src/gromacs/mdlib/ns.c

   1 /* -*- mode: c; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; c-file-style: "stroustrup"; -*-
   2  *
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  *          GROningen MAchine for Chemical Simulations
   9  *
  10  *                        VERSION 3.2.0
  11  * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
  12  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
  13  * Copyright (c) 2001-2004, The GROMACS development team,
  14  * check out http://www.gromacs.org for more information.
  15
  16  * This program is free software; you can redistribute it and/or
  17  * modify it under the terms of the GNU General Public License
  18  * as published by the Free Software Foundation; either version 2
  19  * of the License, or (at your option) any later version.
  20  *
  21  * If you want to redistribute modifications, please consider that
  22  * scientific software is very special. Version control is crucial -
  23  * bugs must be traceable. We will be happy to consider code for
  24  * inclusion in the official distribution, but derived work must not
  25  * be called official GROMACS. Details are found in the README & COPYING
  26  * files - if they are missing, get the official version at www.gromacs.org.
  27  *
  28  * To help us fund GROMACS development, we humbly ask that you cite
  29  * the papers on the package - you can find them in the top README file.
  30  *
  31  * For more info, check our website at http://www.gromacs.org
  32  *
  33  * And Hey:
  34  * GROwing Monsters And Cloning Shrimps
  35  */
  36 #ifdef HAVE_CONFIG_H
  37 #include <config.h>
  38 #endif
  39
  40 #ifdef GMX_THREAD_SHM_FDECOMP
  41 #include <pthread.h>
  42 #endif
  43
  44 #include <math.h>
  45 #include <string.h>
  46 #include "sysstuff.h"
  47 #include "smalloc.h"
  48 #include "macros.h"
  49 #include "maths.h"
  50 #include "vec.h"
  51 #include "network.h"
  52 #include "nsgrid.h"
  53 #include "force.h"
  54 #include "nonbonded.h"
  55 #include "ns.h"
  56 #include "pbc.h"
  57 #include "names.h"
  58 #include "gmx_fatal.h"
  59 #include "nrnb.h"
  60 #include "txtdump.h"
  61 #include "mtop_util.h"
  62
  63 #include "domdec.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 {   e[j] = e[j] | (1<<i); }
  73 static void RMEXCL_(t_excl e[],atom_id i,atom_id j)
  74 { e[j]=e[j] & ~(1<<i); }
  75 static gmx_bool ISEXCL_(t_excl e[],atom_id i,atom_id j)
  76 { return (gmx_bool)(e[j] & (1<<i)); }
  77 static gmx_bool NOTEXCL_(t_excl e[],atom_id i,atom_id j)
  78 {  return !(ISEXCL(e,i,j)); }
  79 #else
  80 #define SETEXCL(e,i,j) (e)[((atom_id) (j))] |= (1<<((atom_id) (i)))
  81 #define RMEXCL(e,i,j)  (e)[((atom_id) (j))] &= (~(1<<((atom_id) (i))))
  82 #define ISEXCL(e,i,j)  (gmx_bool) ((e)[((atom_id) (j))] & (1<<((atom_id) (i))))
  83 #define NOTEXCL(e,i,j) !(ISEXCL(e,i,j))
  84 #endif
  85
  86 /************************************************
  87  *
  88  *  U T I L I T I E S    F O R    N S
  89  *
  90  ************************************************/
  91
  92 static void reallocate_nblist(t_nblist *nl)
  93 {
  94     if (gmx_debug_at)
  95     {
  96         fprintf(debug,"reallocating neigborlist il_code=%d, maxnri=%d\n",
  97                 nl->il_code,nl->maxnri);
  98     }
  99     srenew(nl->iinr,   nl->maxnri);
 100     if (nl->enlist == enlistCG_CG)
 101     {
 102         srenew(nl->iinr_end,nl->maxnri);
 103     }
 104     srenew(nl->gid,    nl->maxnri);
 105     srenew(nl->shift,  nl->maxnri);
 106     srenew(nl->jindex, nl->maxnri+1);
 107 }
 108
 109 /* ivdw/icoul are used to determine the type of interaction, so we
 110  * can set an innerloop index here. The obvious choice for this would have
 111  * been the vdwtype/coultype values in the forcerecord, but unfortunately
 112  * those types are braindead - for instance both Buckingham and normal
 113  * Lennard-Jones use the same value (evdwCUT), and a separate gmx_boolean variable
 114  * to determine which interaction is used. There is further no special value
 115  * for 'no interaction'. For backward compatibility with old TPR files we won't
 116  * change this in the 3.x series, so when calling this routine you should use:
 117  *
 118  * icoul=0 no coulomb interaction
 119  * icoul=1 cutoff standard coulomb
 120  * icoul=2 reaction-field coulomb
 121  * icoul=3 tabulated coulomb
 122  *
 123  * ivdw=0 no vdw interaction
 124  * ivdw=1 standard L-J interaction
 125  * ivdw=2 Buckingham
 126  * ivdw=3 tabulated vdw.
 127  *
 128  * Kind of ugly, but it works.
 129  */
 130 static void init_nblist(t_nblist *nl_sr,t_nblist *nl_lr,
 131                         int maxsr,int maxlr,
 132                         int ivdw, int icoul,
 133                         gmx_bool bfree, int enlist)
 134 {
 135     t_nblist *nl;
 136     int      homenr;
 137     int      i,nn;
 138
 139     int inloop[20] =
 140     {
 141         eNR_NBKERNEL_NONE,
 142         eNR_NBKERNEL010,
 143         eNR_NBKERNEL020,
 144         eNR_NBKERNEL030,
 145         eNR_NBKERNEL100,
 146         eNR_NBKERNEL110,
 147         eNR_NBKERNEL120,
 148         eNR_NBKERNEL130,
 149         eNR_NBKERNEL200,
 150         eNR_NBKERNEL210,
 151         eNR_NBKERNEL220,
 152         eNR_NBKERNEL230,
 153         eNR_NBKERNEL300,
 154         eNR_NBKERNEL310,
 155         eNR_NBKERNEL320,
 156         eNR_NBKERNEL330,
 157         eNR_NBKERNEL400,
 158         eNR_NBKERNEL410,
 159         eNR_NBKERNEL_NONE,
 160         eNR_NBKERNEL430
 161     };
 162
 163     for(i=0; (i<2); i++)
 164     {
 165         nl     = (i == 0) ? nl_sr : nl_lr;
 166         homenr = (i == 0) ? maxsr : maxlr;
 167
 168         if (nl == NULL)
 169         {
 170             continue;
 171         }
 172
 173         /* Set coul/vdw in neighborlist, and for the normal loops we determine
 174          * an index of which one to call.
 175          */
 176         nl->ivdw  = ivdw;
 177         nl->icoul = icoul;
 178         nl->free_energy = bfree;
 179
 180         if (bfree)
 181         {
 182             nl->enlist  = enlistATOM_ATOM;
 183             nl->il_code = eNR_NBKERNEL_FREE_ENERGY;
 184         }
 185         else
 186         {
 187             nl->enlist = enlist;
 188
 189             nn = inloop[4*icoul + ivdw];
 190
 191             /* solvent loops follow directly after the corresponding
 192             * ordinary loops, in the order:
 193             *
 194             * SPC, SPC-SPC, TIP4p, TIP4p-TIP4p
 195             *
 196             */
 197             switch (enlist) {
 198             case enlistATOM_ATOM:
 199             case enlistCG_CG:
 200                 break;
 201             case enlistSPC_ATOM:     nn += 1; break;
 202             case enlistSPC_SPC:      nn += 2; break;
 203             case enlistTIP4P_ATOM:   nn += 3; break;
 204             case enlistTIP4P_TIP4P:  nn += 4; break;
 205             }
 206
 207             nl->il_code = nn;
 208         }
 209
 210         if (debug)
 211             fprintf(debug,"Initiating neighbourlist type %d for %s interactions,\nwith %d SR, %d LR atoms.\n",
 212                     nl->il_code,ENLISTTYPE(enlist),maxsr,maxlr);
 213
 214         /* maxnri is influenced by the number of shifts (maximum is 8)
 215          * and the number of energy groups.
 216          * If it is not enough, nl memory will be reallocated during the run.
 217          * 4 seems to be a reasonable factor, which only causes reallocation
 218          * during runs with tiny and many energygroups.
 219          */
 220         nl->maxnri      = homenr*4;
 221         nl->maxnrj      = 0;
 222         nl->maxlen      = 0;
 223         nl->nri         = -1;
 224         nl->nrj         = 0;
 225         nl->iinr        = NULL;
 226         nl->gid         = NULL;
 227         nl->shift       = NULL;
 228         nl->jindex      = NULL;
 229         reallocate_nblist(nl);
 230         nl->jindex[0] = 0;
 231 #ifdef GMX_THREAD_SHM_FDECOMP
 232         nl->counter = 0;
 233         snew(nl->mtx,1);
 234         pthread_mutex_init(nl->mtx,NULL);
 235 #endif
 236     }
 237 }
 238
 239 void init_neighbor_list(FILE *log,t_forcerec *fr,int homenr)
 240 {
 241    /* Make maxlr tunable! (does not seem to be a big difference though)
 242     * This parameter determines the number of i particles in a long range
 243     * neighbourlist. Too few means many function calls, too many means
 244     * cache trashing.
 245     */
 246    int maxsr,maxsr_wat,maxlr,maxlr_wat;
 247    int icoul,icoulf,ivdw;
 248    int solvent;
 249    int enlist_def,enlist_w,enlist_ww;
 250    int i;
 251    t_nblists *nbl;
 252
 253    /* maxsr     = homenr-fr->nWatMol*3; */
 254    maxsr     = homenr;
 255
 256    if (maxsr < 0)
 257    {
 258      gmx_fatal(FARGS,"%s, %d: Negative number of short range atoms.\n"
 259                  "Call your Gromacs dealer for assistance.",__FILE__,__LINE__);
 260    }
 261    /* This is just for initial allocation, so we do not reallocate
 262     * all the nlist arrays many times in a row.
 263     * The numbers seem very accurate, but they are uncritical.
