */
2148     snew(ns->bHaveVdW, fr->ntype);
2149     for (i = 0; (i < fr->ntype); i++)
2150     {
2151         for (j = 0; (j < fr->ntype); j++)
2152         {
2153             ns->bHaveVdW[i] = (ns->bHaveVdW[i] ||
2154                                (fr->bBHAM ?
2155                                 ((BHAMA(fr->nbfp, fr->ntype, i, j) != 0) ||
2156                                  (BHAMB(fr->nbfp, fr->ntype, i, j) != 0) ||
2157                                  (BHAMC(fr->nbfp, fr->ntype, i, j) != 0)) :
2158                                 ((C6(fr->nbfp, fr->ntype, i, j) != 0) ||
2159                                  (C12(fr->nbfp, fr->ntype, i, j) != 0))));
2160         }
2161     }
2162     if (debug)
2163     {
2164         pr_bvec(debug, 0, "bHaveVdW", ns->bHaveVdW, fr->ntype, TRUE);
2165     }
2166
2167     ns->nra_alloc = 0;
2168     ns->bexcl     = nullptr;
2169     if (!DOMAINDECOMP(cr))
2170     {
2171         ns_realloc_natoms(ns, mtop->natoms);
2172     }
2173
2174     ns->nblist_initialized = FALSE;
2175
2176     /* nbr list debug dump */
2177     {
2178         char *ptr = getenv("GMX_DUMP_NL");
2179         if (ptr)
2180         {
2181             ns->dump_nl = strtol(ptr, nullptr, 10);
2182             if (fplog)
2183             {
2184                 fprintf(fplog, "GMX_DUMP_NL = %d", ns->dump_nl);
2185             }
2186         }
2187         else
2188         {
2189             ns->dump_nl = 0;
2190         }
2191     }
2192 }
2193
2194 void done_ns(gmx_ns_t *ns, int numEnergyGroups)
2195 {
2196     sfree(ns->bExcludeAlleg);
2197     if (ns->ns_buf)
2198     {
2199         for (int i = 0; i < numEnergyGroups; i++)
2200         {
2201             sfree(ns->ns_buf[i]);
2202         }
2203         sfree(ns->ns_buf);
2204     }
2205     sfree(ns->simple_aaj);
2206     sfree(ns->bHaveVdW);
2207     done_grid(ns->grid);
2208     sfree(ns);
2209 }
2210
2211 int search_neighbours(FILE               *log,
2212                       t_forcerec         *fr,
2213                       matrix              box,
2214                       gmx_localtop_t     *top,
2215                       const gmx_groups_t *groups,
2216                       const t_commrec    *cr,
2217                       t_nrnb             *nrnb,
2218                       const t_mdatoms    *md,
2219                       gmx_bool            bFillGrid)
2220 {
2221     const t_block      *cgs = &(top->cgs);
2222     rvec                box_size, grid_x0, grid_x1;
2223     int                 m, ngid;
2224     real                min_size, grid_dens;
2225     int                 nsearch;
2226     gmx_bool            bGrid;
2227     int                 start, end;
2228     gmx_ns_t           *ns;
2229     t_grid             *grid;
2230     gmx_domdec_zones_t *dd_zones;
2231     put_in_list_t      *put_in_list;
2232
2233     ns = fr->ns;
2234
2235     /* Set some local variables */
2236     bGrid = fr->bGrid;
2237     ngid  = groups->grps[egcENER].nr;
2238
2239     for (m = 0; (m < DIM); m++)
2240     {
2241         box_size[m] = box[m][m];
2242     }
2243
2244     if (fr->ePBC != epbcNONE)
2245     {
2246         if (gmx::square(fr->rlist) >= max_cutoff2(fr->ePBC, box))
2247         {
2248             gmx_fatal(FARGS, "One of the box vectors has become shorter than twice the cut-off length or box_yy-|box_zy| or box_zz has become smaller than the cut-off.");
2249         }
2250         if (!bGrid)
2251         {
2252             min_size = std::min(box_size[XX], std::min(box_size[YY], box_size[ZZ]));
2253             if (2*fr->rlist >= min_size)
2254             {
2255                 gmx_fatal(FARGS, "One of the box diagonal elements has become smaller than twice the cut-off length.");
2256             }
2257         }
2258     }
2259
