src/gromacs/mdlib/ns.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * Copyright (c) 2013,2014, by the GROMACS development team, led by
   7  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   8  * and including many others, as listed in the AUTHORS file in the
   9  * top-level source directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37 #ifdef HAVE_CONFIG_H
  38 #include <config.h>
  39 #endif
  40
  41 #include <math.h>
  42 #include <stdlib.h>
  43 #include <string.h>
  44
  45 #include "gromacs/utility/smalloc.h"
  46 #include "macros.h"
  47 #include "gromacs/math/utilities.h"
  48 #include "gromacs/math/vec.h"
  49 #include "types/commrec.h"
  50 #include "network.h"
  51 #include "nsgrid.h"
  52 #include "force.h"
  53 #include "nonbonded.h"
  54 #include "ns.h"
  55 #include "pbc.h"
  56 #include "names.h"
  57 #include "gromacs/utility/fatalerror.h"
  58 #include "nrnb.h"
  59 #include "txtdump.h"
  60 #include "mtop_util.h"
  61
  62 #include "domdec.h"
  63 #include "adress.h"
  64
  65
  66 /*
  67  *    E X C L U S I O N   H A N D L I N G
  68  */
  69
  70 #ifdef DEBUG
  71 static void SETEXCL_(t_excl e[], atom_id i, atom_id j)
  72 {
  73     e[j] = e[j] | (1<<i);
  74 }
  75 static void RMEXCL_(t_excl e[], atom_id i, atom_id j)
  76 {
  77     e[j] = e[j] & ~(1<<i);
  78 }
  79 static gmx_bool ISEXCL_(t_excl e[], atom_id i, atom_id j)
  80 {
  81     return (gmx_bool)(e[j] & (1<<i));
  82 }
  83 static gmx_bool NOTEXCL_(t_excl e[], atom_id i, atom_id j)
  84 {
  85     return !(ISEXCL(e, i, j));
  86 }
  87 #else
  88 #define SETEXCL(e, i, j) (e)[((atom_id) (j))] |= (1<<((atom_id) (i)))
  89 #define RMEXCL(e, i, j)  (e)[((atom_id) (j))] &= (~(1<<((atom_id) (i))))
  90 #define ISEXCL(e, i, j)  (gmx_bool) ((e)[((atom_id) (j))] & (1<<((atom_id) (i))))
  91 #define NOTEXCL(e, i, j) !(ISEXCL(e, i, j))
  92 #endif
  93
  94 static int
  95 round_up_to_simd_width(int length, int simd_width)
  96 {
  97     int offset, newlength;
  98
  99     offset = (simd_width > 0) ? length % simd_width : 0;
 100
 101     return (offset == 0) ? length : length-offset+simd_width;
 102 }
 103 /************************************************
 104  *
 105  *  U T I L I T I E S    F O R    N S
 106  *
 107  ************************************************/
 108
 109 void reallocate_nblist(t_nblist *nl)
 110 {
 111     if (gmx_debug_at)
 112     {
 113         fprintf(debug, "reallocating neigborlist (ielec=%d, ivdw=%d, igeometry=%d, type=%d), maxnri=%d\n",
 114                 nl->ielec, nl->ivdw, nl->igeometry, nl->type, nl->maxnri);
 115     }
 116     srenew(nl->iinr,   nl->maxnri);
 117     if (nl->igeometry == GMX_NBLIST_GEOMETRY_CG_CG)
 118     {
 119         srenew(nl->iinr_end, nl->maxnri);
 120     }
 121     srenew(nl->gid,    nl->maxnri);
 122     srenew(nl->shift,  nl->maxnri);
 123     srenew(nl->jindex, nl->maxnri+1);
 124 }
 125
 126
 127 static void init_nblist(FILE *log, t_nblist *nl_sr, t_nblist *nl_lr,
 128                         int maxsr, int maxlr,
 129                         int ivdw, int ivdwmod,
 130                         int ielec, int ielecmod,
 131                         int igeometry, int type)
 132 {
 133     t_nblist *nl;
 134     int       homenr;
 135     int       i, nn;
 136
 137     for (i = 0; (i < 2); i++)
 138     {
 139         nl     = (i == 0) ? nl_sr : nl_lr;
 140         homenr = (i == 0) ? maxsr : maxlr;
 141
 142         if (nl == NULL)
 143         {
 144             continue;
 145         }
 146
 147
 148         /* Set coul/vdw in neighborlist, and for the normal loops we determine
 149          * an index of which one to call.
 150          */
 151         nl->ivdw        = ivdw;
 152         nl->ivdwmod     = ivdwmod;
 153         nl->ielec       = ielec;
 154         nl->ielecmod    = ielecmod;
 155         nl->type        = type;
 156         nl->igeometry   = igeometry;
 157
 158         if (nl->type == GMX_NBLIST_INTERACTION_FREE_ENERGY)
 159         {
 160             nl->igeometry  = GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE;
 161         }
 162
 163         /* This will also set the simd_padding_width field */
 164         gmx_nonbonded_set_kernel_pointers( (i == 0) ? log : NULL, nl);
 165
 166         /* maxnri is influenced by the number of shifts (maximum is 8)
 167          * and the number of energy groups.
 168          * If it is not enough, nl memory will be reallocated during the run.
 169          * 4 seems to be a reasonable factor, which only causes reallocation
 170          * during runs with tiny and many energygroups.
 171          */
 172         nl->maxnri      = homenr*4;
 173         nl->maxnrj      = 0;
 174         nl->nri         = -1;
 175         nl->nrj         = 0;
 176         nl->iinr        = NULL;
 177         nl->gid         = NULL;
 178         nl->shift       = NULL;
 179         nl->jindex      = NULL;
 180         nl->jjnr        = NULL;
 181         nl->excl_fep    = NULL;
 182         reallocate_nblist(nl);
 183         nl->jindex[0] = 0;
 184
 185         if (debug)
 186         {
 187             fprintf(debug, "Initiating neighbourlist (ielec=%d, ivdw=%d, type=%d) for %s interactions,\nwith %d SR, %d LR atoms.\n",
 188                     nl->ielec, nl->ivdw, nl->type, gmx_nblist_geometry_names[nl->igeometry], maxsr, maxlr);
 189         }
 190     }
 191 }
 192
 193 void init_neighbor_list(FILE *log, t_forcerec *fr, int homenr)
 194 {
 195     /* Make maxlr tunable! (does not seem to be a big difference though)
 196      * This parameter determines the number of i particles in a long range
 197      * neighbourlist. Too few means many function calls, too many means
 198      * cache trashing.
 199      */
 200     int        maxsr, maxsr_wat, maxlr, maxlr_wat;
 201     int        ielec, ielecf, ivdw, ielecmod, ielecmodf, ivdwmod, type;
 202     int        solvent;
 203     int        igeometry_def, igeometry_w, igeometry_ww;
 204     int        i;
 205     t_nblists *nbl;
 206
 207     /* maxsr     = homenr-fr->nWatMol*3; */
 208     maxsr     = homenr;
 209
 210     if (maxsr < 0)
 211     {
 212         gmx_fatal(FARGS, "%s, %d: Negative number of short range atoms.\n"
 213                   "Call your Gromacs dealer for assistance.", __FILE__, __LINE__);
 214     }
 215     /* This is just for initial allocation, so we do not reallocate
 216      * all the nlist arrays many times in a row.
 217      * The numbers seem very accurate, but they are uncritical.
 218      */
 219     maxsr_wat = min(fr->nWatMol, (homenr+2)/3);
 220     if (fr->bTwinRange)
 221     {
 222         maxlr     = 50;
 223         maxlr_wat = min(maxsr_wat, maxlr);
 224     }
 225     else
 226     {
 227         maxlr = maxlr_wat = 0;
 228     }
 229
 230     /* Determine the values for ielec/ivdw. */
 231     ielec    = fr->nbkernel_elec_interaction;
 232     ivdw     = fr->nbkernel_vdw_interaction;
 233     ielecmod = fr->nbkernel_elec_modifier;
 234     ivdwmod  = fr->nbkernel_vdw_modifier;
 235     type     = GMX_NBLIST_INTERACTION_STANDARD;
 236
 237     fr->ns.bCGlist = (getenv("GMX_NBLISTCG") != 0);
 238     if (!fr->ns.bCGlist)
 239     {
 240         igeometry_def = GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE;
 241     }
 242     else
 243     {
 244         igeometry_def = GMX_NBLIST_GEOMETRY_CG_CG;
 245         if (log != NULL)
 246         {
 247             fprintf(log, "\nUsing charge-group - charge-group neighbor lists and kernels\n\n");
 248         }
 249     }
 250
 251     if (fr->solvent_opt == esolTIP4P)
 252     {
 253         igeometry_w  = GMX_NBLIST_GEOMETRY_WATER4_PARTICLE;
 254         igeometry_ww = GMX_NBLIST_GEOMETRY_WATER4_WATER4;
 255     }
 256     else
 257     {
 258         igeometry_w  = GMX_NBLIST_GEOMETRY_WATER3_PARTICLE;
 259         igeometry_ww = GMX_NBLIST_GEOMETRY_WATER3_WATER3;
 260     }
 261
 262     for (i = 0; i < fr->nnblists; i++)
 263     {
 264         nbl = &(fr->nblists[i]);
 265
 266         if ((fr->adress_type != eAdressOff) && (i >= fr->nnblists/2))
 267         {
 268             type = GMX_NBLIST_INTERACTION_ADRESS;
 269         }
 270         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ], &nbl->nlist_lr[eNL_VDWQQ],
 271                     maxsr, maxlr, ivdw, ivdwmod, ielec, ielecmod, igeometry_def, type);
 272         init_nblist(log, &nbl->nlist_sr[eNL_VDW], &nbl->nlist_lr[eNL_VDW],
 273                     maxsr, maxlr, ivdw, ivdwmod, GMX_NBKERNEL_ELEC_NONE, eintmodNONE, igeometry_def, type);
 274         init_nblist(log, &nbl->nlist_sr[eNL_QQ], &nbl->nlist_lr[eNL_QQ],
 275                     maxsr, maxlr, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_def, type);
 276         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_WATER], &nbl->nlist_lr[eNL_VDWQQ_WATER],
 277                     maxsr_wat, maxlr_wat, ivdw, ivdwmod, ielec, ielecmod, igeometry_w, type);
 278         init_nblist(log, &nbl->nlist_sr[eNL_QQ_WATER], &nbl->nlist_lr[eNL_QQ_WATER],
 279                     maxsr_wat, maxlr_wat, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_w, type);
 280         init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_WATERWATER], &nbl->nlist_lr[eNL_VDWQQ_WATERWATER],
 281                     maxsr_wat, maxlr_wat, ivdw, ivdwmod, ielec, ielecmod, igeometry_ww, type);
 282         init_nblist(log, &nbl->nlist_sr[eNL_QQ_WATERWATER], &nbl->nlist_lr[eNL_QQ_WATERWATER],
 283                     maxsr_wat, maxlr_wat, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielec, ielecmod, igeometry_ww, type);
 284
 285         /* Did we get the solvent loops so we can use optimized water kernels? */
 286         if (nbl->nlist_sr[eNL_VDWQQ_WATER].kernelptr_vf == NULL
 287             || nbl->nlist_sr[eNL_QQ_WATER].kernelptr_vf == NULL
 288 #ifndef DISABLE_WATERWATER_NLIST
 289             || nbl->nlist_sr[eNL_VDWQQ_WATERWATER].kernelptr_vf == NULL
 290             || nbl->nlist_sr[eNL_QQ_WATERWATER].kernelptr_vf == NULL
 291 #endif
 292             )
 293         {
 294             fr->solvent_opt = esolNO;
 295             if (log != NULL)
 296             {
 297                 fprintf(log, "Note: The available nonbonded kernels do not support water optimization - disabling.\n");
 298             }
 299         }
 300
 301         if (fr->efep != efepNO)
 302         {
 303             if ((fr->bEwald) && (fr->sc_alphacoul > 0)) /* need to handle long range differently if using softcore */
 304             {
 305                 ielecf    = GMX_NBKERNEL_ELEC_EWALD;
 306                 ielecmodf = eintmodNONE;
 307             }
 308             else
 309             {
 310                 ielecf    = ielec;
 311                 ielecmodf = ielecmod;
 312             }
 313
 314             init_nblist(log, &nbl->nlist_sr[eNL_VDWQQ_FREE], &nbl->nlist_lr[eNL_VDWQQ_FREE],
 315                         maxsr, maxlr, ivdw, ivdwmod, ielecf, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY);
 316             init_nblist(log, &nbl->nlist_sr[eNL_VDW_FREE], &nbl->nlist_lr[eNL_VDW_FREE],
 317                         maxsr, maxlr, ivdw, ivdwmod, GMX_NBKERNEL_ELEC_NONE, eintmodNONE, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY);
 318             init_nblist(log, &nbl->nlist_sr[eNL_QQ_FREE], &nbl->nlist_lr[eNL_QQ_FREE],
 319                         maxsr, maxlr, GMX_NBKERNEL_VDW_NONE, eintmodNONE, ielecf, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_FREE_ENERGY);
 320         }
 321     }
 322     /* QMMM MM list */
 323     if (fr->bQMMM && fr->qr->QMMMscheme != eQMMMschemeoniom)
 324     {
 325         init_nblist(log, &fr->QMMMlist, NULL,
 326                     maxsr, maxlr, 0, 0, ielec, ielecmod, GMX_NBLIST_GEOMETRY_PARTICLE_PARTICLE, GMX_NBLIST_INTERACTION_STANDARD);
 327     }
 328
 329     if (log != NULL)
 330     {
 331         fprintf(log, "\n");
 332     }
 333
 334     fr->ns.nblist_initialized = TRUE;
 335 }
 336
 337 static void reset_nblist(t_nblist *nl)
 338 {
 339     nl->nri       = -1;
 340     nl->nrj       = 0;
 341     if (nl->jindex)
 342     {
 343         nl->jindex[0] = 0;
 344     }
 345 }
 346
 347 static void reset_neighbor_lists(t_forcerec *fr, gmx_bool bResetSR, gmx_bool bResetLR)
 348 {
 349     int n, i;
 350
 351     if (fr->bQMMM)
 352     {
 353         /* only reset the short-range nblist */
 354         reset_nblist(&(fr->QMMMlist));
 355     }
 356
 357     for (n = 0; n < fr->nnblists; n++)
 358     {
 359         for (i = 0; i < eNL_NR; i++)
 360         {
 361             if (bResetSR)
 362             {
 363                 reset_nblist( &(fr->nblists[n].nlist_sr[i]) );
 364             }
 365             if (bResetLR)
 366             {
 367                 reset_nblist( &(fr->nblists[n].nlist_lr[i]) );
 368             }
 369         }
 370     }
 371 }
 372
 373
 374
 375
 376 static gmx_inline void new_i_nblist(t_nblist *nlist, atom_id i_atom, int shift, int gid)
 377 {
 378     int    i, k, nri, nshift;
 379
 380     nri = nlist->nri;
 381
 382     /* Check whether we have to increase the i counter */
 383     if ((nri == -1) ||
 384         (nlist->iinr[nri]  != i_atom) ||
 385         (nlist->shift[nri] != shift) ||
 386         (nlist->gid[nri]   != gid))
 387     {
 388         /* This is something else. Now see if any entries have
 389          * been added in the list of the previous atom.