 264     */
 265    maxsr_wat = min(fr->nWatMol,(homenr+2)/3);
 266    if (fr->bTwinRange)
 267    {
 268        maxlr     = 50;
 269        maxlr_wat = min(maxsr_wat,maxlr);
 270    }
 271    else
 272    {
 273      maxlr = maxlr_wat = 0;
 274    }
 275
 276    /* Determine the values for icoul/ivdw. */
 277    /* Start with GB */
 278    if(fr->bGB)
 279    {
 280        icoul=4;
 281    }
 282    else if (fr->bcoultab)
 283    {
 284        icoul = 3;
 285    }
 286    else if (EEL_RF(fr->eeltype))
 287    {
 288        icoul = 2;
 289    }
 290    else
 291    {
 292        icoul = 1;
 293    }
 294
 295    if (fr->bvdwtab)
 296    {
 297        ivdw = 3;
 298    }
 299    else if (fr->bBHAM)
 300    {
 301        ivdw = 2;
 302    }
 303    else
 304    {
 305        ivdw = 1;
 306    }
 307
 308    fr->ns.bCGlist = (getenv("GMX_NBLISTCG") != 0);
 309    if (!fr->ns.bCGlist)
 310    {
 311        enlist_def = enlistATOM_ATOM;
 312    }
 313    else
 314    {
 315        enlist_def = enlistCG_CG;
 316        if (log != NULL)
 317        {
 318            fprintf(log,"\nUsing charge-group - charge-group neighbor lists and kernels\n\n");
 319        }
 320        if (!fr->bExcl_IntraCGAll_InterCGNone)
 321        {
 322            gmx_fatal(FARGS,"The charge-group - charge-group force loops only support systems with all intra-cg interactions excluded and no inter-cg exclusions, this is not the case for this system.");
 323        }
 324    }
 325
 326    if (fr->solvent_opt == esolTIP4P) {
 327        enlist_w  = enlistTIP4P_ATOM;
 328        enlist_ww = enlistTIP4P_TIP4P;
 329    } else {
 330        enlist_w  = enlistSPC_ATOM;
 331        enlist_ww = enlistSPC_SPC;
 332    }
 333
 334    for(i=0; i<fr->nnblists; i++)
 335    {
 336        nbl = &(fr->nblists[i]);
 337        init_nblist(&nbl->nlist_sr[eNL_VDWQQ],&nbl->nlist_lr[eNL_VDWQQ],
 338                    maxsr,maxlr,ivdw,icoul,FALSE,enlist_def);
 339        init_nblist(&nbl->nlist_sr[eNL_VDW],&nbl->nlist_lr[eNL_VDW],
 340                    maxsr,maxlr,ivdw,0,FALSE,enlist_def);
 341        init_nblist(&nbl->nlist_sr[eNL_QQ],&nbl->nlist_lr[eNL_QQ],
 342                    maxsr,maxlr,0,icoul,FALSE,enlist_def);
 343        init_nblist(&nbl->nlist_sr[eNL_VDWQQ_WATER],&nbl->nlist_lr[eNL_VDWQQ_WATER],
 344                    maxsr_wat,maxlr_wat,ivdw,icoul, FALSE,enlist_w);
 345        init_nblist(&nbl->nlist_sr[eNL_QQ_WATER],&nbl->nlist_lr[eNL_QQ_WATER],
 346                    maxsr_wat,maxlr_wat,0,icoul, FALSE,enlist_w);
 347        init_nblist(&nbl->nlist_sr[eNL_VDWQQ_WATERWATER],&nbl->nlist_lr[eNL_VDWQQ_WATERWATER],
 348                    maxsr_wat,maxlr_wat,ivdw,icoul, FALSE,enlist_ww);
 349        init_nblist(&nbl->nlist_sr[eNL_QQ_WATERWATER],&nbl->nlist_lr[eNL_QQ_WATERWATER],
 350                    maxsr_wat,maxlr_wat,0,icoul, FALSE,enlist_ww);
 351
 352        if (fr->efep != efepNO)
 353        {
 354            if (fr->bEwald)
 355            {
 356                icoulf = 5;
 357            }
 358            else
 359            {
 360                icoulf = icoul;
 361            }
 362
 363            init_nblist(&nbl->nlist_sr[eNL_VDWQQ_FREE],&nbl->nlist_lr[eNL_VDWQQ_FREE],
 364                        maxsr,maxlr,ivdw,icoulf,TRUE,enlistATOM_ATOM);
 365            init_nblist(&nbl->nlist_sr[eNL_VDW_FREE],&nbl->nlist_lr[eNL_VDW_FREE],
 366                        maxsr,maxlr,ivdw,0,TRUE,enlistATOM_ATOM);
 367            init_nblist(&nbl->nlist_sr[eNL_QQ_FREE],&nbl->nlist_lr[eNL_QQ_FREE],
 368                        maxsr,maxlr,0,icoulf,TRUE,enlistATOM_ATOM);
 369        }
 370    }
 371    /* QMMM MM list */
 372    if (fr->bQMMM && fr->qr->QMMMscheme != eQMMMschemeoniom)
 373    {
 374        init_nblist(&fr->QMMMlist,NULL,
 375                    maxsr,maxlr,0,icoul,FALSE,enlistATOM_ATOM);
 376    }
 377
 378    fr->ns.nblist_initialized=TRUE;
 379 }
 380
 381 static void reset_nblist(t_nblist *nl)
 382 {
 383      nl->nri       = -1;
 384      nl->nrj       = 0;
 385      nl->maxlen    = 0;
 386      if (nl->jindex)
 387      {
 388          nl->jindex[0] = 0;
 389      }
 390 }
 391
 392 static void reset_neighbor_list(t_forcerec *fr,gmx_bool bLR,int nls,int eNL)
 393 {
 394     int n,i;
 395
 396     if (bLR)
 397     {
 398         reset_nblist(&(fr->nblists[nls].nlist_lr[eNL]));
 399     }
 400     else
 401     {
 402         for(n=0; n<fr->nnblists; n++)
 403         {
 404             for(i=0; i<eNL_NR; i++)
 405             {
 406                 reset_nblist(&(fr->nblists[n].nlist_sr[i]));
 407             }
 408         }
 409         if (fr->bQMMM)
 410         {
 411             /* only reset the short-range nblist */
 412             reset_nblist(&(fr->QMMMlist));
 413         }
 414     }
 415 }
 416
 417
 418
 419
 420 static inline void new_i_nblist(t_nblist *nlist,
 421                                 gmx_bool bLR,atom_id i_atom,int shift,int gid)
 422 {
 423     int    i,k,nri,nshift;
 424
 425     nri = nlist->nri;
 426
 427     /* Check whether we have to increase the i counter */
 428     if ((nri == -1) ||
 429         (nlist->iinr[nri]  != i_atom) ||
 430         (nlist->shift[nri] != shift) ||
 431         (nlist->gid[nri]   != gid))
 432     {
 433         /* This is something else. Now see if any entries have
 434          * been added in the list of the previous atom.
 435          */
 436         if ((nri == -1) ||
 437             ((nlist->jindex[nri+1] > nlist->jindex[nri]) &&
 438              (nlist->gid[nri] != -1)))
 439         {
 440             /* If so increase the counter */
 441             nlist->nri++;
 442             nri++;
 443             if (nlist->nri >= nlist->maxnri)
 444             {
 445                 nlist->maxnri += over_alloc_large(nlist->nri);
 446                 reallocate_nblist(nlist);
 447             }
 448         }
 449         /* Set the number of neighbours and the atom number */
 450         nlist->jindex[nri+1] = nlist->jindex[nri];
 451         nlist->iinr[nri]     = i_atom;
 452         nlist->gid[nri]      = gid;
 453         nlist->shift[nri]    = shift;
 454     }
 455 }
 456
 457 static inline void close_i_nblist(t_nblist *nlist)
 458 {
 459     int nri = nlist->nri;
 460     int len;
 461
 462     if (nri >= 0)
 463     {
 464         nlist->jindex[nri+1] = nlist->nrj;
 465
 466         len=nlist->nrj -  nlist->jindex[nri];
 467
 468         /* nlist length for water i molecules is treated statically
 469          * in the innerloops
 470          */
 471         if (len > nlist->maxlen)
 472         {
 473             nlist->maxlen = len;
 474         }
 475     }
 476 }
 477
 478 static inline void close_nblist(t_nblist *nlist)
 479 {
 480     /* Only close this nblist when it has been initialized.
 481      * Avoid the creation of i-lists with no j-particles.
 482      */
 483     if (nlist->nrj == 0)
 484     {
 485         /* Some assembly kernels do not support empty lists,
 486          * make sure here that we don't generate any empty lists.
 487          * With the current ns code this branch is taken in two cases:
 488          * No i-particles at all: nri=-1 here
 489          * There are i-particles, but no j-particles; nri=0 here
 490          */
 491         nlist->nri = 0;
 492     }
 493     else
 494     {
 495         /* Close list number nri by incrementing the count */
 496         nlist->nri++;
 497     }
 498 }
 499
 500 static inline void close_neighbor_list(t_forcerec *fr,gmx_bool bLR,int nls,int eNL,
 501                                        gmx_bool bMakeQMMMnblist)
 502 {
 503     int n,i;
 504
 505     if (bMakeQMMMnblist) {
 506         if (!bLR)
 507         {
 508             close_nblist(&(fr->QMMMlist));
 509         }
 510     }
 511     else
 512     {
 513         if (bLR)
 514         {
 515             close_nblist(&(fr->nblists[nls].nlist_lr[eNL]));
 516         }
 517         else
 518         {
 519             for(n=0; n<fr->nnblists; n++)
 520             {
 521                 for(i=0; (i<eNL_NR); i++)
 522                 {
 523                     close_nblist(&(fr->nblists[n].nlist_sr[i]));
 524                 }
 525             }
 526         }
 527     }
 528 }
 529
 530 static inline void add_j_to_nblist(t_nblist *nlist,atom_id j_atom,gmx_bool bLR)
 531 {
 532     int nrj=nlist->nrj;
 533
 534     if (nlist->nrj >= nlist->maxnrj)
 535     {
 536         nlist->maxnrj = over_alloc_small(nlist->nrj + 1);
 537         if (gmx_debug_at)
 538             fprintf(debug,"Increasing %s nblist %s j size to %d\n",
 539                     bLR ? "LR" : "SR",nrnb_str(nlist->il_code),nlist->maxnrj);
 540
 541         srenew(nlist->jjnr,nlist->maxnrj);
 542     }
 543
 544     nlist->jjnr[nrj] = j_atom;
 545     nlist->nrj ++;
 546 }
 547
 548 static inline void add_j_to_nblist_cg(t_nblist *nlist,
 549                                       atom_id j_start,int j_end,
 550                                       t_excl *bexcl,gmx_bool bLR)
 551 {
 552     int nrj=nlist->nrj;
 553     int j;
 554
 555     if (nlist->nrj >= nlist->maxnrj)
 556     {
 557         nlist->maxnrj = over_alloc_small(nlist->nrj + 1);
 558         if (gmx_debug_at)
 559             fprintf(debug,"Increasing %s nblist %s j size to %d\n",
 560                     bLR ? "LR" : "SR",nrnb_str(nlist->il_code),nlist->maxnrj);
 561
 562         srenew(nlist->jjnr    ,nlist->maxnrj);
 563         srenew(nlist->jjnr_end,nlist->maxnrj);
 564         srenew(nlist->excl    ,nlist->maxnrj*MAX_CGCGSIZE);
 565     }
 566
 567     nlist->jjnr[nrj]     = j_start;
 568     nlist->jjnr_end[nrj] = j_end;
 569
 570     if (j_end - j_start > MAX_CGCGSIZE)
 571     {
 572         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);
 573     }
 574
 575     /* Set the exclusions */
 576     for(j=j_start; j<j_end; j++)
 577     {
 578         nlist->excl[nrj*MAX_CGCGSIZE + j - j_start] = bexcl[j];
 579     }
 580
 581     nlist->nrj ++;
 582 }
 583
 584 typedef void
 585 put_in_list_t(gmx_bool              bHaveVdW[],
 586               int               ngid,
 587               t_mdatoms *       md,
 588               int               icg,
 589               int               jgid,
 590               int               nj,
 591               atom_id           jjcg[],
 592               atom_id           index[],
 593               t_excl            bExcl[],
 594               int               shift,
 595               t_forcerec *      fr,
 596               gmx_bool              bLR,
 597               gmx_bool              bDoVdW,
 598               gmx_bool              bDoCoul);
 599
 600 static void
 601 put_in_list_at(gmx_bool              bHaveVdW[],
 602                int               ngid,
 603                t_mdatoms *       md,
 604                int               icg,
 605                int               jgid,
 606                int               nj,
 607                atom_id           jjcg[],
 608                atom_id           index[],
 609                t_excl            bExcl[],
 610                int               shift,
 611                t_forcerec *      fr,
 612                gmx_bool              bLR,
 613                gmx_bool              bDoVdW,
 614                gmx_bool              bDoCoul)
 615 {
 616     /* The a[] index has been removed,
 617      * to put it back in i_atom should be a[i0] and jj should be a[jj].
 618      */
 619     t_nblist *   vdwc;
 620     t_nblist *   vdw;
 621     t_nblist *   coul;
 622     t_nblist *   vdwc_free  = NULL;
 623     t_nblist *   vdw_free   = NULL;
 624     t_nblist *   coul_free  = NULL;
 625     t_nblist *   vdwc_ww    = NULL;
 626     t_nblist *   coul_ww    = NULL;
 627
 628     int             i,j,jcg,igid,gid,nbl_ind,ind_ij;
 629     atom_id   jj,jj0,jj1,i_atom;
 630     int       i0,nicg,len;
 631
 632     int       *cginfo;
 633     int       *type,*typeB;
 634     real      *charge,*chargeB;
 635     real      qi,qiB,qq,rlj;
 636     gmx_bool      bFreeEnergy,bFree,bFreeJ,bNotEx,*bPert;
 637     gmx_bool      bDoVdW_i,bDoCoul_i,bDoCoul_i_sol;
 638     int       iwater,jwater;
 639     t_nblist  *nlist;
 640
 641     /* Copy some pointers */
 642     cginfo  = fr->cginfo;
 643     charge  = md->chargeA;
 644     chargeB = md->chargeB;
 645     type    = md->typeA;
 646     typeB   = md->typeB;
 647     bPert   = md->bPerturbed;
 648
 649     /* Get atom range */
 650     i0     = index[icg];
 651     nicg   = index[icg+1]-i0;
 652
 653     /* Get the i charge group info */
 654     igid   = GET_CGINFO_GID(cginfo[icg]);
 655     iwater = GET_CGINFO_SOLOPT(cginfo[icg]);
 656
 657     bFreeEnergy = FALSE;
 658     if (md->nPerturbed)
 659     {
 660         /* Check if any of the particles involved are perturbed.