2260     if (DOMAINDECOMP(cr))
2261     {
2262         ns_realloc_natoms(ns, cgs->index[cgs->nr]);
2263     }
2264
2265     /* Reset the neighbourlists */
2266     reset_neighbor_lists(fr);
2267
2268     if (bGrid && bFillGrid)
2269     {
2270
2271         grid = ns->grid;
2272         if (DOMAINDECOMP(cr))
2273         {
2274             dd_zones = domdec_zones(cr->dd);
2275         }
2276         else
2277         {
2278             dd_zones = nullptr;
2279
2280             get_nsgrid_boundaries(grid->nboundeddim, box, nullptr, nullptr, nullptr, nullptr,
2281                                   cgs->nr, fr->cg_cm, grid_x0, grid_x1, &grid_dens);
2282
2283             grid_first(log, grid, nullptr, nullptr, box, grid_x0, grid_x1,
2284                        fr->rlist, grid_dens);
2285         }
2286
2287         start = 0;
2288         end   = cgs->nr;
2289
2290         if (DOMAINDECOMP(cr))
2291         {
2292             end = cgs->nr;
2293             fill_grid(dd_zones, grid, end, -1, end, fr->cg_cm);
2294             grid->icg0 = 0;
2295             grid->icg1 = dd_zones->izone[dd_zones->nizone-1].cg1;
2296         }
2297         else
2298         {
2299             fill_grid(nullptr, grid, cgs->nr, fr->cg0, fr->hcg, fr->cg_cm);
2300             grid->icg0 = fr->cg0;
2301             grid->icg1 = fr->hcg;
2302         }
2303
2304         calc_elemnr(grid, start, end, cgs->nr);
2305         calc_ptrs(grid);
2306         grid_last(grid, start, end, cgs->nr);
2307
2308         if (gmx_debug_at)
2309         {
2310             check_grid(grid);
2311             print_grid(debug, grid);
2312         }
2313     }
2314     else if (fr->n_tpi)
2315     {
2316         /* Set the grid cell index for the test particle only.
2317          * The cell to cg index is not corrected, but that does not matter.
2318          */
2319         fill_grid(nullptr, ns->grid, fr->hcg, fr->hcg-1, fr->hcg, fr->cg_cm);
2320     }
2321
2322     if (!fr->ns->bCGlist)
2323     {
2324         put_in_list = put_in_list_at;
2325     }
2326     else
2327     {
2328         put_in_list = put_in_list_cg;
2329     }
2330
2331     /* Do the core! */
2332     if (bGrid)
2333     {
2334         grid    = ns->grid;
2335         nsearch = nsgrid_core(cr, fr, box, ngid, top,
2336                               grid, ns->bexcl, ns->bExcludeAlleg,
2337                               md, put_in_list, ns->bHaveVdW,
2338                               FALSE);
2339
2340         /* neighbour searching withouth QMMM! QM atoms have zero charge in
2341          * the classical calculation. The charge-charge interaction
2342          * between QM and MM atoms is handled in the QMMM core calculation
2343          * (see QMMM.c). The VDW however, we'd like to compute classically
2344          * and the QM MM atom pairs have just been put in the
2345          * corresponding neighbourlists. in case of QMMM we still need to
2346          * fill a special QMMM neighbourlist that contains all neighbours
2347          * of the QM atoms. If bQMMM is true, this list will now be made:
2348          */
2349         if (fr->bQMMM && fr->qr->QMMMscheme != eQMMMschemeoniom)
2350         {
2351             nsearch += nsgrid_core(cr, fr, box, ngid, top,
2352                                    grid, ns->bexcl, ns->bExcludeAlleg,
2353                                    md, put_in_list_qmmm, ns->bHaveVdW,
2354                                    TRUE);
2355         }
2356     }
2357     else
2358     {
2359         nsearch = ns_simple_core(fr, top, md, box, box_size,
2360                                  ns->bexcl, ns->simple_aaj,
2361                                  ngid, ns->ns_buf, put_in_list, ns->bHaveVdW);
2362     }
2363
2364     inc_nrnb(nrnb, eNR_NS, nsearch);
2365
2366     return nsearch;
2367 }