 390          */
 391         if ((nri == -1) ||
 392             ((nlist->jindex[nri+1] > nlist->jindex[nri]) &&
 393              (nlist->gid[nri] != -1)))
 394         {
 395             /* If so increase the counter */
 396             nlist->nri++;
 397             nri++;
 398             if (nlist->nri >= nlist->maxnri)
 399             {
 400                 nlist->maxnri += over_alloc_large(nlist->nri);
 401                 reallocate_nblist(nlist);
 402             }
 403         }
 404         /* Set the number of neighbours and the atom number */
 405         nlist->jindex[nri+1] = nlist->jindex[nri];
 406         nlist->iinr[nri]     = i_atom;
 407         nlist->gid[nri]      = gid;
 408         nlist->shift[nri]    = shift;
 409     }
 410     else
 411     {
 412         /* Adding to previous list. First remove possible previous padding */
 413         if (nlist->simd_padding_width > 1)
 414         {
 415             while (nlist->nrj > 0 && nlist->jjnr[nlist->nrj-1] < 0)
 416             {
 417                 nlist->nrj--;
 418             }
 419         }
 420     }
 421 }
 422
 423 static gmx_inline void close_i_nblist(t_nblist *nlist)
 424 {
 425     int nri = nlist->nri;
 426     int len;
 427
 428     if (nri >= 0)
 429     {
 430         /* Add elements up to padding. Since we allocate memory in units
 431          * of the simd_padding width, we do not have to check for possible
 432          * list reallocation here.
 433          */
 434         while ((nlist->nrj % nlist->simd_padding_width) != 0)
 435         {
 436             /* Use -4 here, so we can write forces for 4 atoms before real data */
 437             nlist->jjnr[nlist->nrj++] = -4;
 438         }
 439         nlist->jindex[nri+1] = nlist->nrj;
 440
 441         len = nlist->nrj -  nlist->jindex[nri];
 442     }
 443 }
 444
 445 static gmx_inline void close_nblist(t_nblist *nlist)
 446 {
 447     /* Only close this nblist when it has been initialized.
 448      * Avoid the creation of i-lists with no j-particles.
 449      */
 450     if (nlist->nrj == 0)
 451     {
 452         /* Some assembly kernels do not support empty lists,
 453          * make sure here that we don't generate any empty lists.
 454          * With the current ns code this branch is taken in two cases:
 455          * No i-particles at all: nri=-1 here
 456          * There are i-particles, but no j-particles; nri=0 here
 457          */
 458         nlist->nri = 0;
 459     }
 460     else
 461     {
 462         /* Close list number nri by incrementing the count */
 463         nlist->nri++;
 464     }
 465 }
 466
 467 static gmx_inline void close_neighbor_lists(t_forcerec *fr, gmx_bool bMakeQMMMnblist)
 468 {
 469     int n, i;
 470
 471     if (bMakeQMMMnblist)
 472     {
 473         close_nblist(&(fr->QMMMlist));
 474     }
 475
 476     for (n = 0; n < fr->nnblists; n++)
 477     {
 478         for (i = 0; (i < eNL_NR); i++)
 479         {
 480             close_nblist(&(fr->nblists[n].nlist_sr[i]));
 481             close_nblist(&(fr->nblists[n].nlist_lr[i]));
 482         }
 483     }
 484 }
 485
 486
 487 static gmx_inline void add_j_to_nblist(t_nblist *nlist, atom_id j_atom, gmx_bool bLR)
 488 {
 489     int nrj = nlist->nrj;
 490
 491     if (nlist->nrj >= nlist->maxnrj)
 492     {
 493         nlist->maxnrj = round_up_to_simd_width(over_alloc_small(nlist->nrj + 1), nlist->simd_padding_width);
 494
 495         if (gmx_debug_at)
 496         {
 497             fprintf(debug, "Increasing %s nblist (ielec=%d,ivdw=%d,type=%d,igeometry=%d) j size to %d\n",
 498                     bLR ? "LR" : "SR", nlist->ielec, nlist->ivdw, nlist->type, nlist->igeometry, nlist->maxnrj);
 499         }
 500
 501         srenew(nlist->jjnr, nlist->maxnrj);
 502     }
 503
 504     nlist->jjnr[nrj] = j_atom;
 505     nlist->nrj++;
 506 }
 507
 508 static gmx_inline void add_j_to_nblist_cg(t_nblist *nlist,
 509                                           atom_id j_start, int j_end,
 510                                           t_excl *bexcl, gmx_bool i_is_j,
 511                                           gmx_bool bLR)
 512 {
 513     int nrj = nlist->nrj;
 514     int j;
 515
 516     if (nlist->nrj >= nlist->maxnrj)
 517     {
 518         nlist->maxnrj = over_alloc_small(nlist->nrj + 1);
 519         if (gmx_debug_at)
 520         {
 521             fprintf(debug, "Increasing %s nblist (ielec=%d,ivdw=%d,type=%d,igeometry=%d) j size to %d\n",
 522                     bLR ? "LR" : "SR", nlist->ielec, nlist->ivdw, nlist->type, nlist->igeometry, nlist->maxnrj);
 523         }
 524
 525         srenew(nlist->jjnr, nlist->maxnrj);
 526         srenew(nlist->jjnr_end, nlist->maxnrj);
 527         srenew(nlist->excl, nlist->maxnrj*MAX_CGCGSIZE);
 528     }
 529
 530     nlist->jjnr[nrj]     = j_start;
 531     nlist->jjnr_end[nrj] = j_end;
 532
 533     if (j_end - j_start > MAX_CGCGSIZE)
 534     {
 535         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);
 536     }
 537
 538     /* Set the exclusions */
 539     for (j = j_start; j < j_end; j++)
 540     {
 541         nlist->excl[nrj*MAX_CGCGSIZE + j - j_start] = bexcl[j];
 542     }
 543     if (i_is_j)
 544     {
 545         /* Avoid double counting of intra-cg interactions */
 546         for (j = 1; j < j_end-j_start; j++)
 547         {
 548             nlist->excl[nrj*MAX_CGCGSIZE + j] |= (1<<j) - 1;
 549         }
 550     }
 551
 552     nlist->nrj++;
 553 }
 554
 555 typedef void
 556     put_in_list_t (gmx_bool              bHaveVdW[],
 557                    int                   ngid,
 558                    t_mdatoms     *       md,
 559                    int                   icg,
 560                    int                   jgid,
 561                    int                   nj,
 562                    atom_id               jjcg[],
 563                    atom_id               index[],
 564                    t_excl                bExcl[],
 565                    int                   shift,
 566                    t_forcerec     *      fr,
 567                    gmx_bool              bLR,
 568                    gmx_bool              bDoVdW,
 569                    gmx_bool              bDoCoul,
 570                    int                   solvent_opt);
 571
 572 static void
 573 put_in_list_at(gmx_bool              bHaveVdW[],
 574                int                   ngid,
 575                t_mdatoms     *       md,
 576                int                   icg,
 577                int                   jgid,
 578                int                   nj,
 579                atom_id               jjcg[],
 580                atom_id               index[],
 581                t_excl                bExcl[],
 582                int                   shift,
 583                t_forcerec     *      fr,
 584                gmx_bool              bLR,
 585                gmx_bool              bDoVdW,
 586                gmx_bool              bDoCoul,
 587                int                   solvent_opt)
 588 {
 589     /* The a[] index has been removed,
 590      * to put it back in i_atom should be a[i0] and jj should be a[jj].
 591      */
 592     t_nblist  *   vdwc;
 593     t_nblist  *   vdw;
 594     t_nblist  *   coul;
 595     t_nblist  *   vdwc_free  = NULL;
 596     t_nblist  *   vdw_free   = NULL;
 597     t_nblist  *   coul_free  = NULL;
 598     t_nblist  *   vdwc_ww    = NULL;
 599     t_nblist  *   coul_ww    = NULL;
 600
 601     int           i, j, jcg, igid, gid, nbl_ind, ind_ij;
 602     atom_id       jj, jj0, jj1, i_atom;
 603     int           i0, nicg, len;
 604
 605     int          *cginfo;
 606     int          *type, *typeB;
 607     real         *charge, *chargeB;
 608     real          qi, qiB, qq, rlj;
 609     gmx_bool      bFreeEnergy, bFree, bFreeJ, bNotEx, *bPert;
 610     gmx_bool      bDoVdW_i, bDoCoul_i, bDoCoul_i_sol;
 611     int           iwater, jwater;
 612     t_nblist     *nlist;
 613
 614     /* Copy some pointers */
 615     cginfo  = fr->cginfo;
 616     charge  = md->chargeA;
 617     chargeB = md->chargeB;
 618     type    = md->typeA;
 619     typeB   = md->typeB;
 620     bPert   = md->bPerturbed;
 621
 622     /* Get atom range */
 623     i0     = index[icg];
 624     nicg   = index[icg+1]-i0;
 625
 626     /* Get the i charge group info */
 627     igid   = GET_CGINFO_GID(cginfo[icg]);
 628
 629     iwater = (solvent_opt != esolNO) ? GET_CGINFO_SOLOPT(cginfo[icg]) : esolNO;
 630
 631     bFreeEnergy = FALSE;
 632     if (md->nPerturbed)
 633     {
 634         /* Check if any of the particles involved are perturbed.
 635          * If not we can do the cheaper normal put_in_list
 636          * and use more solvent optimization.
 637          */
 638         for (i = 0; i < nicg; i++)
 639         {
 640             bFreeEnergy |= bPert[i0+i];
 641         }
 642         /* Loop over the j charge groups */
 643         for (j = 0; (j < nj && !bFreeEnergy); j++)
 644         {
 645             jcg = jjcg[j];
 646             jj0 = index[jcg];
 647             jj1 = index[jcg+1];
 648             /* Finally loop over the atoms in the j-charge group */
 649             for (jj = jj0; jj < jj1; jj++)
 650             {
 651                 bFreeEnergy |= bPert[jj];
 652             }
 653         }
 654     }
 655
 656     /* Unpack pointers to neighbourlist structs */
 657     if (fr->nnblists == 1)
 658     {
 659         nbl_ind = 0;
 660     }
 661     else
 662     {
 663         nbl_ind = fr->gid2nblists[GID(igid, jgid, ngid)];
 664     }
 665     if (bLR)
 666     {
 667         nlist = fr->nblists[nbl_ind].nlist_lr;
 668     }
 669     else
 670     {
 671         nlist = fr->nblists[nbl_ind].nlist_sr;
 672     }
 673
 674     if (iwater != esolNO)
 675     {
 676         vdwc = &nlist[eNL_VDWQQ_WATER];
 677         vdw  = &nlist[eNL_VDW];
 678         coul = &nlist[eNL_QQ_WATER];
 679 #ifndef DISABLE_WATERWATER_NLIST
 680         vdwc_ww = &nlist[eNL_VDWQQ_WATERWATER];
 681         coul_ww = &nlist[eNL_QQ_WATERWATER];
 682 #endif
 683     }
 684     else
 685     {
 686         vdwc = &nlist[eNL_VDWQQ];
 687         vdw  = &nlist[eNL_VDW];
 688         coul = &nlist[eNL_QQ];
 689     }
 690
 691     if (!bFreeEnergy)
 692     {
 693         if (iwater != esolNO)
 694         {
 695             /* Loop over the atoms in the i charge group */
 696             i_atom  = i0;
 697             gid     = GID(igid, jgid, ngid);
 698             /* Create new i_atom for each energy group */
 699             if (bDoCoul && bDoVdW)
 700             {
 701                 new_i_nblist(vdwc, i_atom, shift, gid);
 702 #ifndef DISABLE_WATERWATER_NLIST
 703                 new_i_nblist(vdwc_ww, i_atom, shift, gid);
 704 #endif
 705             }
 706             if (bDoVdW)
 707             {
 708                 new_i_nblist(vdw, i_atom, shift, gid);
 709             }
 710             if (bDoCoul)
 711             {
 712                 new_i_nblist(coul, i_atom, shift, gid);
 713 #ifndef DISABLE_WATERWATER_NLIST
 714                 new_i_nblist(coul_ww, i_atom, shift, gid);
 715 #endif
 716             }
 717             /* Loop over the j charge groups */
 718             for (j = 0; (j < nj); j++)
 719             {
 720                 jcg = jjcg[j];
 721
 722                 if (jcg == icg)
 723                 {
 724                     continue;
 725                 }
 726
 727                 jj0    = index[jcg];
 728                 jwater = GET_CGINFO_SOLOPT(cginfo[jcg]);
 729
 730                 if (iwater == esolSPC && jwater == esolSPC)
 731                 {
 732                     /* Interaction between two SPC molecules */
 733                     if (!bDoCoul)
 734                     {
 735                         /* VdW only - only first atoms in each water interact */
 736                         add_j_to_nblist(vdw, jj0, bLR);
 737                     }
 738                     else
 739                     {
 740 #ifdef DISABLE_WATERWATER_NLIST
 741                         /* Add entries for the three atoms - only do VdW if we need to */
 742                         if (!bDoVdW)
 743                         {
 744                             add_j_to_nblist(coul, jj0, bLR);
 745                         }
 746                         else
 747                         {
 748                             add_j_to_nblist(vdwc, jj0, bLR);
 749                         }
 750                         add_j_to_nblist(coul, jj0+1, bLR);
 751                         add_j_to_nblist(coul, jj0+2, bLR);
 752 #else
 753                         /* One entry for the entire water-water interaction */
 754                         if (!bDoVdW)
 755                         {
 756                             add_j_to_nblist(coul_ww, jj0, bLR);
 757                         }
 758                         else
 759                         {
 760                             add_j_to_nblist(vdwc_ww, jj0, bLR);
 761                         }
 762 #endif
 763                     }
 764                 }
 765                 else if (iwater == esolTIP4P && jwater == esolTIP4P)
 766                 {
 767                     /* Interaction between two TIP4p molecules */
 768                     if (!bDoCoul)
 769                     {
 770                         /* VdW only - only first atoms in each water interact */
 771                         add_j_to_nblist(vdw, jj0, bLR);
 772                     }
 773                     else
 774                     {
 775 #ifdef DISABLE_WATERWATER_NLIST
 776                         /* Add entries for the four atoms - only do VdW if we need to */
 777                         if (bDoVdW)
 778                         {
 779                             add_j_to_nblist(vdw, jj0, bLR);
 780                         }
 781                         add_j_to_nblist(coul, jj0+1, bLR);
 782                         add_j_to_nblist(coul, jj0+2, bLR);
 783                         add_j_to_nblist(coul, jj0+3, bLR);
 784 #else
 785                         /* One entry for the entire water-water interaction */
 786                         if (!bDoVdW)
 787                         {
 788                             add_j_to_nblist(coul_ww, jj0, bLR);
 789                         }
 790                         else
 791                         {
 792                             add_j_to_nblist(vdwc_ww, jj0, bLR);
 793                         }
 794 #endif
 795                     }
 796                 }
 797                 else
 798                 {
 799                     /* j charge group is not water, but i is.