 661          * If not we can do the cheaper normal put_in_list
 662          * and use more solvent optimization.
 663          */
 664         for(i=0; i<nicg; i++)
 665         {
 666             bFreeEnergy |= bPert[i0+i];
 667         }
 668         /* Loop over the j charge groups */
 669         for(j=0; (j<nj && !bFreeEnergy); j++)
 670         {
 671             jcg = jjcg[j];
 672             jj0 = index[jcg];
 673             jj1 = index[jcg+1];
 674             /* Finally loop over the atoms in the j-charge group */
 675             for(jj=jj0; jj<jj1; jj++)
 676             {
 677                 bFreeEnergy |= bPert[jj];
 678             }
 679         }
 680     }
 681
 682     /* Unpack pointers to neighbourlist structs */
 683     if (fr->nnblists == 1)
 684     {
 685         nbl_ind = 0;
 686     }
 687     else
 688     {
 689         nbl_ind = fr->gid2nblists[GID(igid,jgid,ngid)];
 690     }
 691     if (bLR)
 692     {
 693         nlist = fr->nblists[nbl_ind].nlist_lr;
 694     }
 695     else
 696     {
 697         nlist = fr->nblists[nbl_ind].nlist_sr;
 698     }
 699
 700     if (iwater != esolNO)
 701     {
 702         vdwc = &nlist[eNL_VDWQQ_WATER];
 703         vdw  = &nlist[eNL_VDW];
 704         coul = &nlist[eNL_QQ_WATER];
 705 #ifndef DISABLE_WATERWATER_NLIST
 706         vdwc_ww = &nlist[eNL_VDWQQ_WATERWATER];
 707         coul_ww = &nlist[eNL_QQ_WATERWATER];
 708 #endif
 709     }
 710     else
 711     {
 712         vdwc = &nlist[eNL_VDWQQ];
 713         vdw  = &nlist[eNL_VDW];
 714         coul = &nlist[eNL_QQ];
 715     }
 716
 717     if (!bFreeEnergy)
 718     {
 719         if (iwater != esolNO)
 720         {
 721             /* Loop over the atoms in the i charge group */
 722             i_atom  = i0;
 723             gid     = GID(igid,jgid,ngid);
 724             /* Create new i_atom for each energy group */
 725             if (bDoCoul && bDoVdW)
 726             {
 727                 new_i_nblist(vdwc,bLR,i_atom,shift,gid);
 728 #ifndef DISABLE_WATERWATER_NLIST
 729                 new_i_nblist(vdwc_ww,bLR,i_atom,shift,gid);
 730 #endif
 731             }
 732             if (bDoVdW)
 733             {
 734                 new_i_nblist(vdw,bLR,i_atom,shift,gid);
 735             }
 736             if (bDoCoul)
 737             {
 738                 new_i_nblist(coul,bLR,i_atom,shift,gid);
 739 #ifndef DISABLE_WATERWATER_NLIST
 740                 new_i_nblist(coul_ww,bLR,i_atom,shift,gid);
 741 #endif
 742             }
 743           /* Loop over the j charge groups */
 744             for(j=0; (j<nj); j++)
 745             {
 746                 jcg=jjcg[j];
 747
 748                 if (jcg == icg)
 749                 {
 750                     continue;
 751                 }
 752
 753                 jj0 = index[jcg];
 754                 jwater = GET_CGINFO_SOLOPT(cginfo[jcg]);
 755
 756                 if (iwater == esolSPC && jwater == esolSPC)
 757                 {
 758                     /* Interaction between two SPC molecules */
 759                     if (!bDoCoul)
 760                     {
 761                         /* VdW only - only first atoms in each water interact */
 762                         add_j_to_nblist(vdw,jj0,bLR);
 763                     }
 764                     else
 765                     {
 766 #ifdef DISABLE_WATERWATER_NLIST
 767                         /* Add entries for the three atoms - only do VdW if we need to */
 768                         if (!bDoVdW)
 769                         {
 770                             add_j_to_nblist(coul,jj0,bLR);
 771                         }
 772                         else
 773                         {
 774                             add_j_to_nblist(vdwc,jj0,bLR);
 775                         }
 776                         add_j_to_nblist(coul,jj0+1,bLR);
 777                         add_j_to_nblist(coul,jj0+2,bLR);
 778 #else
 779                         /* One entry for the entire water-water interaction */
 780                         if (!bDoVdW)
 781                         {
 782                             add_j_to_nblist(coul_ww,jj0,bLR);
 783                         }
 784                         else
 785                         {
 786                             add_j_to_nblist(vdwc_ww,jj0,bLR);
 787                         }
 788 #endif
 789                     }
 790                 }
 791                 else if (iwater == esolTIP4P && jwater == esolTIP4P)
 792                 {
 793                     /* Interaction between two TIP4p molecules */
 794                     if (!bDoCoul)
 795                     {
 796                         /* VdW only - only first atoms in each water interact */
 797                         add_j_to_nblist(vdw,jj0,bLR);
 798                     }
 799                     else
 800                     {
 801 #ifdef DISABLE_WATERWATER_NLIST
 802                         /* Add entries for the four atoms - only do VdW if we need to */
 803                         if (bDoVdW)
 804                         {
 805                             add_j_to_nblist(vdw,jj0,bLR);
 806                         }
 807                         add_j_to_nblist(coul,jj0+1,bLR);
 808                         add_j_to_nblist(coul,jj0+2,bLR);
 809                         add_j_to_nblist(coul,jj0+3,bLR);
 810 #else
 811                         /* One entry for the entire water-water interaction */
 812                         if (!bDoVdW)
 813                         {
 814                             add_j_to_nblist(coul_ww,jj0,bLR);
 815                         }
 816                         else
 817                         {
 818                             add_j_to_nblist(vdwc_ww,jj0,bLR);
 819                         }
 820 #endif
 821                     }
 822                 }
 823                 else
 824                 {
 825                     /* j charge group is not water, but i is.
 826                      * Add entries to the water-other_atom lists; the geometry of the water
 827                      * molecule doesn't matter - that is taken care of in the nonbonded kernel,
 828                      * so we don't care if it is SPC or TIP4P...
 829                      */
 830
 831                     jj1 = index[jcg+1];
 832
 833                     if (!bDoVdW)
 834                     {
 835                         for(jj=jj0; (jj<jj1); jj++)
 836                         {
 837                             if (charge[jj] != 0)
 838                             {
 839                                 add_j_to_nblist(coul,jj,bLR);
 840                             }
 841                         }
 842                     }
 843                     else if (!bDoCoul)
 844                     {
 845                         for(jj=jj0; (jj<jj1); jj++)
 846                         {
 847                             if (bHaveVdW[type[jj]])
 848                             {
 849                                 add_j_to_nblist(vdw,jj,bLR);
 850                             }
 851                         }
 852                     }
 853                     else
 854                     {
 855                         /* _charge_ _groups_ interact with both coulomb and LJ */
 856                         /* Check which atoms we should add to the lists!       */
 857                         for(jj=jj0; (jj<jj1); jj++)
 858                         {
 859                             if (bHaveVdW[type[jj]])
 860                             {
 861                                 if (charge[jj] != 0)
 862                                 {
 863                                     add_j_to_nblist(vdwc,jj,bLR);
 864                                 }
 865                                 else
 866                                 {
 867                                     add_j_to_nblist(vdw,jj,bLR);
 868                                 }
 869                             }
 870                             else if (charge[jj] != 0)
 871                             {
 872                                 add_j_to_nblist(coul,jj,bLR);
 873                             }
 874                         }
 875                     }
 876                 }
 877             }
 878             close_i_nblist(vdw);
 879             close_i_nblist(coul);
 880             close_i_nblist(vdwc);
 881 #ifndef DISABLE_WATERWATER_NLIST
 882             close_i_nblist(coul_ww);
 883             close_i_nblist(vdwc_ww);
 884 #endif
 885         }
 886         else
 887         {
 888             /* no solvent as i charge group */
 889             /* Loop over the atoms in the i charge group */
 890             for(i=0; i<nicg; i++)
 891             {
 892                 i_atom  = i0+i;
 893                 gid     = GID(igid,jgid,ngid);
 894                 qi      = charge[i_atom];
 895
 896                 /* Create new i_atom for each energy group */
 897                 if (bDoVdW && bDoCoul)
 898                 {
 899                     new_i_nblist(vdwc,bLR,i_atom,shift,gid);
 900                 }
 901                 if (bDoVdW)
 902                 {
 903                     new_i_nblist(vdw,bLR,i_atom,shift,gid);
 904                 }
 905                 if (bDoCoul)
 906                 {
 907                     new_i_nblist(coul,bLR,i_atom,shift,gid);
 908                 }
 909                 bDoVdW_i  = (bDoVdW  && bHaveVdW[type[i_atom]]);
 910                 bDoCoul_i = (bDoCoul && qi!=0);
 911
 912                 if (bDoVdW_i || bDoCoul_i)
 913                 {
 914                     /* Loop over the j charge groups */
 915                     for(j=0; (j<nj); j++)
 916                     {
 917                         jcg=jjcg[j];
 918
 919                         /* Check for large charge groups */
 920                         if (jcg == icg)
 921                         {
 922                             jj0 = i0 + i + 1;
 923                         }
 924                         else
 925                         {
 926                             jj0 = index[jcg];
 927                         }
 928
 929                         jj1=index[jcg+1];
 930                         /* Finally loop over the atoms in the j-charge group */
 931                         for(jj=jj0; jj<jj1; jj++)
 932                         {
 933                             bNotEx = NOTEXCL(bExcl,i,jj);
 934
 935                             if (bNotEx)
 936                             {
 937                                 if (!bDoVdW_i)
 938                                 {
 939                                     if (charge[jj] != 0)
 940                                     {
 941                                         add_j_to_nblist(coul,jj,bLR);
 942                                     }
 943                                 }
 944                                 else if (!bDoCoul_i)
 945                                 {
 946                                     if (bHaveVdW[type[jj]])
 947                                     {
 948                                         add_j_to_nblist(vdw,jj,bLR);
 949                                     }
 950                                 }
 951                                 else
 952                                 {
 953                                     if (bHaveVdW[type[jj]])
 954                                     {
 955                                         if (charge[jj] != 0)
 956                                         {
 957                                             add_j_to_nblist(vdwc,jj,bLR);
 958                                         }
 959                                         else
 960                                         {
 961                                             add_j_to_nblist(vdw,jj,bLR);
 962                                         }
 963                                     }
 964                                     else if (charge[jj] != 0)
 965                                     {
 966                                         add_j_to_nblist(coul,jj,bLR);
 967                                     }
 968                                 }
 969                             }
 970                         }
 971                     }
 972                 }
 973                 close_i_nblist(vdw);
 974                 close_i_nblist(coul);
 975                 close_i_nblist(vdwc);
 976             }
 977         }
 978     }
 979     else
 980     {
 981         /* we are doing free energy */
 982         vdwc_free = &nlist[eNL_VDWQQ_FREE];
 983         vdw_free  = &nlist[eNL_VDW_FREE];
 984         coul_free = &nlist[eNL_QQ_FREE];
 985         /* Loop over the atoms in the i charge group */
 986         for(i=0; i<nicg; i++)
 987         {
 988             i_atom  = i0+i;
 989             gid     = GID(igid,jgid,ngid);
 990             qi      = charge[i_atom];
 991             qiB     = chargeB[i_atom];
 992
 993             /* Create new i_atom for each energy group */
 994             if (bDoVdW && bDoCoul)
 995                 new_i_nblist(vdwc,bLR,i_atom,shift,gid);
 996             if (bDoVdW)
 997                 new_i_nblist(vdw,bLR,i_atom,shift,gid);
 998             if (bDoCoul)
 999                 new_i_nblist(coul,bLR,i_atom,shift,gid);
1000
1001             new_i_nblist(vdw_free,bLR,i_atom,shift,gid);
1002             new_i_nblist(coul_free,bLR,i_atom,shift,gid);
1003             new_i_nblist(vdwc_free,bLR,i_atom,shift,gid);
1004
1005             bDoVdW_i  = (bDoVdW  &&
1006                          (bHaveVdW[type[i_atom]] || bHaveVdW[typeB[i_atom]]));
1007             bDoCoul_i = (bDoCoul && (qi!=0 || qiB!=0));
1008             /* For TIP4P the first atom does not have a charge,
1009              * but the last three do. So we should still put an atom
1010              * without LJ but with charge in the water-atom neighborlist
1011              * for a TIP4p i charge group.
1012              * For SPC type water the first atom has LJ and charge,
1013              * so there is no such problem.
1014              */
1015             if (iwater == esolNO)
1016             {
1017                 bDoCoul_i_sol = bDoCoul_i;
1018             }
1019             else
1020             {
1021                 bDoCoul_i_sol = bDoCoul;
1022             }
1023
1024             if (bDoVdW_i || bDoCoul_i_sol)
1025             {
1026                 /* Loop over the j charge groups */
1027                 for(j=0; (j<nj); j++)
1028                 {
1029                     jcg=jjcg[j];
1030
1031                     /* Check for large charge groups */
1032                     if (jcg == icg)
1033                     {
1034                         jj0 = i0 + i + 1;
1035                     }
1036                     else
1037                     {
1038                         jj0 = index[jcg];
1039                     }
1040
1041                     jj1=index[jcg+1];
1042                     /* Finally loop over the atoms in the j-charge group */
1043                     bFree = bPert[i_atom];
1044                     for(jj=jj0; (jj<jj1); jj++)
1045                     {
1046                         bFreeJ = bFree || bPert[jj];
1047                         /* Complicated if, because the water H's should also
1048                          * see perturbed j-particles
1049                          */
1050                         if (iwater==esolNO || i==0 || bFreeJ)
1051                         {
1052                             bNotEx = NOTEXCL(bExcl,i,jj);
1053
1054                             if (bNotEx)
1055                             {
1056                                 if (bFreeJ)
1057                                 {
1058                                     if (!bDoVdW_i)
1059                                     {
1060                                         if (charge[jj]!=0 || chargeB[jj]!=0)
1061                                         {
1062                                             add_j_to_nblist(coul_free,jj,bLR);
1063                                         }
1064                                     }
1065                                     else if (!bDoCoul_i)
1066                                     {
1067                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
1068                                         {
1069                                             add_j_to_nblist(vdw_free,jj,bLR);
1070                                         }
1071                                     }
1072                                     else
1073                                     {
1074                                         if (bHaveVdW[type[jj]] || bHaveVdW[typeB[jj]])
1075                                         {
1076                                             if (charge[jj]!=0 || chargeB[jj]!=0)
1077                                             {
1078                                                 add_j_to_nblist(vdwc_free,jj,bLR);
1079                                             }
1080                                             else
1081                                             {
1082                                                 add_j_to_nblist(vdw_free,jj,bLR);
1083                                             }
1084                                         }
1085                                         else if (charge[jj]!=0 || chargeB[jj]!=0)
1086                                             add_j_to_nblist(coul_free,jj,bLR);
1087                                     }
1088                                 }
1089                                 else if (!bDoVdW_i)
1090                                 {
1091                                     /* This is done whether or not bWater is set */
1092                                     if (charge[jj] != 0)
1093                                     {
1094                                         add_j_to_nblist(coul,jj,bLR);
1095                                     }
1096                                 }
1097                                 else if (!bDoCoul_i_sol)
1098                                 {
1099                                     if (bHaveVdW[type[jj]])
1100                                     {
1101                                         add_j_to_nblist(vdw,jj,bLR);
1102                                     }
1103                                 }
1104                                 else
1105                                 {
1106                                     if (bHaveVdW[type[jj]])
1107                                     {
1108                                         if (charge[jj] != 0)
1109                                         {
1110                                             add_j_to_nblist(vdwc,jj,bLR);
1111                                         }
1112                                         else
1113                                         {
1114                                             add_j_to_nblist(vdw,jj,bLR);
1115                                         }
1116                                     }
1117                                     else if (charge[jj] != 0)
1118                                     {
1119                                         add_j_to_nblist(coul,jj,bLR);
1120                                     }
1121                                 }
1122                             }
1123                         }
1124                     }
1125                 }
1126             }
1127             close_i_nblist(vdw);
1128             close_i_nblist(coul);
1129             close_i_nblist(vdwc);
1130             close_i_nblist(vdw_free);
1131             close_i_nblist(coul_free);
1132             close_i_nblist(vdwc_free);
1133         }
1134     }
1135 }
1136
1137 static void
1138 put_in_list_qmmm(gmx_bool              bHaveVdW[],
1139                  int               ngid,
1140                  t_mdatoms *       md,
1141                  int               icg,
1142                  int               jgid,
1143                  int               nj,
1144                  atom_id           jjcg[],
1145                  atom_id           index[],
1146                  t_excl            bExcl[],
1147                  int               shift,
1148                  t_forcerec *      fr,
1149                  gmx_bool              bLR,
1150                  gmx_bool              bDoVdW,
1151                  gmx_bool              bDoCoul)
1152 {
1153     t_nblist *   coul;
1154     int           i,j,jcg,igid,gid;
1155     atom_id   jj,jj0,jj1,i_atom;
1156     int       i0,nicg;
1157     gmx_bool      bNotEx;
1158
1159     /* Get atom range */
1160     i0     = index[icg];
1161     nicg   = index[icg+1]-i0;
1162
1163     /* Get the i charge group info */
1164     igid   = GET_CGINFO_GID(fr->cginfo[icg]);
1165
1166     coul = &fr->QMMMlist;
1167
1168     /* Loop over atoms in the ith charge group */
1169     for (i=0;i<nicg;i++)
1170     {
1171         i_atom = i0+i;
1172         gid    = GID(igid,jgid,ngid);
1173         /* Create new i_atom for each energy group */
1174         new_i_nblist(coul,bLR,i_atom,shift,gid);
1175
1176         /* Loop over the j charge groups */
1177         for (j=0;j<nj;j++)
1178         {
1179             jcg=jjcg[j];
1180
1181             /* Charge groups cannot have QM and MM atoms simultaneously */
1182             if (jcg!=icg)
1183             {
1184                 jj0 = index[jcg];
1185                 jj1 = index[jcg+1];
1186                 /* Finally loop over the atoms in the j-charge group */
1187                 for(jj=jj0; jj<jj1; jj++)
1188                 {
1189                     bNotEx = NOTEXCL(bExcl,i,jj);
1190                     if(bNotEx)
1191                         add_j_to_nblist(coul,jj,bLR);
1192                 }
1193             }
1194         }
1195         close_i_nblist(coul);
1196     }
1197 }
1198
1199 static void
1200 put_in_list_cg(gmx_bool              bHaveVdW[],
1201                int               ngid,
1202                t_mdatoms *       md,
1203                int               icg,
1204                int               jgid,
1205                int               nj,
1206                atom_id           jjcg[],
1207                atom_id           index[],
1208                t_excl            bExcl[],
1209                int               shift,
1210                t_forcerec *      fr,
1211                gmx_bool              bLR,
1212                gmx_bool              bDoVdW,
1213                gmx_bool              bDoCoul)
1214 {
1215     int          cginfo;
1216     int          igid,gid,nbl_ind;
1217     t_nblist *   vdwc;
1218     int          j,jcg;
1219
1220     cginfo = fr->cginfo[icg];
1221
1222     igid = GET_CGINFO_GID(cginfo);
1223     gid  = GID(igid,jgid,ngid);
1224
1225     /* Unpack pointers to neighbourlist structs */
1226     if (fr->nnblists == 1)
1227     {
1228         nbl_ind = 0;
1229     }
1230     else
1231     {
1232         nbl_ind = fr->gid2nblists[gid];
1233     }
1234     if (bLR)
1235     {
1236         vdwc = &fr->nblists[nbl_ind].nlist_lr[eNL_VDWQQ];
1237     }
1238     else
1239     {
1240         vdwc = &fr->nblists[nbl_ind].nlist_sr[eNL_VDWQQ];
1241     }
1242
1243     /* Make a new neighbor list for charge group icg.
1244      * Currently simply one neighbor list is made with LJ and Coulomb.
1245      * If required, zero interactions could be removed here
1246      * or in the force loop.
1247      */
1248     new_i_nblist(vdwc,bLR,index[icg],shift,gid);
1249     vdwc->iinr_end[vdwc->nri] = index[icg+1];
1250
1251     for(j=0; (j<nj); j++)
1252     {
1253         jcg = jjcg[j];
1254         /* Skip the icg-icg pairs if all self interactions are excluded */
1255         if (!(jcg == icg && GET_CGINFO_EXCL_INTRA(cginfo)))
1256         {
1257             /* Here we add the j charge group jcg to the list,
1258              * exclusions are also added to the list.
1259              */
1260             add_j_to_nblist_cg(vdwc,index[jcg],index[jcg+1],bExcl,bLR);
1261         }
1262     }
1263
1264     close_i_nblist(vdwc);
1265 }
1266
1267 static void setexcl(atom_id start,atom_id end,t_blocka *excl,gmx_bool b,
1268                     t_excl bexcl[])
1269 {
1270     atom_id i,k;
1271
1272     if (b)
1273     {
1274         for(i=start; i<end; i++)
1275         {
1276             for(k=excl->index[i]; k<excl->index[i+1]; k++)
1277             {
1278                 SETEXCL(bexcl,i-start,excl->a[k]);
1279             }
1280         }
1281     }
1282     else
1283     {
1284         for(i=start; i<end; i++)
1285         {
1286             for(k=excl->index[i]; k<excl->index[i+1]; k++)
1287             {
1288                 RMEXCL(bexcl,i-start,excl->a[k]);
1289             }
1290         }
1291     }
1292 }
1293
1294 int calc_naaj(int icg,int cgtot)
1295 {
1296     int naaj;
1297
1298     if ((cgtot % 2) == 1)
1299     {
1300         /* Odd number of charge groups, easy */
1301         naaj = 1 + (cgtot/2);
1302     }
1303     else if ((cgtot % 4) == 0)
1304     {
1305     /* Multiple of four is hard */
1306         if (icg < cgtot/2)
1307         {
1308             if ((icg % 2) == 0)
1309             {
1310                 naaj=1+(cgtot/2);
1311             }
1312             else
1313             {
1314                 naaj=cgtot/2;
1315             }
1316         }
1317         else
1318         {
1319             if ((icg % 2) == 1)
1320             {
1321                 naaj=1+(cgtot/2);
1322             }
1323             else
1324             {
1325                 naaj=cgtot/2;
1326             }
1327         }
1328     }
1329     else
1330     {
1331         /* cgtot/2 = odd */
1332         if ((icg % 2) == 0)
1333         {
1334             naaj=1+(cgtot/2);
1335         }
1336         else
1337         {
1338             naaj=cgtot/2;
1339         }
1340     }
1341 #ifdef DEBUG
1342     fprintf(log,"naaj=%d\n",naaj);
1343 #endif
1344
1345     return naaj;
1346 }
1347
1348 /************************************************
1349  *
1350  *  S I M P L E      C O R E     S T U F F
1351  *
1352  ************************************************/
1353
1354 static real calc_image_tric(rvec xi,rvec xj,matrix box,
1355                             rvec b_inv,int *shift)
1356 {
1357     /* This code assumes that the cut-off is smaller than
1358      * a half times the smallest diagonal element of the box.