 800                      * Add entries to the water-other_atom lists; the geometry of the water
 801                      * molecule doesn't matter - that is taken care of in the nonbonded kernel,
 802                      * so we don't care if it is SPC or TIP4P...
 803                      */
 804
 805                     jj1 = index[jcg+1];
 806
 807                     if (!bDoVdW)
 808                     {
 809                         for (jj = jj0; (jj < jj1); jj++)
 810                         {
 811                             if (charge[jj] != 0)
 812                             {
 813                                 add_j_to_nblist(coul, jj, bLR);
 814                             }
 815                         }
 816                     }
 817                     else if (!bDoCoul)
 818                     {
 819                         for (jj = jj0; (jj < jj1); jj++)
 820                         {
 821                             if (bHaveVdW[type[jj]])
 822                             {
 823                                 add_j_to_nblist(vdw, jj, bLR);
 824                             }
 825                         }
 826                     }
 827                     else
 828                     {
 829                         /* _charge_ _groups_ interact with both coulomb and LJ */
 830                         /* Check which atoms we should add to the lists!       */
 831                         for (jj = jj0; (jj < jj1); jj++)
 832                         {
 833                             if (bHaveVdW[type[jj]])
 834                             {
 835                                 if (charge[jj] != 0)
 836                                 {
 837                                     add_j_to_nblist(vdwc, jj, bLR);
 838                                 }
 839                                 else
 840                                 {
 841                                     add_j_to_nblist(vdw, jj, bLR);
 842                                 }
 843                             }
 844                             else if (charge[jj] != 0)
 845                             {
 846                                 add_j_to_nblist(coul, jj, bLR);
 847                             }
 848                         }
 849                     }
 850                 }
 851             }
 852             close_i_nblist(vdw);
 853             close_i_nblist(coul);
 854             close_i_nblist(vdwc);
 855 #ifndef DISABLE_WATERWATER_NLIST
 856             close_i_nblist(coul_ww);
 857             close_i_nblist(vdwc_ww);
 858 #endif
 859         }
 860         else
 861         {
 862             /* no solvent as i charge group */
 863             /* Loop over the atoms in the i charge group */
 864             for (i = 0; i < nicg; i++)
 865             {
 866                 i_atom  = i0+i;
 867                 gid     = GID(igid, jgid, ngid);
 868                 qi      = charge[i_atom];
 869
 870                 /* Create new i_atom for each energy group */
 871                 if (bDoVdW && bDoCoul)
 872                 {
 873                     new_i_nblist(vdwc, i_atom, shift, gid);
 874                 }
 875                 if (bDoVdW)
 876                 {
 877                     new_i_nblist(vdw, i_atom, shift, gid);
 878                 }
 879                 if (bDoCoul)
 880                 {
 881                     new_i_nblist(coul, i_atom, shift, gid);
 882                 }
 883                 bDoVdW_i  = (bDoVdW  && bHaveVdW[type[i_atom]]);
 884                 bDoCoul_i = (bDoCoul && qi != 0);
 885
 886                 if (bDoVdW_i || bDoCoul_i)
 887                 {
 888                     /* Loop over the j charge groups */
 889                     for (j = 0; (j < nj); j++)
 890                     {
 891                         jcg = jjcg[j];
 892
 893                         /* Check for large charge groups */
 894                         if (jcg == icg)
 895                         {
 896                             jj0 = i0 + i + 1;
 897                         }
 898                         else
 899                         {
 900                             jj0 = index[jcg];
 901                         }
 902
 903                         jj1 = index[jcg+1];
 904                         /* Finally loop over the atoms in the j-charge group */
 905                         for (jj = jj0; jj < jj1; jj++)
 906                         {
 907                             bNotEx = NOTEXCL(bExcl, i, jj);
 908
 909                             if (bNotEx)
 910                             {
 911                                 if (!bDoVdW_i)
 912                                 {
 913                                     if (charge[jj] != 0)
 914                                     {
 915                                         add_j_to_nblist(coul, jj, bLR);
 916                                     }
 917                                 }
 918                                 else if (!bDoCoul_i)
 919                                 {
 920                                     if (bHaveVdW[type[jj]])
 921                                     {
 922                                         add_j_to_nblist(vdw, jj, bLR);
 923                                     }
 924                                 }
 925                                 else
 926                                 {
 927                                     if (bHaveVdW[type[jj]])
 928                                     {
 929                                         if (charge[jj] != 0)
 930                                         {
 931                                             add_j_to_nblist(vdwc, jj, bLR);
 932                                         }
 933                                         else
 934                                         {
 935                                             add_j_to_nblist(vdw, jj, bLR);
 936                                         }
 937                                     }
 938                                     else if (charge[jj] != 0)
 939                                     {
 940                                         add_j_to_nblist(coul, jj, bLR);
 941                                     }
 942                                 }
 943                             }
 944                         }
 945                     }
 946                 }
 947                 close_i_nblist(vdw);
 948                 close_i_nblist(coul);
 949                 close_i_nblist(vdwc);
 950             }
 951         }
 952     }
 953     else
 954     {
 955         /* we are doing free energy */
 956         vdwc_free = &nlist[eNL_VDWQQ_FREE];
 957         vdw_free  = &nlist[eNL_VDW_FREE];
 958         coul_free = &nlist[eNL_QQ_FREE];
 959         /* Loop over the atoms in the i charge group */
 960         for (i = 0; i < nicg; i++)
 961         {
 962             i_atom  = i0+i;
 963             gid     = GID(igid, jgid, ngid);
 964             qi      = charge[i_atom];
 965             qiB     = chargeB[i_atom];
 966
 967             /* Create new i_atom for each energy group */
 968             if (bDoVdW && bDoCoul)
 969             {
 970                 new_i_nblist(vdwc, i_atom, shift, gid);
 971             }
 972             if (bDoVdW)
 973             {
 974                 new_i_nblist(vdw, i_atom, shift, gid);
 975             }
 976             if (bDoCoul)
 977             {
 978                 new_i_nblist(coul, i_atom, shift, gid);
 979             }
 980
 981             new_i_nblist(vdw_free, i_atom, shift, gid);
 982             new_i_nblist(coul_free, i_atom, shift, gid);
 983             new_i_nblist(vdwc_free, i_atom, shift, gid);
 984
 985             bDoVdW_i  = (bDoVdW  &&
 986                          (bHaveVdW[type[i_atom]] || bHaveVdW[typeB[i_atom]]));
 987             bDoCoul_i = (bDoCoul && (qi != 0 || qiB != 0));
 988             /* For TIP4P the first atom does not have a charge,
 989              * but the last three do. So we should still put an atom
 990              * without LJ but with charge in the water-atom neighborlist
 991              * for a TIP4p i charge group.
 992              * For SPC type water the first atom has LJ and charge,
 993              * so there is no such problem.
 994              */
 995             if (iwater == esolNO)
 996             {
 997                 bDoCoul_i_sol = bDoCoul_i;
 998             }
 999             else
1000             {
1001                 bDoCoul_i_sol = bDoCoul;
1002             }
1003
1004             if (bDoVdW_i || bDoCoul_i_sol)
1005             {
1006                 /* Loop over the j charge groups */
1007                 for (j = 0; (j < nj); j++)
1008                 {
1009                     jcg = jjcg[j];
1010
1011                     /* Check for large charge groups */
1012                     if (jcg == icg)
1013                     {
1014                         jj0 = i0 + i + 1;
1015                     }
1016                     else
1017                     {
1018                         jj0 = index[jcg];
1019                     }
1020
1021                     jj1 = index[jcg+1];
1022                     /* Finally loop over the atoms in the j-charge group */
1023                     bFree = bPert[i_atom];
1024                     for (jj = jj0; (jj < jj1); jj++)
1025                     {
1026                         bFreeJ = bFree || bPert[jj];
1027                         /* Complicated if, because the water H's should also
1028                          * see perturbed j-particles
1029                          */
1030                         if (iwater == esolNO || i == 0 || bFreeJ)
1031                         {
1032                             bNotEx = NOTEXCL(bExcl, i, jj);
1033
1034                             if (bNotEx)
1035                             {
1036                                 if (bFreeJ)
1037                                 {
1038                                     if (!bDoVdW_i)
1039                                     {
1040                                         if (charge[jj] != 0 || chargeB[jj] != 0)
1041                                         {
1042                                             add_j_to_nblist(coul_free, jj, bLR);
1043                                         }
1044                                     }
1045                                     else if (!bDoCoul_i)
1046                                     {
1047                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
1048                                         {
1049                                             add_j_to_nblist(vdw_free, jj, bLR);
1050                                         }
1051                                     }
1052                                     else
1053                                     {
1054                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
1055                                         {
1056                                             if (charge[jj] != 0 || chargeB[jj] != 0)
1057                                             {
1058                                                 add_j_to_nblist(vdwc_free, jj, bLR);
1059                                             }
1060                                             else
1061                                             {
1062                                                 add_j_to_nblist(vdw_free, jj, bLR);
1063                                             }
1064                                         }
1065                                         else if (charge[jj] != 0 || chargeB[jj] != 0)
1066                                         {
1067                                             add_j_to_nblist(coul_free, jj, bLR);
1068                                         }
1069                                     }
1070                                 }
1071                                 else if (!bDoVdW_i)
1072                                 {
1073                                     /* This is done whether or not bWater is set */
1074                                     if (charge[jj] != 0)
1075                                     {
1076                                         add_j_to_nblist(coul, jj, bLR);
1077                                     }
1078                                 }
1079                                 else if (!bDoCoul_i_sol)
1080                                 {
1081                                     if (bHaveVdW[type[jj]])
1082                                     {
1083                                         add_j_to_nblist(vdw, jj, bLR);
1084                                     }
1085                                 }
1086                                 else
1087                                 {
1088                                     if (bHaveVdW[type[jj]])
1089                                     {
1090                                         if (charge[jj] != 0)
1091                                         {
1092                                             add_j_to_nblist(vdwc, jj, bLR);
1093                                         }
1094                                         else
1095                                         {
1096                                             add_j_to_nblist(vdw, jj, bLR);
1097                                         }
1098                                     }
1099                                     else if (charge[jj] != 0)
1100                                     {
1101                                         add_j_to_nblist(coul, jj, bLR);
1102                                     }
1103                                 }
1104                             }
1105                         }
1106                     }
1107                 }
1108             }
1109             close_i_nblist(vdw);
1110             close_i_nblist(coul);
1111             close_i_nblist(vdwc);
1112             close_i_nblist(vdw_free);
1113             close_i_nblist(coul_free);
1114             close_i_nblist(vdwc_free);
1115         }
1116     }
1117 }
1118
1119 static void
1120 put_in_list_adress(gmx_bool              bHaveVdW[],
1121                    int                   ngid,
1122                    t_mdatoms     *       md,
1123                    int                   icg,
1124                    int                   jgid,
1125                    int                   nj,
1126                    atom_id               jjcg[],
1127                    atom_id               index[],
1128                    t_excl                bExcl[],
1129                    int                   shift,
1130                    t_forcerec     *      fr,
1131                    gmx_bool              bLR,
1132                    gmx_bool              bDoVdW,
1133                    gmx_bool              bDoCoul,
1134                    int                   solvent_opt)
1135 {
1136     /* The a[] index has been removed,
1137      * to put it back in i_atom should be a[i0] and jj should be a[jj].
1138      */
1139     t_nblist  *   vdwc;
1140     t_nblist  *   vdw;
1141     t_nblist  *   coul;
1142     t_nblist  *   vdwc_adress  = NULL;
1143     t_nblist  *   vdw_adress   = NULL;
1144     t_nblist  *   coul_adress  = NULL;
1145     t_nblist  *   vdwc_ww      = NULL;
1146     t_nblist  *   coul_ww      = NULL;
1147
1148     int           i, j, jcg, igid, gid, nbl_ind, nbl_ind_adress;
1149     atom_id       jj, jj0, jj1, i_atom;
1150     int           i0, nicg, len;
1151
1152     int          *cginfo;
1153     int          *type, *typeB;
1154     real         *charge, *chargeB;
1155     real         *wf;
1156     real          qi, qiB, qq, rlj;
1157     gmx_bool      bFreeEnergy, bFree, bFreeJ, bNotEx, *bPert;
1158     gmx_bool      bDoVdW_i, bDoCoul_i, bDoCoul_i_sol;
1159     gmx_bool      b_hybrid;
1160     gmx_bool      j_all_atom;
1161     int           iwater, jwater;
1162     t_nblist     *nlist, *nlist_adress;
1163     gmx_bool      bEnergyGroupCG;
1164
1165     /* Copy some pointers */
1166     cginfo  = fr->cginfo;
1167     charge  = md->chargeA;
1168     chargeB = md->chargeB;
1169     type    = md->typeA;
1170     typeB   = md->typeB;
1171     bPert   = md->bPerturbed;
1172     wf      = md->wf;
1173
1174     /* Get atom range */
1175     i0     = index[icg];
1176     nicg   = index[icg+1]-i0;
1177
1178     /* Get the i charge group info */
1179     igid   = GET_CGINFO_GID(cginfo[icg]);
1180
1181     iwater = (solvent_opt != esolNO) ? GET_CGINFO_SOLOPT(cginfo[icg]) : esolNO;
1182
1183     if (md->nPerturbed)
1184     {
1185         gmx_fatal(FARGS, "AdResS does not support free energy pertubation\n");
1186     }
1187
1188     /* Unpack pointers to neighbourlist structs */
1189     if (fr->nnblists == 2)
1190     {
1191         nbl_ind        = 0;
1192         nbl_ind_adress = 1;
1193     }
1194     else
1195     {
1196         nbl_ind        = fr->gid2nblists[GID(igid, jgid, ngid)];
1197         nbl_ind_adress = nbl_ind+fr->nnblists/2;
1198     }
1199     if (bLR)
1200     {
1201         nlist        = fr->nblists[nbl_ind].nlist_lr;
1202         nlist_adress = fr->nblists[nbl_ind_adress].nlist_lr;
1203     }
1204     else
1205     {
1206         nlist        = fr->nblists[nbl_ind].nlist_sr;
1207         nlist_adress = fr->nblists[nbl_ind_adress].nlist_sr;
1208     }
1209
1210
1211     vdwc = &nlist[eNL_VDWQQ];
1212     vdw  = &nlist[eNL_VDW];
1213     coul = &nlist[eNL_QQ];
1214
1215     vdwc_adress = &nlist_adress[eNL_VDWQQ];
1216     vdw_adress  = &nlist_adress[eNL_VDW];
1217     coul_adress = &nlist_adress[eNL_QQ];
1218
1219     /* We do not support solvent optimization with AdResS for now.
1220        For this we would need hybrid solvent-other kernels */
1221
1222     /* no solvent as i charge group */
1223     /* Loop over the atoms in the i charge group */
1224     for (i = 0; i < nicg; i++)
1225     {
1226         i_atom  = i0+i;
1227         gid     = GID(igid, jgid, ngid);
1228         qi      = charge[i_atom];
1229
1230         /* Create new i_atom for each energy group */
1231         if (bDoVdW && bDoCoul)
1232         {
1233             new_i_nblist(vdwc, i_atom, shift, gid);
1234             new_i_nblist(vdwc_adress, i_atom, shift, gid);
1235
1236         }
1237         if (bDoVdW)
1238         {
1239             new_i_nblist(vdw, i_atom, shift, gid);
1240             new_i_nblist(vdw_adress, i_atom, shift, gid);
1241
1242         }
1243         if (bDoCoul)
1244         {
1245             new_i_nblist(coul, i_atom, shift, gid);
1246             new_i_nblist(coul_adress, i_atom, shift, gid);
1247         }
1248         bDoVdW_i  = (bDoVdW  && bHaveVdW[type[i_atom]]);
1249         bDoCoul_i = (bDoCoul && qi != 0);
1250
1251         /* Here we find out whether the energy groups interaction belong to a
1252          * coarse-grained (vsite) or atomistic interaction. Note that, beacuse
1253          * interactions between coarse-grained and other (atomistic) energygroups
1254          * are excluded automatically by grompp, it is sufficient to check for
1255          * the group id of atom i (igid) */
1256         bEnergyGroupCG = !egp_explicit(fr, igid);
1257
1258         if (bDoVdW_i || bDoCoul_i)
1259         {
1260             /* Loop over the j charge groups */
1261             for (j = 0; (j < nj); j++)
1262             {
1263                 jcg = jjcg[j];
1264
1265                 /* Check for large charge groups */
1266                 if (jcg == icg)
1267                 {
1268                     jj0 = i0 + i + 1;
1269                 }
1270                 else
1271                 {
1272                     jj0 = index[jcg];
1273                 }
1274
1275                 jj1 = index[jcg+1];
1276                 /* Finally loop over the atoms in the j-charge group */
1277                 for (jj = jj0; jj < jj1; jj++)
1278                 {
1279                     bNotEx = NOTEXCL(bExcl, i, jj);
1280
1281                     /* Now we have to exclude interactions which will be zero
1282                      * anyway due to the AdResS weights (in previous implementations
1283                      * this was done in the force kernel). This is necessary as
1284                      * pure interactions (those with b_hybrid=false, i.e. w_i*w_j==1 or 0)
1285                      * are put into neighbour lists which will be passed to the
1286                      * standard (optimized) kernels for speed. The interactions with
1287                      * b_hybrid=true are placed into the _adress neighbour lists and
1288                      * processed by the generic AdResS kernel.
1289                      */
1290                     if ( (bEnergyGroupCG &&
1291                           wf[i_atom] >= 1-GMX_REAL_EPS && wf[jj] >= 1-GMX_REAL_EPS ) ||
1292                          ( !bEnergyGroupCG && wf[jj] <= GMX_REAL_EPS ) )
1293                     {
1294                         continue;
1295                     }
1296
1297                     b_hybrid = !((wf[i_atom] >= 1-GMX_REAL_EPS && wf[jj] >= 1-GMX_REAL_EPS) ||
1298                                  (wf[i_atom] <= GMX_REAL_EPS && wf[jj] <= GMX_REAL_EPS));
1299
1300                     if (bNotEx)
1301                     {
1302                         if (!bDoVdW_i)
1303                         {
1304                             if (charge[jj] != 0)
1305                             {
1306                                 if (!b_hybrid)
1307                                 {
1308                                     add_j_to_nblist(coul, jj, bLR);
1309                                 }
1310                                 else
1311                                 {
1312                                     add_j_to_nblist(coul_adress, jj, bLR);
1313                                 }
1314                             }
1315                         }
1316                         else if (!bDoCoul_i)
1317                         {
1318                             if (bHaveVdW[type[jj]])
1319                             {
1320                                 if (!b_hybrid)
1321                                 {
1322                                     add_j_to_nblist(vdw, jj, bLR);
1323                                 }
1324                                 else
1325                                 {
1326                                     add_j_to_nblist(vdw_adress, jj, bLR);
1327                                 }
1328                             }
1329                         }
1330                         else
1331                         {
1332                             if (bHaveVdW[type[jj]])
1333                             {
1334                                 if (charge[jj] != 0)
1335                                 {
1336                                     if (!b_hybrid)
1337                                     {
1338                                         add_j_to_nblist(vdwc, jj, bLR);
1339                                     }
1340                                     else
1341                                     {
1342                                         add_j_to_nblist(vdwc_adress, jj, bLR);
1343                                     }
1344                                 }
1345                                 else
1346                                 {
1347                                     if (!b_hybrid)
1348                                     {
1349                                         add_j_to_nblist(vdw, jj, bLR);
1350                                     }
1351                                     else
1352                                     {
1353                                         add_j_to_nblist(vdw_adress, jj, bLR);
1354                                     }
1355
1356                                 }
1357                             }
1358                             else if (charge[jj] != 0)
1359                             {
1360                                 if (!b_hybrid)
1361                                 {
1362                                     add_j_to_nblist(coul, jj, bLR);
1363                                 }
1364                                 else
1365                                 {
1366                                     add_j_to_nblist(coul_adress, jj, bLR);
1367                                 }
1368
1369                             }
1370                         }
1371                     }
1372                 }
1373             }
1374
1375             close_i_nblist(vdw);
1376             close_i_nblist(coul);
1377             close_i_nblist(vdwc);
1378             close_i_nblist(vdw_adress);
1379             close_i_nblist(coul_adress);
1380             close_i_nblist(vdwc_adress);
1381         }
1382     }
1383 }
1384
1385 static void
1386 put_in_list_qmmm(gmx_bool gmx_unused              bHaveVdW[],
1387                  int                              ngid,
1388                  t_mdatoms gmx_unused     *       md,
1389                  int                              icg,
1390                  int                              jgid,
1391                  int                              nj,
1392                  atom_id                          jjcg[],
1393                  atom_id                          index[],
1394                  t_excl                           bExcl[],
1395                  int                              shift,
1396                  t_forcerec                *      fr,
1397                  gmx_bool                         bLR,
1398                  gmx_bool  gmx_unused             bDoVdW,
1399                  gmx_bool  gmx_unused             bDoCoul,
1400                  int       gmx_unused             solvent_opt)
1401 {
1402     t_nblist  *   coul;
1403     int           i, j, jcg, igid, gid;
1404     atom_id       jj, jj0, jj1, i_atom;
1405     int           i0, nicg;
1406     gmx_bool      bNotEx;
1407
1408     /* Get atom range */
1409     i0     = index[icg];
1410     nicg   = index[icg+1]-i0;
1411
1412     /* Get the i charge group info */
1413     igid   = GET_CGINFO_GID(fr->cginfo[icg]);
1414
1415     coul = &fr->QMMMlist;
1416
1417     /* Loop over atoms in the ith charge group */
1418     for (i = 0; i < nicg; i++)
1419     {
1420         i_atom = i0+i;
1421         gid    = GID(igid, jgid, ngid);
1422         /* Create new i_atom for each energy group */
1423         new_i_nblist(coul, i_atom, shift, gid);
1424
1425         /* Loop over the j charge groups */
1426         for (j = 0; j < nj; j++)
1427         {
1428             jcg = jjcg[j];
1429
1430             /* Charge groups cannot have QM and MM atoms simultaneously */
1431             if (jcg != icg)
1432             {
1433                 jj0 = index[jcg];
1434                 jj1 = index[jcg+1];
1435                 /* Finally loop over the atoms in the j-charge group */
1436                 for (jj = jj0; jj < jj1; jj++)
1437                 {
1438                     bNotEx = NOTEXCL(bExcl, i, jj);
1439                     if (bNotEx)
1440                     {
1441                         add_j_to_nblist(coul, jj, bLR);
1442                     }
1443                 }
1444             }
1445         }
1446         close_i_nblist(coul);
1447     }
1448 }
1449
1450 static void
1451 put_in_list_cg(gmx_bool  gmx_unused             bHaveVdW[],
1452                int                              ngid,
1453                t_mdatoms  gmx_unused    *       md,
1454                int                              icg,
1455                int                              jgid,
1456                int                              nj,
1457                atom_id                          jjcg[],
1458                atom_id                          index[],
1459                t_excl                           bExcl[],
1460                int                              shift,
1461                t_forcerec                *      fr,
1462                gmx_bool                         bLR,
1463                gmx_bool   gmx_unused            bDoVdW,
1464                gmx_bool   gmx_unused            bDoCoul,
1465                int        gmx_unused            solvent_opt)
1466 {
1467     int          cginfo;
1468     int          igid, gid, nbl_ind;
1469     t_nblist *   vdwc;
1470     int          j, jcg;
1471
1472     cginfo = fr->cginfo[icg];
1473
1474     igid = GET_CGINFO_GID(cginfo);
1475     gid  = GID(igid, jgid, ngid);
1476
1477     /* Unpack pointers to neighbourlist structs */
1478     if (fr->nnblists == 1)
1479     {
1480         nbl_ind = 0;
1481     }
1482     else
1483     {
1484         nbl_ind = fr->gid2nblists[gid];
1485     }
1486     if (bLR)
1487     {
1488         vdwc = &fr->nblists[nbl_ind].nlist_lr[eNL_VDWQQ];
1489     }
1490     else
1491     {
1492         vdwc = &fr->nblists[nbl_ind].nlist_sr[eNL_VDWQQ];
1493     }
1494
1495     /* Make a new neighbor list for charge group icg.
1496      * Currently simply one neighbor list is made with LJ and Coulomb.
1497      * If required, zero interactions could be removed here
1498      * or in the force loop.
1499      */
1500     new_i_nblist(vdwc, index[icg], shift, gid);
1501     vdwc->iinr_end[vdwc->nri] = index[icg+1];
1502
1503     for (j = 0; (j < nj); j++)
1504     {
1505         jcg = jjcg[j];
1506         /* Skip the icg-icg pairs if all self interactions are excluded */
1507         if (!(jcg == icg && GET_CGINFO_EXCL_INTRA(cginfo)))
1508         {
1509             /* Here we add the j charge group jcg to the list,
1510              * exclusions are also added to the list.