1359      */
1360     const real h25=2.5;
1361     real dx,dy,dz;
1362     real r2;
1363     int  tx,ty,tz;
1364
1365     /* Compute diff vector */
1366     dz = xj[ZZ] - xi[ZZ];
1367     dy = xj[YY] - xi[YY];
1368     dx = xj[XX] - xi[XX];
1369
1370   /* Perform NINT operation, using trunc operation, therefore
1371    * we first add 2.5 then subtract 2 again
1372    */
1373     tz = dz*b_inv[ZZ] + h25;
1374     tz -= 2;
1375     dz -= tz*box[ZZ][ZZ];
1376     dy -= tz*box[ZZ][YY];
1377     dx -= tz*box[ZZ][XX];
1378
1379     ty = dy*b_inv[YY] + h25;
1380     ty -= 2;
1381     dy -= ty*box[YY][YY];
1382     dx -= ty*box[YY][XX];
1383
1384     tx = dx*b_inv[XX]+h25;
1385     tx -= 2;
1386     dx -= tx*box[XX][XX];
1387
1388     /* Distance squared */
1389     r2 = (dx*dx) + (dy*dy) + (dz*dz);
1390
1391     *shift = XYZ2IS(tx,ty,tz);
1392
1393     return r2;
1394 }
1395
1396 static real calc_image_rect(rvec xi,rvec xj,rvec box_size,
1397                             rvec b_inv,int *shift)
1398 {
1399     const real h15=1.5;
1400     real ddx,ddy,ddz;
1401     real dx,dy,dz;
1402     real r2;
1403     int  tx,ty,tz;
1404
1405     /* Compute diff vector */
1406     dx = xj[XX] - xi[XX];
1407     dy = xj[YY] - xi[YY];
1408     dz = xj[ZZ] - xi[ZZ];
1409
1410     /* Perform NINT operation, using trunc operation, therefore
1411      * we first add 1.5 then subtract 1 again
1412      */
1413     tx = dx*b_inv[XX] + h15;
1414     ty = dy*b_inv[YY] + h15;
1415     tz = dz*b_inv[ZZ] + h15;
1416     tx--;
1417     ty--;
1418     tz--;
1419
1420     /* Correct diff vector for translation */
1421     ddx = tx*box_size[XX] - dx;
1422     ddy = ty*box_size[YY] - dy;
1423     ddz = tz*box_size[ZZ] - dz;
1424
1425     /* Distance squared */
1426     r2 = (ddx*ddx) + (ddy*ddy) + (ddz*ddz);
1427
1428     *shift = XYZ2IS(tx,ty,tz);
1429
1430     return r2;
1431 }
1432
1433 static void add_simple(t_ns_buf *nsbuf,int nrj,atom_id cg_j,
1434                        gmx_bool bHaveVdW[],int ngid,t_mdatoms *md,
1435                        int icg,int jgid,t_block *cgs,t_excl bexcl[],
1436                        int shift,t_forcerec *fr,put_in_list_t *put_in_list)
1437 {
1438     if (nsbuf->nj + nrj > MAX_CG)
1439     {
1440         put_in_list(bHaveVdW,ngid,md,icg,jgid,nsbuf->ncg,nsbuf->jcg,
1441                     cgs->index,bexcl,shift,fr,FALSE,TRUE,TRUE);
1442         /* Reset buffer contents */
1443         nsbuf->ncg = nsbuf->nj = 0;
1444     }
1445     nsbuf->jcg[nsbuf->ncg++] = cg_j;
1446     nsbuf->nj += nrj;
1447 }
1448
1449 static void ns_inner_tric(rvec x[],int icg,int *i_egp_flags,
1450                           int njcg,atom_id jcg[],
1451                           matrix box,rvec b_inv,real rcut2,
1452                           t_block *cgs,t_ns_buf **ns_buf,
1453                           gmx_bool bHaveVdW[],int ngid,t_mdatoms *md,
1454                           t_excl bexcl[],t_forcerec *fr,
1455                           put_in_list_t *put_in_list)
1456 {
1457     int      shift;
1458     int      j,nrj,jgid;
1459     int      *cginfo=fr->cginfo;
1460     atom_id  cg_j,*cgindex;
1461     t_ns_buf *nsbuf;
1462
1463     cgindex = cgs->index;
1464     shift   = CENTRAL;
1465     for(j=0; (j<njcg); j++)
1466     {
1467         cg_j   = jcg[j];
1468         nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1469         if (calc_image_tric(x[icg],x[cg_j],box,b_inv,&shift) < rcut2)
1470         {
1471             jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1472             if (!(i_egp_flags[jgid] & EGP_EXCL))
1473             {
1474                 add_simple(&ns_buf[jgid][shift],nrj,cg_j,
1475                            bHaveVdW,ngid,md,icg,jgid,cgs,bexcl,shift,fr,
1476                            put_in_list);
1477             }
1478         }
1479     }
1480 }
1481
1482 static void ns_inner_rect(rvec x[],int icg,int *i_egp_flags,
1483                           int njcg,atom_id jcg[],
1484                           gmx_bool bBox,rvec box_size,rvec b_inv,real rcut2,
1485                           t_block *cgs,t_ns_buf **ns_buf,
1486                           gmx_bool bHaveVdW[],int ngid,t_mdatoms *md,
1487                           t_excl bexcl[],t_forcerec *fr,
1488                           put_in_list_t *put_in_list)
1489 {
1490     int      shift;
1491     int      j,nrj,jgid;
1492     int      *cginfo=fr->cginfo;
1493     atom_id  cg_j,*cgindex;
1494     t_ns_buf *nsbuf;
1495
1496     cgindex = cgs->index;
1497     if (bBox)
1498     {
1499         shift = CENTRAL;
1500         for(j=0; (j<njcg); j++)
1501         {
1502             cg_j   = jcg[j];
1503             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1504             if (calc_image_rect(x[icg],x[cg_j],box_size,b_inv,&shift) < rcut2)
1505             {
1506                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1507                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1508                 {
1509                     add_simple(&ns_buf[jgid][shift],nrj,cg_j,
1510                                bHaveVdW,ngid,md,icg,jgid,cgs,bexcl,shift,fr,
1511                                put_in_list);
1512                 }
1513             }
1514         }
1515     }
1516     else
1517     {
1518         for(j=0; (j<njcg); j++)
1519         {
1520             cg_j   = jcg[j];
1521             nrj    = cgindex[cg_j+1]-cgindex[cg_j];
1522             if ((rcut2 == 0) || (distance2(x[icg],x[cg_j]) < rcut2)) {
1523                 jgid  = GET_CGINFO_GID(cginfo[cg_j]);
1524                 if (!(i_egp_flags[jgid] & EGP_EXCL))
1525                 {
1526                     add_simple(&ns_buf[jgid][CENTRAL],nrj,cg_j,
1527                                bHaveVdW,ngid,md,icg,jgid,cgs,bexcl,CENTRAL,fr,
1528                                put_in_list);
1529                 }
1530             }
1531         }
1532     }
1533 }
1534
1535 /* ns_simple_core needs to be adapted for QMMM still 2005 */
1536
1537 static int ns_simple_core(t_forcerec *fr,
1538                           gmx_localtop_t *top,
1539                           t_mdatoms *md,
1540                           matrix box,rvec box_size,
1541                           t_excl bexcl[],atom_id *aaj,
1542                           int ngid,t_ns_buf **ns_buf,
1543                           put_in_list_t *put_in_list,gmx_bool bHaveVdW[])
1544 {
1545     int      naaj,k;
1546     real     rlist2;
1547     int      nsearch,icg,jcg,igid,i0,nri,nn;
1548     int      *cginfo;
1549     t_ns_buf *nsbuf;
1550     /* atom_id  *i_atoms; */
1551     t_block  *cgs=&(top->cgs);
1552     t_blocka *excl=&(top->excls);
1553     rvec     b_inv;
1554     int      m;
1555     gmx_bool     bBox,bTriclinic;
1556     int      *i_egp_flags;
1557
1558     rlist2 = sqr(fr->rlist);
1559
1560     bBox = (fr->ePBC != epbcNONE);
1561     if (bBox)
1562     {
1563         for(m=0; (m<DIM); m++)
1564         {
1565             b_inv[m] = divide_err(1.0,box_size[m]);
1566         }
1567         bTriclinic = TRICLINIC(box);
1568     }
1569     else
1570     {
1571         bTriclinic = FALSE;
1572     }
1573
1574     cginfo = fr->cginfo;
1575
1576     nsearch=0;
1577     for (icg=fr->cg0; (icg<fr->hcg); icg++)
1578     {
1579         /*
1580           i0        = cgs->index[icg];
1581           nri       = cgs->index[icg+1]-i0;
1582           i_atoms   = &(cgs->a[i0]);
1583           i_eg_excl = fr->eg_excl + ngid*md->cENER[*i_atoms];
1584           setexcl(nri,i_atoms,excl,TRUE,bexcl);
1585         */
1586         igid = GET_CGINFO_GID(cginfo[icg]);
1587         i_egp_flags = fr->egp_flags + ngid*igid;
1588         setexcl(cgs->index[icg],cgs->index[icg+1],excl,TRUE,bexcl);
1589
1590         naaj=calc_naaj(icg,cgs->nr);
1591         if (bTriclinic)
1592         {
1593             ns_inner_tric(fr->cg_cm,icg,i_egp_flags,naaj,&(aaj[icg]),
1594                           box,b_inv,rlist2,cgs,ns_buf,
1595                           bHaveVdW,ngid,md,bexcl,fr,put_in_list);
1596         }
1597         else
1598         {
1599             ns_inner_rect(fr->cg_cm,icg,i_egp_flags,naaj,&(aaj[icg]),
1600                           bBox,box_size,b_inv,rlist2,cgs,ns_buf,
1601                           bHaveVdW,ngid,md,bexcl,fr,put_in_list);
1602         }
1603         nsearch += naaj;
1604
1605         for(nn=0; (nn<ngid); nn++)
1606         {
1607             for(k=0; (k<SHIFTS); k++)
1608             {
1609                 nsbuf = &(ns_buf[nn][k]);
1610                 if (nsbuf->ncg > 0)
1611                 {
1612                     put_in_list(bHaveVdW,ngid,md,icg,nn,nsbuf->ncg,nsbuf->jcg,
1613                                 cgs->index,bexcl,k,fr,FALSE,TRUE,TRUE);
1614                     nsbuf->ncg=nsbuf->nj=0;
1615                 }
1616             }
1617         }
1618         /* setexcl(nri,i_atoms,excl,FALSE,bexcl); */
1619         setexcl(cgs->index[icg],cgs->index[icg+1],excl,FALSE,bexcl);
1620     }
1621     close_neighbor_list(fr,FALSE,-1,-1,FALSE);
1622
1623     return nsearch;
1624 }
1625
1626 /************************************************
1627  *
1628  *    N S 5     G R I D     S T U F F
1629  *
1630  ************************************************/
1631
1632 static inline void get_dx(int Nx,real gridx,real rc2,int xgi,real x,
1633                           int *dx0,int *dx1,real *dcx2)
1634 {
1635     real dcx,tmp;
1636     int  xgi0,xgi1,i;
1637
1638     if (xgi < 0)
1639     {
1640         *dx0 = 0;
1641         xgi0 = -1;
1642         *dx1 = -1;
1643         xgi1 = 0;
1644     }
1645     else if (xgi >= Nx)
1646     {
1647         *dx0 = Nx;
1648         xgi0 = Nx-1;
1649         *dx1 = Nx-1;
1650         xgi1 = Nx;
1651     }
1652     else
1653     {
1654         dcx2[xgi] = 0;
1655         *dx0 = xgi;
1656         xgi0 = xgi-1;
1657         *dx1 = xgi;
1658         xgi1 = xgi+1;
1659     }
1660
1661     for(i=xgi0; i>=0; i--)
1662     {
1663         dcx = (i+1)*gridx-x;
1664         tmp = dcx*dcx;
1665         if (tmp >= rc2)
1666             break;
1667         *dx0 = i;
1668         dcx2[i] = tmp;
1669     }
1670     for(i=xgi1; i<Nx; i++)
1671     {
1672         dcx = i*gridx-x;
1673         tmp = dcx*dcx;
1674         if (tmp >= rc2)
1675         {
1676             break;
1677         }
1678         *dx1 = i;
1679         dcx2[i] = tmp;
1680     }
1681 }
1682
1683 static inline void get_dx_dd(int Nx,real gridx,real rc2,int xgi,real x,
1684                              int ncpddc,int shift_min,int shift_max,
1685                              int *g0,int *g1,real *dcx2)
1686 {
1687     real dcx,tmp;
1688     int  g_min,g_max,shift_home;
1689
1690     if (xgi < 0)
1691     {
1692         g_min = 0;
1693         g_max = Nx - 1;
1694         *g0   = 0;
1695         *g1   = -1;
1696     }
1697     else if (xgi >= Nx)
1698     {
1699         g_min = 0;
1700         g_max = Nx - 1;