1511              */
1512             add_j_to_nblist_cg(vdwc, index[jcg], index[jcg+1], bExcl, icg == jcg, bLR);
1513         }
1514     }
1515
1516     close_i_nblist(vdwc);
1517 }
1518
1519 static void setexcl(atom_id start, atom_id end, t_blocka *excl, gmx_bool b,
1520                     t_excl bexcl[])
1521 {
1522     atom_id i, k;
1523
1524     if (b)
1525     {
1526         for (i = start; i < end; i++)
1527         {
1528             for (k = excl->index[i]; k < excl->index[i+1]; k++)
1529             {
1530                 SETEXCL(bexcl, i-start, excl->a[k]);
1531             }
1532         }
1533     }
1534     else
1535     {
1536         for (i = start; i < end; i++)
1537         {
1538             for (k = excl->index[i]; k < excl->index[i+1]; k++)
1539             {
1540                 RMEXCL(bexcl, i-start, excl->a[k]);
1541             }
1542         }
1543     }
1544 }
1545
1546 int calc_naaj(int icg, int cgtot)
1547 {
1548     int naaj;
1549
1550     if ((cgtot % 2) == 1)
1551     {
1552         /* Odd number of charge groups, easy */
1553         naaj = 1 + (cgtot/2);
1554     }
1555     else if ((cgtot % 4) == 0)
1556     {
1557         /* Multiple of four is hard */
1558         if (icg < cgtot/2)
1559         {
1560             if ((icg % 2) == 0)
1561             {
1562                 naaj = 1+(cgtot/2);
1563             }
1564             else
1565             {
1566                 naaj = cgtot/2;
1567             }
1568         }
1569         else
1570         {
1571             if ((icg % 2) == 1)
1572             {
1573                 naaj = 1+(cgtot/2);
1574             }
1575             else
1576             {
1577                 naaj = cgtot/2;
1578             }
1579         }
1580     }
1581     else
1582     {
1583         /* cgtot/2 = odd */
1584         if ((icg % 2) == 0)
1585         {
1586             naaj = 1+(cgtot/2);
1587         }
1588         else
1589         {
1590             naaj = cgtot/2;
1591         }
1592     }
1593 #ifdef DEBUG
1594     fprintf(log, "naaj=%d\n", naaj);
1595 #endif
1596
1597     return naaj;
1598 }
1599
1600 /************************************************
1601  *
1602  *  S I M P L E      C O R E     S T U F F
1603  *
1604  ************************************************/
1605
1606 static real calc_image_tric(rvec xi, rvec xj, matrix box,
1607                             rvec b_inv, int *shift)
1608 {
1609     /* This code assumes that the cut-off is smaller than
1610      * a half times the smallest diagonal element of the box.
1611      */
1612     const real h25 = 2.5;
1613     real       dx, dy, dz;
1614     real       r2;
1615     int        tx, ty, tz;
1616
1617     /* Compute diff vector */
1618     dz = xj[ZZ] - xi[ZZ];
1619     dy = xj[YY] - xi[YY];
1620     dx = xj[XX] - xi[XX];
1621
1622     /* Perform NINT operation, using trunc operation, therefore
1623      * we first add 2.5 then subtract 2 again
1624      */
1625     tz  = dz*b_inv[ZZ] + h25;
1626     tz -= 2;
1627     dz -= tz*box[ZZ][ZZ];
1628     dy -= tz*box[ZZ][YY];
1629     dx -= tz*box[ZZ][XX];
1630
1631     ty  = dy*b_inv[YY] + h25;
1632     ty -= 2;
1633     dy -= ty*box[YY][YY];
1634     dx -= ty*box[YY][XX];
1635
1636     tx  = dx*b_inv[XX]+h25;
1637     tx -= 2;
1638     dx -= tx*box[XX][XX];
1639
1640     /* Distance squared */
1641     r2 = (dx*dx) + (dy*dy) + (dz*dz);
1642
1643     *shift = XYZ2IS(tx, ty, tz);
1644
1645     return r2;
1646 }
1647
1648 static real calc_image_rect(rvec xi, rvec xj, rvec box_size,
1649                             rvec b_inv, int *shift)
1650 {
1651     const real h15 = 1.5;
1652     real       ddx, ddy, ddz;
1653     real       dx, dy, dz;
1654     real       r2;
1655     int        tx, ty, tz;
1656
1657     /* Compute diff vector */
1658     dx = xj[XX] - xi[XX];
1659     dy = xj[YY] - xi[YY];
1660     dz = xj[ZZ] - xi[ZZ];
1661
1662     /* Perform NINT operation, using trunc operation, therefore
1663      * we first add 1.5 then subtract 1 again
1664      */
1665     tx = dx*b_inv[XX] + h15;
1666     ty = dy*b_inv[YY] + h15;
1667     tz = dz*b_inv[ZZ] + h15;
1668     tx--;
1669     ty--;
1670     tz--;
1671
1672     /* Correct diff vector for translation */
1673     ddx = tx*box_size[XX] - dx;
1674     ddy = ty*box_size[YY] - dy;
1675     ddz = tz*box_size[ZZ] - dz;
1676
1677     /* Distance squared */
1678     r2 = (ddx*ddx) + (ddy*ddy) + (ddz*ddz);
1679
1680     *shift = XYZ2IS(tx, ty, tz);
1681
1682     return r2;
1683 }
1684
1685 static void add_simple(t_ns_buf *nsbuf, int nrj, atom_id cg_j,
1686                        gmx_bool bHaveVdW[], int ngid, t_mdatoms *md,
1687                        int icg, int jgid, t_block *cgs, t_excl bexcl[],
1688                        int shift, t_forcerec *fr, put_in_list_t *put_in_list)
1689 {
1690     if (nsbuf->nj + nrj > MAX_CG)
1691     {
1692         put_in_list(bHaveVdW, ngid, md, icg, jgid, nsbuf->ncg, nsbuf->jcg,
1693                     cgs->index, bexcl, shift, fr, FALSE, TRUE, TRUE, fr->solvent_opt);
1694         /* Reset buffer contents */
1695         nsbuf->ncg = nsbuf->nj = 0;
1696     }
1697     nsbuf->jcg[nsbuf->ncg++] = cg_j;
1698     nsbuf->nj               += nrj;
1699 }
1700
1701 static void ns_inner_tric(rvec x[], int icg, int *i_egp_flags,
1702                           int njcg, atom_id jcg[],
1703                           matrix box, rvec b_inv, real rcut2,
1704                           t_block *cgs, t_ns_buf **ns_buf,
1705                           gmx_bool bHaveVdW[], int ngid, t_mdatoms *md,
1706                           t_excl bexcl[], t_forcerec *fr,
1707                           put_in_list_t *put_in_list)
1708 {
1709     int       shift;
1710     int       j, nrj, jgid;
1711     int      *cginfo = fr->cginfo;
1712     atom_id   cg_j, *cgindex;
1713     t_ns_buf *nsbuf;
1714
1715     cgindex = cgs->index;
1716     shift   = CENTRAL;
1717     for (j = 0; (j < njcg); j++)
1718     {
1719         cg_j   = jcg[j];
1720         nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1721         if (calc_image_tric(x[icg], x[cg_j], box, b_inv, &shift) < rcut2)
1722         {
1723             jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1724             if (!(i_egp_flags[jgid] & EGP_EXCL))
1725             {
1726                 add_simple(&ns_buf[jgid][shift], nrj, cg_j,
1727                            bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, shift, fr,
1728                            put_in_list);
1729             }
1730         }
1731     }
1732 }
1733
1734 static void ns_inner_rect(rvec x[], int icg, int *i_egp_flags,
1735                           int njcg, atom_id jcg[],
1736                           gmx_bool bBox, rvec box_size, rvec b_inv, real rcut2,
1737                           t_block *cgs, t_ns_buf **ns_buf,
1738                           gmx_bool bHaveVdW[], int ngid, t_mdatoms *md,
1739                           t_excl bexcl[], t_forcerec *fr,
1740                           put_in_list_t *put_in_list)
1741 {
1742     int       shift;
1743     int       j, nrj, jgid;
1744     int      *cginfo = fr->cginfo;
1745     atom_id   cg_j, *cgindex;
1746     t_ns_buf *nsbuf;
1747
1748     cgindex = cgs->index;
1749     if (bBox)
1750     {
1751         shift = CENTRAL;
1752         for (j = 0; (j < njcg); j++)
1753         {
1754             cg_j   = jcg[j];
1755             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1756             if (calc_image_rect(x[icg], x[cg_j], box_size, b_inv, &shift) < rcut2)
1757             {
1758                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1759                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1760                 {
1761                     add_simple(&ns_buf[jgid][shift], nrj, cg_j,
1762                                bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, shift, fr,
1763                                put_in_list);
1764                 }
1765             }
1766         }
1767     }
1768     else
1769     {
1770         for (j = 0; (j < njcg); j++)
1771         {
1772             cg_j   = jcg[j];
1773             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1774             if ((rcut2 == 0) || (distance2(x[icg], x[cg_j]) < rcut2))
1775             {
1776                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1777                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1778                 {
1779                     add_simple(&ns_buf[jgid][CENTRAL], nrj, cg_j,
1780                                bHaveVdW, ngid, md, icg, jgid, cgs, bexcl, CENTRAL, fr,
1781                                put_in_list);
1782                 }
1783             }
1784         }
1785     }
1786 }
1787
1788 /* ns_simple_core needs to be adapted for QMMM still 2005 */
1789
1790 static int ns_simple_core(t_forcerec *fr,
1791                           gmx_localtop_t *top,
1792                           t_mdatoms *md,
1793                           matrix box, rvec box_size,
1794                           t_excl bexcl[], atom_id *aaj,
1795                           int ngid, t_ns_buf **ns_buf,
1796                           put_in_list_t *put_in_list, gmx_bool bHaveVdW[])
1797 {
1798     int          naaj, k;
1799     real         rlist2;
1800     int          nsearch, icg, jcg, igid, i0, nri, nn;
1801     int         *cginfo;
1802     t_ns_buf    *nsbuf;
1803     /* atom_id  *i_atoms; */
1804     t_block     *cgs  = &(top->cgs);
1805     t_blocka    *excl = &(top->excls);
1806     rvec         b_inv;
1807     int          m;
1808     gmx_bool     bBox, bTriclinic;
1809     int         *i_egp_flags;
1810
1811     rlist2 = sqr(fr->rlist);
1812
1813     bBox = (fr->ePBC != epbcNONE);
1814     if (bBox)
1815     {
1816         for (m = 0; (m < DIM); m++)
1817         {
1818             b_inv[m] = divide_err(1.0, box_size[m]);
1819         }
1820         bTriclinic = TRICLINIC(box);
1821     }
1822     else
1823     {
1824         bTriclinic = FALSE;
1825     }
1826
1827     cginfo = fr->cginfo;
1828
1829     nsearch = 0;
1830     for (icg = fr->cg0; (icg < fr->hcg); icg++)
1831     {
1832         /*
1833            i0        = cgs->index[icg];
1834            nri       = cgs->index[icg+1]-i0;
1835            i_atoms   = &(cgs->a[i0]);
1836            i_eg_excl = fr->eg_excl + ngid*md->cENER[*i_atoms];
1837            setexcl(nri,i_atoms,excl,TRUE,bexcl);
1838          */
1839         igid        = GET_CGINFO_GID(cginfo[icg]);
1840         i_egp_flags = fr->egp_flags + ngid*igid;
1841         setexcl(cgs->index[icg], cgs->index[icg+1], excl, TRUE, bexcl);
1842
1843         naaj = calc_naaj(icg, cgs->nr);
1844         if (bTriclinic)
1845         {
1846             ns_inner_tric(fr->cg_cm, icg, i_egp_flags, naaj, &(aaj[icg]),
1847                           box, b_inv, rlist2, cgs, ns_buf,
1848                           bHaveVdW, ngid, md, bexcl, fr, put_in_list);
1849         }
1850         else
1851         {
1852             ns_inner_rect(fr->cg_cm, icg, i_egp_flags, naaj, &(aaj[icg]),
1853                           bBox, box_size, b_inv, rlist2, cgs, ns_buf,
1854                           bHaveVdW, ngid, md, bexcl, fr, put_in_list);
1855         }
1856         nsearch += naaj;
1857
1858         for (nn = 0; (nn < ngid); nn++)
1859         {
1860             for (k = 0; (k < SHIFTS); k++)
1861             {
1862                 nsbuf = &(ns_buf[nn][k]);
1863                 if (nsbuf->ncg > 0)
1864                 {
1865                     put_in_list(bHaveVdW, ngid, md, icg, nn, nsbuf->ncg, nsbuf->jcg,
1866                                 cgs->index, bexcl, k, fr, FALSE, TRUE, TRUE, fr->solvent_opt);
1867                     nsbuf->ncg = nsbuf->nj = 0;
1868                 }
1869             }
1870         }
1871         /* setexcl(nri,i_atoms,excl,FALSE,bexcl); */
1872         setexcl(cgs->index[icg], cgs->index[icg+1], excl, FALSE, bexcl);
1873     }
1874     close_neighbor_lists(fr, FALSE);
1875
1876     return nsearch;
1877 }
1878
1879 /************************************************
1880  *
1881  *    N S 5     G R I D     S T U F F
1882  *
1883  ************************************************/
1884
1885 static gmx_inline void get_dx(int Nx, real gridx, real rc2, int xgi, real x,
1886                               int *dx0, int *dx1, real *dcx2)
1887 {
1888     real dcx, tmp;
1889     int  xgi0, xgi1, i;
1890
1891     if (xgi < 0)
1892     {
1893         *dx0 = 0;
1894         xgi0 = -1;
1895         *dx1 = -1;
1896         xgi1 = 0;
1897     }
1898     else if (xgi >= Nx)
1899     {
1900         *dx0 = Nx;
1901         xgi0 = Nx-1;
1902         *dx1 = Nx-1;
1903         xgi1 = Nx;
1904     }
1905     else
1906     {
1907         dcx2[xgi] = 0;
1908         *dx0      = xgi;
1909         xgi0      = xgi-1;
1910         *dx1      = xgi;
1911         xgi1      = xgi+1;
1912     }
1913
1914     for (i = xgi0; i >= 0; i--)
1915     {
1916         dcx = (i+1)*gridx-x;
1917         tmp = dcx*dcx;
1918         if (tmp >= rc2)
1919         {
1920             break;
1921         }
1922         *dx0    = i;
1923         dcx2[i] = tmp;
1924     }
1925     for (i = xgi1; i < Nx; i++)
1926     {
1927         dcx = i*gridx-x;
1928         tmp = dcx*dcx;
1929         if (tmp >= rc2)
1930         {
1931             break;
1932         }
1933         *dx1    = i;
1934         dcx2[i] = tmp;
1935     }
1936 }
1937
1938 static gmx_inline void get_dx_dd(int Nx, real gridx, real rc2, int xgi, real x,
1939                                  int ncpddc, int shift_min, int shift_max,
1940                                  int *g0, int *g1, real *dcx2)
1941 {
1942     real dcx, tmp;
1943     int  g_min, g_max, shift_home;
1944
1945     if (xgi < 0)
1946     {
1947         g_min = 0;
1948         g_max = Nx - 1;
1949         *g0   = 0;
1950         *g1   = -1;
1951     }
1952     else if (xgi >= Nx)
1953     {
1954         g_min = 0;
1955         g_max = Nx - 1;
1956         *g0   = Nx;
1957         *g1   = Nx - 1;
1958     }
1959     else
1960     {
1961         if (ncpddc == 0)
1962         {
1963             g_min = 0;
1964             g_max = Nx - 1;
1965         }
1966         else
1967         {
1968             if (xgi < ncpddc)
1969             {
1970                 shift_home = 0;
1971             }
1972             else
1973             {
1974                 shift_home = -1;
1975             }
1976             g_min = (shift_min == shift_home ? 0          : ncpddc);
1977             g_max = (shift_max == shift_home ? ncpddc - 1 : Nx - 1);
1978         }
1979         if (shift_min > 0)
1980         {
1981             *g0 = g_min;
1982             *g1 = g_min - 1;
1983         }
1984         else if (shift_max < 0)
1985         {
1986             *g0 = g_max + 1;
1987             *g1 = g_max;
1988         }
1989         else
1990         {
1991             *g0       = xgi;
1992             *g1       = xgi;
1993             dcx2[xgi] = 0;
1994         }
1995     }
1996
1997     while (*g0 > g_min)
1998     {
1999         /* Check one grid cell down */
2000         dcx = ((*g0 - 1) + 1)*gridx - x;
2001         tmp = dcx*dcx;
2002         if (tmp >= rc2)
2003         {
2004             break;
2005         }
2006         (*g0)--;
2007         dcx2[*g0] = tmp;
2008     }
2009
2010     while (*g1 < g_max)
2011     {
2012         /* Check one grid cell up */
2013         dcx = (*g1 + 1)*gridx - x;
2014         tmp = dcx*dcx;
2015         if (tmp >= rc2)
2016         {
2017             break;
2018         }
2019         (*g1)++;
2020         dcx2[*g1] = tmp;
2021     }
2022 }
2023
2024
2025 #define sqr(x) ((x)*(x))
2026 #define calc_dx2(XI, YI, ZI, y) (sqr(XI-y[XX]) + sqr(YI-y[YY]) + sqr(ZI-y[ZZ]))
2027 #define calc_cyl_dx2(XI, YI, y) (sqr(XI-y[XX]) + sqr(YI-y[YY]))
2028 /****************************************************
2029  *
2030  *    F A S T   N E I G H B O R  S E A R C H I N G
2031  *
2032  *    Optimized neighboursearching routine using grid
2033  *    at least 1x1x1, see GROMACS manual
2034  *
2035  ****************************************************/
2036
2037
2038 static void get_cutoff2(t_forcerec *fr, gmx_bool bDoLongRange,
2039                         real *rvdw2, real *rcoul2,
2040                         real *rs2, real *rm2, real *rl2)
2041 {
2042     *rs2 = sqr(fr->rlist);
2043
2044     if (bDoLongRange && fr->bTwinRange)
2045     {
2046         /* With plain cut-off or RF we need to make the list exactly
2047          * up to the cut-off and the cut-off's can be different,
2048          * so we can not simply set them to rlistlong.
2049          * To keep this code compatible with (exotic) old cases,
2050          * we also create lists up to rvdw/rcoulomb for PME and Ewald.
2051          * The interaction check should correspond to:
2052          * !ir_vdw/coulomb_might_be_zero_at_cutoff from inputrec.c.
2053          */
2054         if (((fr->vdwtype == evdwCUT || fr->vdwtype == evdwPME) &&
2055              fr->vdw_modifier == eintmodNONE) ||
2056             fr->rvdw <= fr->rlist)
2057         {
2058             *rvdw2  = sqr(fr->rvdw);
2059         }
2060         else
2061         {
2062             *rvdw2  = sqr(fr->rlistlong);
2063         }
2064         if (((fr->eeltype == eelCUT ||
2065               (EEL_RF(fr->eeltype) && fr->eeltype != eelRF_ZERO) ||
2066               fr->eeltype == eelPME ||
2067               fr->eeltype == eelEWALD) &&
2068              fr->coulomb_modifier == eintmodNONE) ||
2069             fr->rcoulomb <= fr->rlist)
2070         {
2071             *rcoul2 = sqr(fr->rcoulomb);
2072         }
2073         else
2074         {
2075             *rcoul2 = sqr(fr->rlistlong);
2076         }
2077     }
2078     else
2079     {
2080         /* Workaround for a gcc -O3 or -ffast-math problem */
2081         *rvdw2  = *rs2;
2082         *rcoul2 = *rs2;
2083     }
2084     *rm2 = min(*rvdw2, *rcoul2);
2085     *rl2 = max(*rvdw2, *rcoul2);
2086 }
2087
2088 static void init_nsgrid_lists(t_forcerec *fr, int ngid, gmx_ns_t *ns)
2089 {
2090     real rvdw2, rcoul2, rs2, rm2, rl2;
2091     int  j;
2092
2093     get_cutoff2(fr, TRUE, &rvdw2, &rcoul2, &rs2, &rm2, &rl2);
2094
2095     /* Short range buffers */
2096     snew(ns->nl_sr, ngid);
2097     /* Counters */
2098     snew(ns->nsr, ngid);
2099     snew(ns->nlr_ljc, ngid);
2100     snew(ns->nlr_one, ngid);
2101
2102     /* Always allocate both list types, since rcoulomb might now change with PME load balancing */
2103     /* Long range VdW and Coul buffers */
2104     snew(ns->nl_lr_ljc, ngid);
2105     /* Long range VdW or Coul only buffers */
2106     snew(ns->nl_lr_one, ngid);
2107
2108     for (j = 0; (j < ngid); j++)
2109     {
2110         snew(ns->nl_sr[j], MAX_CG);
2111         snew(ns->nl_lr_ljc[j], MAX_CG);
2112         snew(ns->nl_lr_one[j], MAX_CG);
2113     }
2114     if (debug)
2115     {
2116         fprintf(debug,
2117                 "ns5_core: rs2 = %g, rm2 = %g, rl2 = %g (nm^2)\n",
2118                 rs2, rm2, rl2);
2119     }
2120 }
2121
2122 static int nsgrid_core(t_commrec *cr, t_forcerec *fr,
2123                        matrix box, int ngid,
2124                        gmx_localtop_t *top,
2125                        t_grid *grid,
2126                        t_excl bexcl[], gmx_bool *bExcludeAlleg,
2127                        t_mdatoms *md,
2128                        put_in_list_t *put_in_list,
2129                        gmx_bool bHaveVdW[],
2130                        gmx_bool bDoLongRange, gmx_bool bMakeQMMMnblist)
2131 {
2132     gmx_ns_t     *ns;
2133     atom_id     **nl_lr_ljc, **nl_lr_one, **nl_sr;
2134     int          *nlr_ljc, *nlr_one, *nsr;
2135     gmx_domdec_t *dd     = NULL;
2136     t_block      *cgs    = &(top->cgs);
2137     int          *cginfo = fr->cginfo;
2138     /* atom_id *i_atoms,*cgsindex=cgs->index; */
2139     ivec          sh0, sh1, shp;
2140     int           cell_x, cell_y, cell_z;
2141     int           d, tx, ty, tz, dx, dy, dz, cj;
2142 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
2143     int           zsh_ty, zsh_tx, ysh_tx;
2144 #endif
2145     int           dx0, dx1, dy0, dy1, dz0, dz1;
2146     int           Nx, Ny, Nz, shift = -1, j, nrj, nns, nn = -1;
2147     real          gridx, gridy, gridz, grid_x, grid_y, grid_z;
2148     real         *dcx2, *dcy2, *dcz2;
2149     int           zgi, ygi, xgi;
2150     int           cg0, cg1, icg = -1, cgsnr, i0, igid, nri, naaj, max_jcg;
2151     int           jcg0, jcg1, jjcg, cgj0, jgid;
2152     int          *grida, *gridnra, *gridind;
2153     gmx_bool      rvdw_lt_rcoul, rcoul_lt_rvdw;
2154     rvec          xi, *cgcm, grid_offset;
2155     real          r2, rs2, rvdw2, rcoul2, rm2, rl2, XI, YI, ZI, dcx, dcy, dcz, tmp1, tmp2;
2156     int          *i_egp_flags;
2157     gmx_bool      bDomDec, bTriclinicX, bTriclinicY;
2158     ivec          ncpddc;
2159
2160     ns = &fr->ns;
2161
2162     bDomDec = DOMAINDECOMP(cr);
2163     if (bDomDec)
2164     {
2165         dd = cr->dd;
2166     }
2167
2168     bTriclinicX = ((YY < grid->npbcdim &&
2169                     (!bDomDec || dd->nc[YY] == 1) && box[YY][XX] != 0) ||
2170                    (ZZ < grid->npbcdim &&
2171                     (!bDomDec || dd->nc[ZZ] == 1) && box[ZZ][XX] != 0));
2172     bTriclinicY =  (ZZ < grid->npbcdim &&
2173                     (!bDomDec || dd->nc[ZZ] == 1) && box[ZZ][YY] != 0);
2174
2175     cgsnr    = cgs->nr;
2176
2177     get_cutoff2(fr, bDoLongRange, &rvdw2, &rcoul2, &rs2, &rm2, &rl2);
2178
2179     rvdw_lt_rcoul = (rvdw2 >= rcoul2);
2180     rcoul_lt_rvdw = (rcoul2 >= rvdw2);
2181
2182     if (bMakeQMMMnblist)
2183     {
2184         rm2 = rl2;
2185         rs2 = rl2;
2186     }
2187
2188     nl_sr     = ns->nl_sr;
2189     nsr       = ns->nsr;
2190     nl_lr_ljc = ns->nl_lr_ljc;
2191     nl_lr_one = ns->nl_lr_one;
2192     nlr_ljc   = ns->nlr_ljc;
2193     nlr_one   = ns->nlr_one;
2194
2195     /* Unpack arrays */
2196     cgcm    = fr->cg_cm;
2197     Nx      = grid->n[XX];
2198     Ny      = grid->n[YY];
2199     Nz      = grid->n[ZZ];
2200     grida   = grid->a;
2201     gridind = grid->index;
2202     gridnra = grid->nra;
2203     nns     = 0;
2204
2205     gridx      = grid->cell_size[XX];
2206     gridy      = grid->cell_size[YY];
2207     gridz      = grid->cell_size[ZZ];
2208     grid_x     = 1/gridx;
2209     grid_y     = 1/gridy;
2210     grid_z     = 1/gridz;
2211     copy_rvec(grid->cell_offset, grid_offset);
2212     copy_ivec(grid->ncpddc, ncpddc);
2213     dcx2       = grid->dcx2;
2214     dcy2       = grid->dcy2;
2215     dcz2       = grid->dcz2;
2216
2217 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
2218     zsh_ty = floor(-box[ZZ][YY]/box[YY][YY]+0.5);
2219     zsh_tx = floor(-box[ZZ][XX]/box[XX][XX]+0.5);
2220     ysh_tx = floor(-box[YY][XX]/box[XX][XX]+0.5);
2221     if (zsh_tx != 0 && ysh_tx != 0)
2222     {
2223         /* This could happen due to rounding, when both ratios are 0.5 */
2224         ysh_tx = 0;
2225     }
2226 #endif
2227
2228     debug_gmx();
2229
2230     if (fr->n_tpi)
2231     {
2232         /* We only want a list for the test particle */
2233         cg0 = cgsnr - 1;
2234     }
2235     else
2236     {
2237         cg0 = grid->icg0;
2238     }
2239     cg1 = grid->icg1;
2240
2241     /* Set the shift range */
2242     for (d = 0; d < DIM; d++)
2243     {
2244         sh0[d] = -1;
2245         sh1[d] = 1;
2246         /* Check if we need periodicity shifts.
2247          * Without PBC or with domain decomposition we don't need them.