1701         *g0   = Nx;
1702         *g1   = Nx - 1;
1703     }
1704     else
1705     {
1706         if (ncpddc == 0)
1707         {
1708             g_min = 0;
1709             g_max = Nx - 1;
1710         }
1711         else
1712         {
1713             if (xgi < ncpddc)
1714             {
1715                 shift_home = 0;
1716             }
1717             else
1718             {
1719                 shift_home = -1;
1720             }
1721             g_min = (shift_min == shift_home ? 0          : ncpddc);
1722             g_max = (shift_max == shift_home ? ncpddc - 1 : Nx - 1);
1723         }
1724         if (shift_min > 0)
1725         {
1726             *g0 = g_min;
1727             *g1 = g_min - 1;
1728         }
1729         else if (shift_max < 0)
1730         {
1731             *g0 = g_max + 1;
1732             *g1 = g_max;
1733         }
1734         else
1735         {
1736             *g0 = xgi;
1737             *g1 = xgi;
1738             dcx2[xgi] = 0;
1739         }
1740     }
1741
1742     while (*g0 > g_min)
1743     {
1744         /* Check one grid cell down */
1745         dcx = ((*g0 - 1) + 1)*gridx - x;
1746         tmp = dcx*dcx;
1747         if (tmp >= rc2)
1748         {
1749             break;
1750         }
1751         (*g0)--;
1752         dcx2[*g0] = tmp;
1753     }
1754
1755     while (*g1 < g_max)
1756     {
1757         /* Check one grid cell up */
1758         dcx = (*g1 + 1)*gridx - x;
1759         tmp = dcx*dcx;
1760         if (tmp >= rc2)
1761         {
1762             break;
1763         }
1764         (*g1)++;
1765         dcx2[*g1] = tmp;
1766     }
1767 }
1768
1769
1770 #define sqr(x) ((x)*(x))
1771 #define calc_dx2(XI,YI,ZI,y) (sqr(XI-y[XX]) + sqr(YI-y[YY]) + sqr(ZI-y[ZZ]))
1772 #define calc_cyl_dx2(XI,YI,y) (sqr(XI-y[XX]) + sqr(YI-y[YY]))
1773 /****************************************************
1774  *
1775  *    F A S T   N E I G H B O R  S E A R C H I N G
1776  *
1777  *    Optimized neighboursearching routine using grid
1778  *    at least 1x1x1, see GROMACS manual
1779  *
1780  ****************************************************/
1781
1782 static void do_longrange(t_commrec *cr,gmx_localtop_t *top,t_forcerec *fr,
1783                          int ngid,t_mdatoms *md,int icg,
1784                          int jgid,int nlr,
1785                          atom_id lr[],t_excl bexcl[],int shift,
1786                          rvec x[],rvec box_size,t_nrnb *nrnb,
1787                          real lambda,real *dvdlambda,
1788                          gmx_grppairener_t *grppener,
1789                          gmx_bool bDoVdW,gmx_bool bDoCoul,
1790                          gmx_bool bEvaluateNow,put_in_list_t *put_in_list,
1791                          gmx_bool bHaveVdW[],
1792                          gmx_bool bDoForces,rvec *f)
1793 {
1794     int n,i;
1795     t_nblist *nl;
1796
1797     for(n=0; n<fr->nnblists; n++)
1798     {
1799         for(i=0; (i<eNL_NR); i++)
1800         {
1801             nl = &fr->nblists[n].nlist_lr[i];
1802             if ((nl->nri > nl->maxnri-32) || bEvaluateNow)
1803             {
1804                 close_neighbor_list(fr,TRUE,n,i,FALSE);
1805                 /* Evaluate the energies and forces */
1806                 do_nonbonded(cr,fr,x,f,md,NULL,
1807                              grppener->ener[fr->bBHAM ? egBHAMLR : egLJLR],
1808                              grppener->ener[egCOULLR],
1809                                                          grppener->ener[egGB],box_size,
1810                              nrnb,lambda,dvdlambda,n,i,
1811                              GMX_DONB_LR | GMX_DONB_FORCES);
1812
1813                 reset_neighbor_list(fr,TRUE,n,i);
1814             }
1815         }
1816     }
1817
1818     if (!bEvaluateNow)
1819     {
1820         /* Put the long range particles in a list */
1821         /* do_longrange is never called for QMMM  */
1822         put_in_list(bHaveVdW,ngid,md,icg,jgid,nlr,lr,top->cgs.index,
1823                     bexcl,shift,fr,TRUE,bDoVdW,bDoCoul);
1824     }
1825 }
1826
1827 static void get_cutoff2(t_forcerec *fr,gmx_bool bDoLongRange,
1828                         real *rvdw2,real *rcoul2,
1829                         real *rs2,real *rm2,real *rl2)
1830 {
1831     *rs2 = sqr(fr->rlist);
1832     if (bDoLongRange && fr->bTwinRange)
1833     {
1834         /* The VdW and elec. LR cut-off's could be different,
1835          * so we can not simply set them to rlistlong.
1836          */
1837         if (EVDW_MIGHT_BE_ZERO_AT_CUTOFF(fr->vdwtype) &&
1838             fr->rvdw > fr->rlist)
1839         {
1840             *rvdw2  = sqr(fr->rlistlong);
1841         }
1842         else
1843         {
1844             *rvdw2  = sqr(fr->rvdw);
1845         }
1846         if (EEL_MIGHT_BE_ZERO_AT_CUTOFF(fr->eeltype) &&
1847             fr->rcoulomb > fr->rlist)
1848         {
1849             *rcoul2 = sqr(fr->rlistlong);
1850         }
1851         else
1852         {
1853             *rcoul2 = sqr(fr->rcoulomb);
1854         }
1855     }
1856     else
1857     {
1858         /* Workaround for a gcc -O3 or -ffast-math problem */
1859         *rvdw2  = *rs2;
1860         *rcoul2 = *rs2;
1861     }
1862     *rm2 = min(*rvdw2,*rcoul2);
1863     *rl2 = max(*rvdw2,*rcoul2);
1864 }
1865
1866 static void init_nsgrid_lists(t_forcerec *fr,int ngid,gmx_ns_t *ns)
1867 {
1868     real rvdw2,rcoul2,rs2,rm2,rl2;
1869     int j;
1870
1871     get_cutoff2(fr,TRUE,&rvdw2,&rcoul2,&rs2,&rm2,&rl2);
1872
1873     /* Short range buffers */
1874     snew(ns->nl_sr,ngid);
1875     /* Counters */
1876     snew(ns->nsr,ngid);
1877     snew(ns->nlr_ljc,ngid);
1878     snew(ns->nlr_one,ngid);
1879
1880     if (rm2 > rs2)
1881     {
1882             /* Long range VdW and Coul buffers */
1883         snew(ns->nl_lr_ljc,ngid);
1884     }
1885     if (rl2 > rm2)
1886     {
1887         /* Long range VdW or Coul only buffers */
1888         snew(ns->nl_lr_one,ngid);
1889     }
1890     for(j=0; (j<ngid); j++) {
1891         snew(ns->nl_sr[j],MAX_CG);
1892         if (rm2 > rs2)
1893         {
1894             snew(ns->nl_lr_ljc[j],MAX_CG);
1895         }
1896         if (rl2 > rm2)
1897         {
1898             snew(ns->nl_lr_one[j],MAX_CG);
1899         }
1900     }
1901     if (debug)
1902     {
1903         fprintf(debug,
1904                 "ns5_core: rs2 = %g, rm2 = %g, rl2 = %g (nm^2)\n",
1905                 rs2,rm2,rl2);
1906     }
1907 }
1908
1909 static int nsgrid_core(FILE *log,t_commrec *cr,t_forcerec *fr,
1910                        matrix box,rvec box_size,int ngid,
1911                        gmx_localtop_t *top,
1912                        t_grid *grid,rvec x[],
1913                        t_excl bexcl[],gmx_bool *bExcludeAlleg,
1914                        t_nrnb *nrnb,t_mdatoms *md,
1915                        real lambda,real *dvdlambda,
1916                        gmx_grppairener_t *grppener,
1917                        put_in_list_t *put_in_list,
1918                        gmx_bool bHaveVdW[],
1919                        gmx_bool bDoLongRange,gmx_bool bDoForces,rvec *f,
1920                        gmx_bool bMakeQMMMnblist)
1921 {
1922     gmx_ns_t *ns;
1923     atom_id **nl_lr_ljc,**nl_lr_one,**nl_sr;
1924     int     *nlr_ljc,*nlr_one,*nsr;
1925     gmx_domdec_t *dd=NULL;
1926     t_block *cgs=&(top->cgs);
1927     int     *cginfo=fr->cginfo;
1928     /* atom_id *i_atoms,*cgsindex=cgs->index; */
1929     ivec    sh0,sh1,shp;
1930     int     cell_x,cell_y,cell_z;
1931     int     d,tx,ty,tz,dx,dy,dz,cj;
1932 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
1933     int     zsh_ty,zsh_tx,ysh_tx;
1934 #endif
1935     int     dx0,dx1,dy0,dy1,dz0,dz1;
1936     int     Nx,Ny,Nz,shift=-1,j,nrj,nns,nn=-1;
1937     real    gridx,gridy,gridz,grid_x,grid_y,grid_z;
1938     real    *dcx2,*dcy2,*dcz2;
1939     int     zgi,ygi,xgi;
1940     int     cg0,cg1,icg=-1,cgsnr,i0,igid,nri,naaj,max_jcg;
1941     int     jcg0,jcg1,jjcg,cgj0,jgid;
1942     int     *grida,*gridnra,*gridind;
1943     gmx_bool    rvdw_lt_rcoul,rcoul_lt_rvdw;
1944     rvec    xi,*cgcm,grid_offset;
1945     real    r2,rs2,rvdw2,rcoul2,rm2,rl2,XI,YI,ZI,dcx,dcy,dcz,tmp1,tmp2;
1946     int     *i_egp_flags;
1947     gmx_bool    bDomDec,bTriclinicX,bTriclinicY;
1948     ivec    ncpddc;
1949
1950     ns = &fr->ns;
1951
1952     bDomDec = DOMAINDECOMP(cr);
1953     if (bDomDec)
1954     {
1955         dd = cr->dd;
1956     }
1957
1958     bTriclinicX = ((YY < grid->npbcdim &&
1959                     (!bDomDec || dd->nc[YY]==1) && box[YY][XX] != 0) ||
1960                    (ZZ < grid->npbcdim &&
1961                     (!bDomDec || dd->nc[ZZ]==1) && box[ZZ][XX] != 0));
1962     bTriclinicY =  (ZZ < grid->npbcdim &&
1963                     (!bDomDec || dd->nc[ZZ]==1) && box[ZZ][YY] != 0);
1964
1965     cgsnr    = cgs->nr;
1966
1967     get_cutoff2(fr,bDoLongRange,&rvdw2,&rcoul2,&rs2,&rm2,&rl2);
1968
1969     rvdw_lt_rcoul = (rvdw2 >= rcoul2);
1970     rcoul_lt_rvdw = (rcoul2 >= rvdw2);
1971
1972     if (bMakeQMMMnblist)
1973     {
1974         rm2 = rl2;
1975         rs2 = rl2;
1976     }
1977
1978     nl_sr     = ns->nl_sr;
1979     nsr       = ns->nsr;
1980     nl_lr_ljc = ns->nl_lr_ljc;
1981     nl_lr_one = ns->nl_lr_one;
1982     nlr_ljc   = ns->nlr_ljc;
1983     nlr_one   = ns->nlr_one;
1984
1985     /* Unpack arrays */
1986     cgcm    = fr->cg_cm;
1987     Nx      = grid->n[XX];
1988     Ny      = grid->n[YY];
1989     Nz      = grid->n[ZZ];
1990     grida   = grid->a;
1991     gridind = grid->index;
1992     gridnra = grid->nra;
1993     nns     = 0;
1994
1995     gridx      = grid->cell_size[XX];
1996     gridy      = grid->cell_size[YY];
1997     gridz      = grid->cell_size[ZZ];
1998     grid_x     = 1/gridx;
1999     grid_y     = 1/gridy;
2000     grid_z     = 1/gridz;
2001     copy_rvec(grid->cell_offset,grid_offset);
2002     copy_ivec(grid->ncpddc,ncpddc);
2003     dcx2       = grid->dcx2;
2004     dcy2       = grid->dcy2;
2005     dcz2       = grid->dcz2;
2006
2007 #ifdef ALLOW_OFFDIAG_LT_HALFDIAG
2008     zsh_ty = floor(-box[ZZ][YY]/box[YY][YY]+0.5);
2009     zsh_tx = floor(-box[ZZ][XX]/box[XX][XX]+0.5);
2010     ysh_tx = floor(-box[YY][XX]/box[XX][XX]+0.5);
2011     if (zsh_tx!=0 && ysh_tx!=0)
2012     {
2013         /* This could happen due to rounding, when both ratios are 0.5 */
2014         ysh_tx = 0;
2015     }
2016 #endif
2017
2018     debug_gmx();
2019
2020     if (fr->n_tpi)
2021     {
2022         /* We only want a list for the test particle */
2023         cg0 = cgsnr - 1;
2024     }
2025     else
2026     {
2027         cg0 = grid->icg0;
2028     }
2029     cg1 = grid->icg1;
2030
2031     /* Set the shift range */
2032     for(d=0; d<DIM; d++)
2033     {
2034         sh0[d] = -1;
2035         sh1[d] = 1;
2036         /* Check if we need periodicity shifts.