2248          */
2249         if (d >= ePBC2npbcdim(fr->ePBC) || (bDomDec && dd->nc[d] > 1))
2250         {
2251             shp[d] = 0;
2252         }
2253         else
2254         {
2255             if (d == XX &&
2256                 box[XX][XX] - fabs(box[YY][XX]) - fabs(box[ZZ][XX]) < sqrt(rl2))
2257             {
2258                 shp[d] = 2;
2259             }
2260             else
2261             {
2262                 shp[d] = 1;
2263             }
2264         }
2265     }
2266
2267     /* Loop over charge groups */
2268     for (icg = cg0; (icg < cg1); icg++)
2269     {
2270         igid = GET_CGINFO_GID(cginfo[icg]);
2271         /* Skip this charge group if all energy groups are excluded! */
2272         if (bExcludeAlleg[igid])
2273         {
2274             continue;
2275         }
2276
2277         i0   = cgs->index[icg];
2278
2279         if (bMakeQMMMnblist)
2280         {
2281             /* Skip this charge group if it is not a QM atom while making a
2282              * QM/MM neighbourlist
2283              */
2284             if (md->bQM[i0] == FALSE)
2285             {
2286                 continue; /* MM particle, go to next particle */
2287             }
2288
2289             /* Compute the number of charge groups that fall within the control
2290              * of this one (icg)
2291              */
2292             naaj    = calc_naaj(icg, cgsnr);
2293             jcg0    = icg;
2294             jcg1    = icg + naaj;
2295             max_jcg = cgsnr;
2296         }
2297         else
2298         {
2299             /* make a normal neighbourlist */
2300
2301             if (bDomDec)
2302             {
2303                 /* Get the j charge-group and dd cell shift ranges */
2304                 dd_get_ns_ranges(cr->dd, icg, &jcg0, &jcg1, sh0, sh1);
2305                 max_jcg = 0;
2306             }
2307             else
2308             {
2309                 /* Compute the number of charge groups that fall within the control
2310                  * of this one (icg)
2311                  */
2312                 naaj = calc_naaj(icg, cgsnr);
2313                 jcg0 = icg;
2314                 jcg1 = icg + naaj;
2315
2316                 if (fr->n_tpi)
2317                 {
2318                     /* The i-particle is awlways the test particle,
2319                      * so we want all j-particles
2320                      */
2321                     max_jcg = cgsnr - 1;
2322                 }
2323                 else
2324                 {
2325                     max_jcg  = jcg1 - cgsnr;
2326                 }
2327             }
2328         }
2329
2330         i_egp_flags = fr->egp_flags + igid*ngid;
2331
2332         /* Set the exclusions for the atoms in charge group icg using a bitmask */
2333         setexcl(i0, cgs->index[icg+1], &top->excls, TRUE, bexcl);
2334
2335         ci2xyz(grid, icg, &cell_x, &cell_y, &cell_z);
2336
2337         /* Changed iicg to icg, DvdS 990115
2338          * (but see consistency check above, DvdS 990330)
2339          */
2340 #ifdef NS5DB
2341         fprintf(log, "icg=%5d, naaj=%5d, cell %d %d %d\n",
2342                 icg, naaj, cell_x, cell_y, cell_z);
2343 #endif
2344         /* Loop over shift vectors in three dimensions */
2345         for (tz = -shp[ZZ]; tz <= shp[ZZ]; tz++)
2346         {
2347             ZI = cgcm[icg][ZZ]+tz*box[ZZ][ZZ];
2348             /* Calculate range of cells in Z direction that have the shift tz */
2349             zgi = cell_z + tz*Nz;
2350 #define FAST_DD_NS
2351 #ifndef FAST_DD_NS
2352             get_dx(Nz, gridz, rl2, zgi, ZI, &dz0, &dz1, dcz2);
2353 #else
2354             get_dx_dd(Nz, gridz, rl2, zgi, ZI-grid_offset[ZZ],
2355                       ncpddc[ZZ], sh0[ZZ], sh1[ZZ], &dz0, &dz1, dcz2);
2356 #endif
2357             if (dz0 > dz1)
2358             {
2359                 continue;
2360             }
2361             for (ty = -shp[YY]; ty <= shp[YY]; ty++)
2362             {
2363                 YI = cgcm[icg][YY]+ty*box[YY][YY]+tz*box[ZZ][YY];
2364                 /* Calculate range of cells in Y direction that have the shift ty */
2365                 if (bTriclinicY)
2366                 {
2367                     ygi = (int)(Ny + (YI - grid_offset[YY])*grid_y) - Ny;
2368                 }
2369                 else
2370                 {
2371                     ygi = cell_y + ty*Ny;
2372                 }
2373 #ifndef FAST_DD_NS
2374                 get_dx(Ny, gridy, rl2, ygi, YI, &dy0, &dy1, dcy2);
2375 #else
2376                 get_dx_dd(Ny, gridy, rl2, ygi, YI-grid_offset[YY],
2377                           ncpddc[YY], sh0[YY], sh1[YY], &dy0, &dy1, dcy2);
2378 #endif
2379                 if (dy0 > dy1)
2380                 {
2381                     continue;
2382                 }
2383                 for (tx = -shp[XX]; tx <= shp[XX]; tx++)
2384                 {
2385                     XI = cgcm[icg][XX]+tx*box[XX][XX]+ty*box[YY][XX]+tz*box[ZZ][XX];
2386                     /* Calculate range of cells in X direction that have the shift tx */
2387                     if (bTriclinicX)
2388                     {
2389                         xgi = (int)(Nx + (XI - grid_offset[XX])*grid_x) - Nx;
2390                     }
2391                     else
2392                     {
2393                         xgi = cell_x + tx*Nx;
2394                     }
2395 #ifndef FAST_DD_NS
2396                     get_dx(Nx, gridx, rl2, xgi*Nx, XI, &dx0, &dx1, dcx2);
2397 #else
2398                     get_dx_dd(Nx, gridx, rl2, xgi, XI-grid_offset[XX],
2399                               ncpddc[XX], sh0[XX], sh1[XX], &dx0, &dx1, dcx2);
2400 #endif
2401                     if (dx0 > dx1)
2402                     {
2403                         continue;
2404                     }
2405                     /* Adress: an explicit cg that has a weigthing function of 0 is excluded
2406                      *  from the neigbour list as it will not interact  */
2407                     if (fr->adress_type != eAdressOff)
2408                     {
2409                         if (md->wf[cgs->index[icg]] <= GMX_REAL_EPS && egp_explicit(fr, igid))
2410                         {
2411                             continue;
2412                         }
2413                     }
2414                     /* Get shift vector */
2415                     shift = XYZ2IS(tx, ty, tz);
2416 #ifdef NS5DB
2417                     range_check(shift, 0, SHIFTS);
2418 #endif
2419                     for (nn = 0; (nn < ngid); nn++)
2420                     {
2421                         nsr[nn]      = 0;
2422                         nlr_ljc[nn]  = 0;
2423                         nlr_one[nn]  = 0;
2424                     }
2425 #ifdef NS5DB
2426                     fprintf(log, "shift: %2d, dx0,1: %2d,%2d, dy0,1: %2d,%2d, dz0,1: %2d,%2d\n",
2427                             shift, dx0, dx1, dy0, dy1, dz0, dz1);
2428                     fprintf(log, "cgcm: %8.3f  %8.3f  %8.3f\n", cgcm[icg][XX],
2429                             cgcm[icg][YY], cgcm[icg][ZZ]);
2430                     fprintf(log, "xi:   %8.3f  %8.3f  %8.3f\n", XI, YI, ZI);
2431 #endif
2432                     for (dx = dx0; (dx <= dx1); dx++)
2433                     {
2434                         tmp1 = rl2 - dcx2[dx];
2435                         for (dy = dy0; (dy <= dy1); dy++)
2436                         {
2437                             tmp2 = tmp1 - dcy2[dy];
2438                             if (tmp2 > 0)
2439                             {
2440                                 for (dz = dz0; (dz <= dz1); dz++)
2441                                 {
2442                                     if (tmp2 > dcz2[dz])
2443                                     {
2444                                         /* Find grid-cell cj in which possible neighbours are */
2445                                         cj   = xyz2ci(Ny, Nz, dx, dy, dz);
2446
2447                                         /* Check out how many cgs (nrj) there in this cell */
2448                                         nrj  = gridnra[cj];
2449
2450                                         /* Find the offset in the cg list */
2451                                         cgj0 = gridind[cj];
2452
2453                                         /* Check if all j's are out of range so we
2454                                          * can skip the whole cell.
2455                                          * Should save some time, especially with DD.
2456                                          */
2457                                         if (nrj == 0 ||
2458                                             (grida[cgj0] >= max_jcg &&
2459                                              (grida[cgj0] >= jcg1 || grida[cgj0+nrj-1] < jcg0)))
2460                                         {
2461                                             continue;
2462                                         }
2463
2464                                         /* Loop over cgs */
2465                                         for (j = 0; (j < nrj); j++)
2466                                         {
2467                                             jjcg = grida[cgj0+j];
2468
2469                                             /* check whether this guy is in range! */
2470                                             if ((jjcg >= jcg0 && jjcg < jcg1) ||
2471                                                 (jjcg < max_jcg))
2472                                             {
2473                                                 r2 = calc_dx2(XI, YI, ZI, cgcm[jjcg]);
2474                                                 if (r2 < rl2)
2475                                                 {
2476                                                     /* jgid = gid[cgsatoms[cgsindex[jjcg]]]; */
2477                                                     jgid = GET_CGINFO_GID(cginfo[jjcg]);
2478                                                     /* check energy group exclusions */
2479                                                     if (!(i_egp_flags[jgid] & EGP_EXCL))
2480                                                     {
2481                                                         if (r2 < rs2)
2482                                                         {
2483                                                             if (nsr[jgid] >= MAX_CG)
2484                                                             {
2485                                                                 /* Add to short-range list */
2486                                                                 put_in_list(bHaveVdW, ngid, md, icg, jgid,
2487                                                                             nsr[jgid], nl_sr[jgid],
2488                                                                             cgs->index, /* cgsatoms, */ bexcl,
2489                                                                             shift, fr, FALSE, TRUE, TRUE, fr->solvent_opt);
2490                                                                 nsr[jgid] = 0;
2491                                                             }
2492                                                             nl_sr[jgid][nsr[jgid]++] = jjcg;
2493                                                         }
2494                                                         else if (r2 < rm2)
2495                                                         {
2496                                                             if (nlr_ljc[jgid] >= MAX_CG)
2497                                                             {
2498                                                                 /* Add to LJ+coulomb long-range list */
2499                                                                 put_in_list(bHaveVdW, ngid, md, icg, jgid,
2500                                                                             nlr_ljc[jgid], nl_lr_ljc[jgid], top->cgs.index,
2501                                                                             bexcl, shift, fr, TRUE, TRUE, TRUE, fr->solvent_opt);
2502                                                                 nlr_ljc[jgid] = 0;
2503                                                             }
2504                                                             nl_lr_ljc[jgid][nlr_ljc[jgid]++] = jjcg;
2505                                                         }
2506                                                         else
2507                                                         {
2508                                                             if (nlr_one[jgid] >= MAX_CG)
2509                                                             {
2510                                                                 /* Add to long-range list with only coul, or only LJ */
2511                                                                 put_in_list(bHaveVdW, ngid, md, icg, jgid,
2512                                                                             nlr_one[jgid], nl_lr_one[jgid], top->cgs.index,
2513                                                                             bexcl, shift, fr, TRUE, rvdw_lt_rcoul, rcoul_lt_rvdw, fr->solvent_opt);
2514                                                                 nlr_one[jgid] = 0;
2515                                                             }
2516                                                             nl_lr_one[jgid][nlr_one[jgid]++] = jjcg;
2517                                                         }
2518                                                     }
2519                                                 }
2520                                                 nns++;
2521                                             }
2522                                         }
2523                                     }
2524                                 }
2525                             }
2526                         }
2527                     }
2528                     /* CHECK whether there is anything left in the buffers */
2529                     for (nn = 0; (nn < ngid); nn++)
2530                     {
2531                         if (nsr[nn] > 0)
2532                         {
2533                             put_in_list(bHaveVdW, ngid, md, icg, nn, nsr[nn], nl_sr[nn],
2534                                         cgs->index, /* cgsatoms, */ bexcl,
2535                                         shift, fr, FALSE, TRUE, TRUE, fr->solvent_opt);
2536                         }
2537
2538                         if (nlr_ljc[nn] > 0)
2539                         {
2540                             put_in_list(bHaveVdW, ngid, md, icg, nn, nlr_ljc[nn],
2541                                         nl_lr_ljc[nn], top->cgs.index,
2542                                         bexcl, shift, fr, TRUE, TRUE, TRUE, fr->solvent_opt);
2543                         }
2544
2545                         if (nlr_one[nn] > 0)
2546                         {
2547                             put_in_list(bHaveVdW, ngid, md, icg, nn, nlr_one[nn],
2548                                         nl_lr_one[nn], top->cgs.index,
2549                                         bexcl, shift, fr, TRUE, rvdw_lt_rcoul, rcoul_lt_rvdw, fr->solvent_opt);
2550                         }
2551                     }
2552                 }
2553             }
2554         }
2555         /* setexcl(nri,i_atoms,&top->atoms.excl,FALSE,bexcl); */
2556         setexcl(cgs->index[icg], cgs->index[icg+1], &top->excls, FALSE, bexcl);
2557     }
2558     /* No need to perform any left-over force calculations anymore (as we used to do here)
2559      * since we now save the proper long-range lists for later evaluation.
2560      */
2561
2562     debug_gmx();
2563
2564     /* Close neighbourlists */
2565     close_neighbor_lists(fr, bMakeQMMMnblist);
2566
2567     return nns;
2568 }
2569
2570 void ns_realloc_natoms(gmx_ns_t *ns, int natoms)
2571 {
2572     int i;
2573
2574     if (natoms > ns->nra_alloc)
2575     {
2576         ns->nra_alloc = over_alloc_dd(natoms);
2577         srenew(ns->bexcl, ns->nra_alloc);
2578         for (i = 0; i < ns->nra_alloc; i++)
2579         {
2580             ns->bexcl[i] = 0;
2581         }
2582     }
2583 }
2584
2585 void init_ns(FILE *fplog, const t_commrec *cr,
2586              gmx_ns_t *ns, t_forcerec *fr,
2587              const gmx_mtop_t *mtop)
2588 {
2589     int  mt, icg, nr_in_cg, maxcg, i, j, jcg, ngid, ncg;
2590     t_block *cgs;
2591     char *ptr;
2592
2593     /* Compute largest charge groups size (# atoms) */
2594     nr_in_cg = 1;
2595     for (mt = 0; mt < mtop->nmoltype; mt++)
2596     {
2597         cgs = &mtop->moltype[mt].cgs;
2598         for (icg = 0; (icg < cgs->nr); icg++)
2599         {
2600             nr_in_cg = max(nr_in_cg, (int)(cgs->index[icg+1]-cgs->index[icg]));
2601         }
2602     }
2603
2604     /* Verify whether largest charge group is <= max cg.
2605      * This is determined by the type of the local exclusion type
2606      * Exclusions are stored in bits. (If the type is not large
2607      * enough, enlarge it, unsigned char -> unsigned short -> unsigned long)
2608      */
2609     maxcg = sizeof(t_excl)*8;
2610     if (nr_in_cg > maxcg)
2611     {
2612         gmx_fatal(FARGS, "Max #atoms in a charge group: %d > %d\n",
2613                   nr_in_cg, maxcg);
2614     }
2615
2616     ngid = mtop->groups.grps[egcENER].nr;
2617     snew(ns->bExcludeAlleg, ngid);
2618     for (i = 0; i < ngid; i++)
2619     {
2620         ns->bExcludeAlleg[i] = TRUE;
2621         for (j = 0; j < ngid; j++)
2622         {
2623             if (!(fr->egp_flags[i*ngid+j] & EGP_EXCL))
2624             {
2625                 ns->bExcludeAlleg[i] = FALSE;
2626             }
2627         }
2628     }
2629
2630     if (fr->bGrid)
2631     {
2632         /* Grid search */
2633         ns->grid = init_grid(fplog, fr);
2634         init_nsgrid_lists(fr, ngid, ns);
2635     }
2636     else
2637     {
2638         /* Simple search */
2639         snew(ns->ns_buf, ngid);
2640         for (i = 0; (i < ngid); i++)
2641         {
2642             snew(ns->ns_buf[i], SHIFTS);
2643         }
2644         ncg = ncg_mtop(mtop);
2645         snew(ns->simple_aaj, 2*ncg);
2646         for (jcg = 0; (jcg < ncg); jcg++)
2647         {
2648             ns->simple_aaj[jcg]     = jcg;
2649             ns->simple_aaj[jcg+ncg] = jcg;
2650         }
2651     }
2652
2653     /* Create array that determines whether or not atoms have VdW */
2654     snew(ns->bHaveVdW, fr->ntype);
2655     for (i = 0; (i < fr->ntype); i++)
2656     {
2657         for (j = 0; (j < fr->ntype); j++)
2658         {
2659             ns->bHaveVdW[i] = (ns->bHaveVdW[i] ||
2660                                (fr->bBHAM ?
2661                                 ((BHAMA(fr->nbfp, fr->ntype, i, j) != 0) ||
2662                                  (BHAMB(fr->nbfp, fr->ntype, i, j) != 0) ||
2663                                  (BHAMC(fr->nbfp, fr->ntype, i, j) != 0)) :
2664                                 ((C6(fr->nbfp, fr->ntype, i, j) != 0) ||
2665                                  (C12(fr->nbfp, fr->ntype, i, j) != 0))));
2666         }
2667     }
2668     if (debug)
2669     {
2670         pr_bvec(debug, 0, "bHaveVdW", ns->bHaveVdW, fr->ntype, TRUE);
2671     }
2672
2673     ns->nra_alloc = 0;
2674     ns->bexcl     = NULL;
2675     if (!DOMAINDECOMP(cr))
2676     {
2677         ns_realloc_natoms(ns, mtop->natoms);
2678     }
2679
2680     ns->nblist_initialized = FALSE;
2681
2682     /* nbr list debug dump */
2683     {
2684         char *ptr = getenv("GMX_DUMP_NL");
2685         if (ptr)
2686         {
2687             ns->dump_nl = strtol(ptr, NULL, 10);
2688             if (fplog)
2689             {
2690                 fprintf(fplog, "GMX_DUMP_NL = %d", ns->dump_nl);
2691             }
2692         }
2693         else
2694         {
2695             ns->dump_nl = 0;
2696         }
2697     }
2698 }
2699
2700
2701 int search_neighbours(FILE *log, t_forcerec *fr,
2702                       matrix box,
2703                       gmx_localtop_t *top,
2704                       gmx_groups_t *groups,
2705                       t_commrec *cr,
2706                       t_nrnb *nrnb, t_mdatoms *md,
2707                       gmx_bool bFillGrid,
2708                       gmx_bool bDoLongRangeNS)
2709 {
2710     t_block  *cgs = &(top->cgs);
2711     rvec     box_size, grid_x0, grid_x1;
2712     int      i, j, m, ngid;
2713     real     min_size, grid_dens;
2714     int      nsearch;
2715     gmx_bool     bGrid;
2716     char     *ptr;
2717     gmx_bool     *i_egp_flags;
2718     int      cg_start, cg_end, start, end;
2719     gmx_ns_t *ns;
2720     t_grid   *grid;
2721     gmx_domdec_zones_t *dd_zones;
2722     put_in_list_t *put_in_list;
2723
2724     ns = &fr->ns;
2725
2726     /* Set some local variables */
2727     bGrid = fr->bGrid;
2728     ngid  = groups->grps[egcENER].nr;
2729
2730     for (m = 0; (m < DIM); m++)
2731     {
2732         box_size[m] = box[m][m];
2733     }
2734
2735     if (fr->ePBC != epbcNONE)
2736     {
2737         if (sqr(fr->rlistlong) >= max_cutoff2(fr->ePBC, box))
2738         {
2739             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.");
2740         }
2741         if (!bGrid)
2742         {
2743             min_size = min(box_size[XX], min(box_size[YY], box_size[ZZ]));
2744             if (2*fr->rlistlong >= min_size)
2745             {
2746                 gmx_fatal(FARGS, "One of the box diagonal elements has become smaller than twice the cut-off length.");
2747             }
2748         }
2749     }
2750
2751     if (DOMAINDECOMP(cr))
2752     {
2753         ns_realloc_natoms(ns, cgs->index[cgs->nr]);
2754     }
2755     debug_gmx();
2756
2757     /* Reset the neighbourlists */
2758     reset_neighbor_lists(fr, TRUE, TRUE);
2759
2760     if (bGrid && bFillGrid)
2761     {
2762
2763         grid = ns->grid;
2764         if (DOMAINDECOMP(cr))
2765         {
2766             dd_zones = domdec_zones(cr->dd);
2767         }
2768         else
2769         {
2770             dd_zones = NULL;
2771
2772             get_nsgrid_boundaries(grid->nboundeddim, box, NULL, NULL, NULL, NULL,
2773                                   cgs->nr, fr->cg_cm, grid_x0, grid_x1, &grid_dens);
2774
2775             grid_first(log, grid, NULL, NULL, box, grid_x0, grid_x1,
2776                        fr->rlistlong, grid_dens);
2777         }
2778         debug_gmx();
2779
2780         start = 0;
2781         end   = cgs->nr;
2782
2783         if (DOMAINDECOMP(cr))
2784         {
2785             end = cgs->nr;
2786             fill_grid(dd_zones, grid, end, -1, end, fr->cg_cm);
2787             grid->icg0 = 0;
2788             grid->icg1 = dd_zones->izone[dd_zones->nizone-1].cg1;
2789         }
2790         else
2791         {
2792             fill_grid(NULL, grid, cgs->nr, fr->cg0, fr->hcg, fr->cg_cm);
2793             grid->icg0 = fr->cg0;
2794             grid->icg1 = fr->hcg;
2795             debug_gmx();
2796         }
2797
2798         calc_elemnr(grid, start, end, cgs->nr);
2799         calc_ptrs(grid);
2800         grid_last(grid, start, end, cgs->nr);
2801
2802         if (gmx_debug_at)
2803         {
2804             check_grid(grid);
2805             print_grid(debug, grid);
2806         }
2807     }
2808     else if (fr->n_tpi)
2809     {
2810         /* Set the grid cell index for the test particle only.
2811          * The cell to cg index is not corrected, but that does not matter.
2812          */
2813         fill_grid(NULL, ns->grid, fr->hcg, fr->hcg-1, fr->hcg, fr->cg_cm);
2814     }
2815     debug_gmx();
2816
2817     if (fr->adress_type == eAdressOff)
2818     {
2819         if (!fr->ns.bCGlist)
2820         {
2821             put_in_list = put_in_list_at;
2822         }
2823         else
2824         {
2825             put_in_list = put_in_list_cg;
2826         }
2827     }
2828     else
2829     {
2830         put_in_list = put_in_list_adress;
2831     }
2832
2833     /* Do the core! */
2834     if (bGrid)
2835     {
2836         grid    = ns->grid;
2837         nsearch = nsgrid_core(cr, fr, box, ngid, top,
2838                               grid, ns->bexcl, ns->bExcludeAlleg,
2839                               md, put_in_list, ns->bHaveVdW,
2840                               bDoLongRangeNS, FALSE);
2841
2842         /* neighbour searching withouth QMMM! QM atoms have zero charge in
2843          * the classical calculation. The charge-charge interaction
2844          * between QM and MM atoms is handled in the QMMM core calculation
2845          * (see QMMM.c). The VDW however, we'd like to compute classically
2846          * and the QM MM atom pairs have just been put in the
2847          * corresponding neighbourlists. in case of QMMM we still need to
2848          * fill a special QMMM neighbourlist that contains all neighbours
2849          * of the QM atoms. If bQMMM is true, this list will now be made:
2850          */
2851         if (fr->bQMMM && fr->qr->QMMMscheme != eQMMMschemeoniom)
2852         {
2853             nsearch += nsgrid_core(cr, fr, box, ngid, top,
2854                                    grid, ns->bexcl, ns->bExcludeAlleg,
2855                                    md, put_in_list_qmmm, ns->bHaveVdW,
2856                                    bDoLongRangeNS, TRUE);
2857         }
2858     }
2859     else
2860     {
2861         nsearch = ns_simple_core(fr, top, md, box, box_size,
2862                                  ns->bexcl, ns->simple_aaj,
2863                                  ngid, ns->ns_buf, put_in_list, ns->bHaveVdW);
2864     }
2865     debug_gmx();
2866
2867 #ifdef DEBUG
2868     pr_nsblock(log);
2869 #endif
2870
2871     inc_nrnb(nrnb, eNR_NS, nsearch);
2872     /* inc_nrnb(nrnb,eNR_LR,fr->nlr); */
2873
2874     return nsearch;
2875 }
2876
2877 int natoms_beyond_ns_buffer(t_inputrec *ir, t_forcerec *fr, t_block *cgs,
2878                             matrix scale_tot, rvec *x)
2879 {
2880     int  cg0, cg1, cg, a0, a1, a, i, j;
2881     real rint, hbuf2, scale;
2882     rvec *cg_cm, cgsc;
2883     gmx_bool bIsotropic;
2884     int  nBeyond;
2885
2886     nBeyond = 0;
2887
2888     rint = max(ir->rcoulomb, ir->rvdw);
2889     if (ir->rlist < rint)
2890     {
2891         gmx_fatal(FARGS, "The neighbor search buffer has negative size: %f nm",
2892                   ir->rlist - rint);
2893     }
2894     cg_cm = fr->cg_cm;
2895
2896     cg0 = fr->cg0;
2897     cg1 = fr->hcg;
2898
2899     if (!EI_DYNAMICS(ir->eI) || !DYNAMIC_BOX(*ir))
2900     {
2901         hbuf2 = sqr(0.5*(ir->rlist - rint));
2902         for (cg = cg0; cg < cg1; cg++)
2903         {
2904             a0 = cgs->index[cg];
2905             a1 = cgs->index[cg+1];
2906             for (a = a0; a < a1; a++)
2907             {
2908                 if (distance2(cg_cm[cg], x[a]) > hbuf2)
2909                 {
2910                     nBeyond++;
2911                 }
2912             }
2913         }
2914     }
2915     else
2916     {
2917         bIsotropic = TRUE;
2918         scale      = scale_tot[0][0];
2919         for (i = 1; i < DIM; i++)
2920         {
2921             /* With anisotropic scaling, the original spherical ns volumes become
2922              * ellipsoids. To avoid costly transformations we use the minimum
2923              * eigenvalue of the scaling matrix for determining the buffer size.
2924              * Since the lower half is 0, the eigenvalues are the diagonal elements.
2925              */
2926             scale = min(scale, scale_tot[i][i]);
2927             if (scale_tot[i][i] != scale_tot[i-1][i-1])
2928             {
2929                 bIsotropic = FALSE;
2930             }
2931             for (j = 0; j < i; j++)
2932             {
2933                 if (scale_tot[i][j] != 0)
2934                 {
2935                     bIsotropic = FALSE;
2936                 }
2937             }
2938         }
2939         hbuf2 = sqr(0.5*(scale*ir->rlist - rint));
2940         if (bIsotropic)
2941         {
2942             for (cg = cg0; cg < cg1; cg++)
2943             {
2944                 svmul(scale, cg_cm[cg], cgsc);
2945                 a0 = cgs->index[cg];
2946                 a1 = cgs->index[cg+1];
2947                 for (a = a0; a < a1; a++)
2948                 {
2949                     if (distance2(cgsc, x[a]) > hbuf2)
2950                     {
2951                         nBeyond++;
2952                     }
2953                 }
2954             }
2955         }
2956         else
2957         {
2958             /* Anistropic scaling */
2959             for (cg = cg0; cg < cg1; cg++)
2960             {
2961                 /* Since scale_tot contains the transpose of the scaling matrix,
2962                  * we need to multiply with the transpose.
2963                  */
2964                 tmvmul_ur0(scale_tot, cg_cm[cg], cgsc);
2965                 a0 = cgs->index[cg];
2966                 a1 = cgs->index[cg+1];
2967                 for (a = a0; a < a1; a++)
2968                 {
2969                     if (distance2(cgsc, x[a]) > hbuf2)
2970                     {
2971                         nBeyond++;
2972                     }
2973                 }
2974             }
2975         }
2976     }
2977
2978     return nBeyond;
2979 }