2037          * Without PBC or with domain decomposition we don't need them.
2038          */
2039         if (d >= ePBC2npbcdim(fr->ePBC) || (bDomDec && dd->nc[d] > 1))
2040         {
2041             shp[d] = 0;
2042         }
2043         else
2044         {
2045             if (d == XX &&
2046                 box[XX][XX] - fabs(box[YY][XX]) - fabs(box[ZZ][XX]) < sqrt(rl2))
2047             {
2048                 shp[d] = 2;
2049             }
2050             else
2051             {
2052                 shp[d] = 1;
2053             }
2054         }
2055     }
2056
2057     /* Loop over charge groups */
2058     for(icg=cg0; (icg < cg1); icg++)
2059     {
2060         igid = GET_CGINFO_GID(cginfo[icg]);
2061         /* Skip this charge group if all energy groups are excluded! */
2062         if (bExcludeAlleg[igid])
2063         {
2064             continue;
2065         }
2066
2067         i0   = cgs->index[icg];
2068
2069         if (bMakeQMMMnblist)
2070         {
2071             /* Skip this charge group if it is not a QM atom while making a
2072              * QM/MM neighbourlist
2073              */
2074             if (md->bQM[i0]==FALSE)
2075             {
2076                 continue; /* MM particle, go to next particle */
2077             }
2078
2079             /* Compute the number of charge groups that fall within the control
2080              * of this one (icg)
2081              */
2082             naaj    = calc_naaj(icg,cgsnr);
2083             jcg0    = icg;
2084             jcg1    = icg + naaj;
2085             max_jcg = cgsnr;
2086         }
2087         else
2088         {
2089             /* make a normal neighbourlist */
2090
2091             if (bDomDec)
2092             {
2093                 /* Get the j charge-group and dd cell shift ranges */
2094                 dd_get_ns_ranges(cr->dd,icg,&jcg0,&jcg1,sh0,sh1);
2095                 max_jcg = 0;
2096             }
2097             else
2098             {
2099                 /* Compute the number of charge groups that fall within the control
2100                  * of this one (icg)
2101                  */
2102                 naaj = calc_naaj(icg,cgsnr);
2103                 jcg0 = icg;
2104                 jcg1 = icg + naaj;
2105
2106                 if (fr->n_tpi)
2107                 {
2108                     /* The i-particle is awlways the test particle,
2109                      * so we want all j-particles
2110                      */
2111                     max_jcg = cgsnr - 1;
2112                 }
2113                 else
2114                 {
2115                     max_jcg  = jcg1 - cgsnr;
2116                 }
2117             }
2118         }
2119
2120         i_egp_flags = fr->egp_flags + igid*ngid;
2121
2122         /* Set the exclusions for the atoms in charge group icg using a bitmask */
2123         setexcl(i0,cgs->index[icg+1],&top->excls,TRUE,bexcl);
2124
2125         ci2xyz(grid,icg,&cell_x,&cell_y,&cell_z);
2126
2127         /* Changed iicg to icg, DvdS 990115
2128          * (but see consistency check above, DvdS 990330)
2129          */
2130 #ifdef NS5DB
2131         fprintf(log,"icg=%5d, naaj=%5d, cell %d %d %d\n",
2132                 icg,naaj,cell_x,cell_y,cell_z);
2133 #endif
2134         /* Loop over shift vectors in three dimensions */
2135         for (tz=-shp[ZZ]; tz<=shp[ZZ]; tz++)
2136         {
2137             ZI = cgcm[icg][ZZ]+tz*box[ZZ][ZZ];
2138             /* Calculate range of cells in Z direction that have the shift tz */
2139             zgi = cell_z + tz*Nz;
2140 #define FAST_DD_NS
2141 #ifndef FAST_DD_NS
2142             get_dx(Nz,gridz,rl2,zgi,ZI,&dz0,&dz1,dcz2);
2143 #else
2144             get_dx_dd(Nz,gridz,rl2,zgi,ZI-grid_offset[ZZ],
2145                       ncpddc[ZZ],sh0[ZZ],sh1[ZZ],&dz0,&dz1,dcz2);
2146 #endif
2147             if (dz0 > dz1)
2148             {
2149                 continue;
2150             }
2151             for (ty=-shp[YY]; ty<=shp[YY]; ty++)
2152             {
2153                 YI = cgcm[icg][YY]+ty*box[YY][YY]+tz*box[ZZ][YY];
2154                 /* Calculate range of cells in Y direction that have the shift ty */
2155                 if (bTriclinicY)
2156                 {
2157                     ygi = (int)(Ny + (YI - grid_offset[YY])*grid_y) - Ny;
2158                 }
2159                 else
2160                 {
2161                     ygi = cell_y + ty*Ny;
2162                 }
2163 #ifndef FAST_DD_NS
2164                 get_dx(Ny,gridy,rl2,ygi,YI,&dy0,&dy1,dcy2);
2165 #else
2166                 get_dx_dd(Ny,gridy,rl2,ygi,YI-grid_offset[YY],
2167                           ncpddc[YY],sh0[YY],sh1[YY],&dy0,&dy1,dcy2);
2168 #endif
2169                 if (dy0 > dy1)
2170                 {
2171                     continue;
2172                 }
2173                 for (tx=-shp[XX]; tx<=shp[XX]; tx++)
2174                 {
2175                     XI = cgcm[icg][XX]+tx*box[XX][XX]+ty*box[YY][XX]+tz*box[ZZ][XX];
2176                     /* Calculate range of cells in X direction that have the shift tx */
2177                     if (bTriclinicX)
2178                     {
2179                         xgi = (int)(Nx + (XI - grid_offset[XX])*grid_x) - Nx;
2180                     }
2181                     else
2182                     {
2183                         xgi = cell_x + tx*Nx;
2184                     }
2185 #ifndef FAST_DD_NS
2186                     get_dx(Nx,gridx,rl2,xgi*Nx,XI,&dx0,&dx1,dcx2);
2187 #else
2188                     get_dx_dd(Nx,gridx,rl2,xgi,XI-grid_offset[XX],
2189                               ncpddc[XX],sh0[XX],sh1[XX],&dx0,&dx1,dcx2);
2190 #endif
2191                     if (dx0 > dx1)
2192                     {
2193                         continue;
2194                     }
2195                     /* Get shift vector */
2196                     shift=XYZ2IS(tx,ty,tz);
2197 #ifdef NS5DB
2198                     range_check(shift,0,SHIFTS);
2199 #endif
2200                     for(nn=0; (nn<ngid); nn++)
2201                     {
2202                         nsr[nn]      = 0;
2203                         nlr_ljc[nn]  = 0;
2204                         nlr_one[nn] = 0;
2205                     }
2206 #ifdef NS5DB
2207                     fprintf(log,"shift: %2d, dx0,1: %2d,%2d, dy0,1: %2d,%2d, dz0,1: %2d,%2d\n",
2208                             shift,dx0,dx1,dy0,dy1,dz0,dz1);
2209                     fprintf(log,"cgcm: %8.3f  %8.3f  %8.3f\n",cgcm[icg][XX],
2210                             cgcm[icg][YY],cgcm[icg][ZZ]);
2211                     fprintf(log,"xi:   %8.3f  %8.3f  %8.3f\n",XI,YI,ZI);
2212 #endif
2213                     for (dx=dx0; (dx<=dx1); dx++)
2214                     {
2215                         tmp1 = rl2 - dcx2[dx];
2216                         for (dy=dy0; (dy<=dy1); dy++)
2217                         {
2218                             tmp2 = tmp1 - dcy2[dy];
2219                             if (tmp2 > 0)
2220                             {
2221                                 for (dz=dz0; (dz<=dz1); dz++) {
2222                                     if (tmp2 > dcz2[dz]) {
2223                                         /* Find grid-cell cj in which possible neighbours are */
2224                                         cj   = xyz2ci(Ny,Nz,dx,dy,dz);
2225
2226                                         /* Check out how many cgs (nrj) there in this cell */
2227                                         nrj  = gridnra[cj];
2228
2229                                         /* Find the offset in the cg list */
2230                                         cgj0 = gridind[cj];
2231
2232                                         /* Check if all j's are out of range so we
2233                                          * can skip the whole cell.
2234                                          * Should save some time, especially with DD.
2235                                          */
2236                                         if (nrj == 0 ||
2237                                             (grida[cgj0] >= max_jcg &&
2238                                              (grida[cgj0] >= jcg1 || grida[cgj0+nrj-1] < jcg0)))
2239                                         {
2240                                             continue;
2241                                         }
2242
2243                                         /* Loop over cgs */
2244                                         for (j=0; (j<nrj); j++)
2245                                         {
2246                                             jjcg = grida[cgj0+j];
2247
2248                                             /* check whether this guy is in range! */
2249                                             if ((jjcg >= jcg0 && jjcg < jcg1) ||
2250                                                 (jjcg < max_jcg))
2251                                             {
2252                                                 r2=calc_dx2(XI,YI,ZI,cgcm[jjcg]);
2253                                                 if (r2 < rl2) {
2254                                                     /* jgid = gid[cgsatoms[cgsindex[jjcg]]]; */
2255                                                     jgid = GET_CGINFO_GID(cginfo[jjcg]);
2256                                                     /* check energy group exclusions */
2257                                                     if (!(i_egp_flags[jgid] & EGP_EXCL))
2258                                                     {
2259                                                         if (r2 < rs2)
2260                                                         {
2261                                                             if (nsr[jgid] >= MAX_CG)
2262                                                             {
2263                                                                 put_in_list(bHaveVdW,ngid,md,icg,jgid,
2264                                                                             nsr[jgid],nl_sr[jgid],
2265                                                                             cgs->index,/* cgsatoms, */ bexcl,
2266                                                                             shift,fr,FALSE,TRUE,TRUE);
2267                                                                 nsr[jgid]=0;
2268                                                             }
2269                                                             nl_sr[jgid][nsr[jgid]++]=jjcg;
2270                                                         }
2271                                                         else if (r2 < rm2)
2272                                                         {
2273                                                             if (nlr_ljc[jgid] >= MAX_CG)
2274                                                             {
2275                                                                 do_longrange(cr,top,fr,ngid,md,icg,jgid,
2276                                                                              nlr_ljc[jgid],
2277                                                                              nl_lr_ljc[jgid],bexcl,shift,x,
2278                                                                              box_size,nrnb,
2279                                                                              lambda,dvdlambda,
2280                                                                              grppener,
2281                                                                              TRUE,TRUE,FALSE,
2282                                                                              put_in_list,
2283                                                                              bHaveVdW,
2284                                                                              bDoForces,f);
2285                                                                 nlr_ljc[jgid]=0;
2286                                                             }
2287                                                             nl_lr_ljc[jgid][nlr_ljc[jgid]++]=jjcg;
2288                                                         }
2289                                                         else
2290                                                         {
2291                                                             if (nlr_one[jgid] >= MAX_CG) {
2292                                                                 do_longrange(cr,top,fr,ngid,md,icg,jgid,
2293                                                                              nlr_one[jgid],
2294                                                                              nl_lr_one[jgid],bexcl,shift,x,
2295                                                                              box_size,nrnb,
2296                                                                              lambda,dvdlambda,
2297                                                                              grppener,
2298                                                                              rvdw_lt_rcoul,rcoul_lt_rvdw,FALSE,
2299                                                                              put_in_list,
2300                                                                              bHaveVdW,
2301                                                                              bDoForces,f);
2302                                                                 nlr_one[jgid]=0;
2303                                                             }
2304                                                             nl_lr_one[jgid][nlr_one[jgid]++]=jjcg;
2305                                                         }
2306                                                     }
2307                                                 }
2308                                                 nns++;
2309                                             }
2310                                         }
2311                                     }
2312                                 }
2313                             }
2314                         }
2315                     }
2316                     /* CHECK whether there is anything left in the buffers */
2317                     for(nn=0; (nn<ngid); nn++)
2318                     {
2319                         if (nsr[nn] > 0)
2320                         {
2321                             put_in_list(bHaveVdW,ngid,md,icg,nn,nsr[nn],nl_sr[nn],
2322                                         cgs->index, /* cgsatoms, */ bexcl,
2323                                         shift,fr,FALSE,TRUE,TRUE);
2324                         }
2325
2326                         if (nlr_ljc[nn] > 0)
2327                         {
2328                             do_longrange(cr,top,fr,ngid,md,icg,nn,nlr_ljc[nn],
2329                                          nl_lr_ljc[nn],bexcl,shift,x,box_size,nrnb,
2330                                          lambda,dvdlambda,grppener,TRUE,TRUE,FALSE,
2331                                          put_in_list,bHaveVdW,bDoForces,f);
2332                         }
2333
2334                         if (nlr_one[nn] > 0)
2335                         {
2336                             do_longrange(cr,top,fr,ngid,md,icg,nn,nlr_one[nn],
2337                                          nl_lr_one[nn],bexcl,shift,x,box_size,nrnb,
2338                                          lambda,dvdlambda,grppener,
2339                                          rvdw_lt_rcoul,rcoul_lt_rvdw,FALSE,
2340                                          put_in_list,bHaveVdW,bDoForces,f);
2341                         }
2342                     }
2343                 }
2344             }
2345         }
2346         /* setexcl(nri,i_atoms,&top->atoms.excl,FALSE,bexcl); */
2347         setexcl(cgs->index[icg],cgs->index[icg+1],&top->excls,FALSE,bexcl);
2348     }
2349     /* Perform any left over force calculations */
2350     for (nn=0; (nn<ngid); nn++)
2351     {
2352         if (rm2 > rs2)
2353         {
2354             do_longrange(cr,top,fr,0,md,icg,nn,nlr_ljc[nn],
2355                          nl_lr_ljc[nn],bexcl,shift,x,box_size,nrnb,
2356                          lambda,dvdlambda,grppener,
2357                          TRUE,TRUE,TRUE,put_in_list,bHaveVdW,bDoForces,f);
2358         }
2359         if (rl2 > rm2) {
2360             do_longrange(cr,top,fr,0,md,icg,nn,nlr_one[nn],
2361                          nl_lr_one[nn],bexcl,shift,x,box_size,nrnb,
2362                          lambda,dvdlambda,grppener,
2363                          rvdw_lt_rcoul,rcoul_lt_rvdw,
2364                          TRUE,put_in_list,bHaveVdW,bDoForces,f);
2365         }
2366     }
2367     debug_gmx();
2368
2369     /* Close off short range neighbourlists */
2370     close_neighbor_list(fr,FALSE,-1,-1,bMakeQMMMnblist);
2371
2372     return nns;
2373 }
2374
2375 void ns_realloc_natoms(gmx_ns_t *ns,int natoms)
2376 {
2377     int i;
2378
2379     if (natoms > ns->nra_alloc)
2380     {
2381         ns->nra_alloc = over_alloc_dd(natoms);
2382         srenew(ns->bexcl,ns->nra_alloc);
2383         for(i=0; i<ns->nra_alloc; i++)
2384         {
2385             ns->bexcl[i] = 0;
2386         }
2387     }
2388 }
2389
2390 void init_ns(FILE *fplog,const t_commrec *cr,
2391              gmx_ns_t *ns,t_forcerec *fr,
2392              const gmx_mtop_t *mtop,
2393              matrix box)
2394 {
2395     int  mt,icg,nr_in_cg,maxcg,i,j,jcg,ngid,ncg;
2396     t_block *cgs;
2397     char *ptr;
2398
2399     /* Compute largest charge groups size (# atoms) */
2400     nr_in_cg=1;
2401     for(mt=0; mt<mtop->nmoltype; mt++) {
2402         cgs = &mtop->moltype[mt].cgs;
2403         for (icg=0; (icg < cgs->nr); icg++)
2404         {
2405             nr_in_cg=max(nr_in_cg,(int)(cgs->index[icg+1]-cgs->index[icg]));
2406         }
2407     }
2408
2409     /* Verify whether largest charge group is <= max cg.
2410      * This is determined by the type of the local exclusion type
2411      * Exclusions are stored in bits. (If the type is not large
2412      * enough, enlarge it, unsigned char -> unsigned short -> unsigned long)
2413      */
2414     maxcg = sizeof(t_excl)*8;
2415     if (nr_in_cg > maxcg)
2416     {
2417         gmx_fatal(FARGS,"Max #atoms in a charge group: %d > %d\n",
2418                   nr_in_cg,maxcg);
2419     }
2420
2421     ngid = mtop->groups.grps[egcENER].nr;
2422     snew(ns->bExcludeAlleg,ngid);
2423     for(i=0; i<ngid; i++) {
2424         ns->bExcludeAlleg[i] = TRUE;
2425         for(j=0; j<ngid; j++)
2426         {
2427             if (!(fr->egp_flags[i*ngid+j] & EGP_EXCL))
2428             {
2429                 ns->bExcludeAlleg[i] = FALSE;
2430             }
2431         }
2432     }
2433
2434     if (fr->bGrid) {
2435         /* Grid search */
2436         ns->grid = init_grid(fplog,fr);
2437         init_nsgrid_lists(fr,ngid,ns);
2438     }
2439     else
2440     {
2441         /* Simple search */
2442         snew(ns->ns_buf,ngid);
2443         for(i=0; (i<ngid); i++)
2444         {
2445             snew(ns->ns_buf[i],SHIFTS);
2446         }
2447         ncg = ncg_mtop(mtop);
2448         snew(ns->simple_aaj,2*ncg);
2449         for(jcg=0; (jcg<ncg); jcg++)
2450         {
2451             ns->simple_aaj[jcg]     = jcg;
2452             ns->simple_aaj[jcg+ncg] = jcg;
2453         }
2454     }
2455
2456     /* Create array that determines whether or not atoms have VdW */
2457     snew(ns->bHaveVdW,fr->ntype);
2458     for(i=0; (i<fr->ntype); i++)
2459     {
2460         for(j=0; (j<fr->ntype); j++)
2461         {
2462             ns->bHaveVdW[i] = (ns->bHaveVdW[i] ||
2463                                (fr->bBHAM ?
2464                                 ((BHAMA(fr->nbfp,fr->ntype,i,j) != 0) ||
2465                                  (BHAMB(fr->nbfp,fr->ntype,i,j) != 0) ||
2466                                  (BHAMC(fr->nbfp,fr->ntype,i,j) != 0)) :
2467                                 ((C6(fr->nbfp,fr->ntype,i,j) != 0) ||
2468                                  (C12(fr->nbfp,fr->ntype,i,j) != 0))));
2469         }
2470     }
2471     if (debug)
2472         pr_bvec(debug,0,"bHaveVdW",ns->bHaveVdW,fr->ntype,TRUE);
2473
2474     ns->nra_alloc = 0;
2475     ns->bexcl = NULL;
2476     if (!DOMAINDECOMP(cr))
2477     {
2478         /* This could be reduced with particle decomposition */
2479         ns_realloc_natoms(ns,mtop->natoms);
2480     }
2481
2482     ns->nblist_initialized=FALSE;
2483
2484     /* nbr list debug dump */
2485     {
2486         char *ptr=getenv("GMX_DUMP_NL");
2487         if (ptr)
2488         {
2489             ns->dump_nl=strtol(ptr,NULL,10);
2490             if (fplog)
2491             {
2492                 fprintf(fplog, "GMX_DUMP_NL = %d", ns->dump_nl);
2493             }
2494         }
2495         else
2496         {
2497             ns->dump_nl=0;
2498         }
2499     }
2500 }
2501
2502
2503 int search_neighbours(FILE *log,t_forcerec *fr,
2504                       rvec x[],matrix box,
2505                       gmx_localtop_t *top,
2506                       gmx_groups_t *groups,
2507                       t_commrec *cr,
2508                       t_nrnb *nrnb,t_mdatoms *md,
2509                       real lambda,real *dvdlambda,
2510                       gmx_grppairener_t *grppener,
2511                       gmx_bool bFillGrid,
2512                       gmx_bool bDoLongRange,
2513                       gmx_bool bDoForces,rvec *f)
2514 {
2515     t_block  *cgs=&(top->cgs);
2516     rvec     box_size,grid_x0,grid_x1;
2517     int      i,j,m,ngid;
2518     real     min_size,grid_dens;
2519     int      nsearch;
2520     gmx_bool     bGrid;
2521     char     *ptr;
2522     gmx_bool     *i_egp_flags;
2523     int      cg_start,cg_end,start,end;
2524     gmx_ns_t *ns;
2525     t_grid   *grid;
2526     gmx_domdec_zones_t *dd_zones;
2527     put_in_list_t *put_in_list;
2528
2529     ns = &fr->ns;
2530
2531     /* Set some local variables */
2532     bGrid = fr->bGrid;
2533     ngid = groups->grps[egcENER].nr;
2534
2535     for(m=0; (m<DIM); m++)
2536     {
2537         box_size[m] = box[m][m];
2538     }
2539
2540     if (fr->ePBC != epbcNONE)
2541     {
2542         if (sqr(fr->rlistlong) >= max_cutoff2(fr->ePBC,box))
2543         {
2544             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.");
2545         }
2546         if (!bGrid)
2547         {
2548             min_size = min(box_size[XX],min(box_size[YY],box_size[ZZ]));
2549             if (2*fr->rlistlong >= min_size)
2550                 gmx_fatal(FARGS,"One of the box diagonal elements has become smaller than twice the cut-off length.");
2551         }
2552     }
2553
2554     if (DOMAINDECOMP(cr))
2555     {
2556         ns_realloc_natoms(ns,cgs->index[cgs->nr]);
2557     }
2558     debug_gmx();
2559
2560     /* Reset the neighbourlists */
2561     reset_neighbor_list(fr,FALSE,-1,-1);
2562
2563     if (bGrid && bFillGrid)
2564     {
2565
2566         grid = ns->grid;
2567         if (DOMAINDECOMP(cr))
2568         {
2569             dd_zones = domdec_zones(cr->dd);
2570         }
2571         else
2572         {
2573             dd_zones = NULL;
2574
2575             get_nsgrid_boundaries(grid,NULL,box,NULL,NULL,NULL,
2576                                   cgs->nr,fr->cg_cm,grid_x0,grid_x1,&grid_dens);
2577
2578             grid_first(log,grid,NULL,NULL,fr->ePBC,box,grid_x0,grid_x1,
2579                        fr->rlistlong,grid_dens);
2580         }
2581         debug_gmx();
2582
2583         /* Don't know why this all is... (DvdS 3/99) */
2584 #ifndef SEGV
2585         start = 0;
2586         end   = cgs->nr;
2587 #else
2588         start = fr->cg0;
2589         end   = (cgs->nr+1)/2;
2590 #endif
2591
2592         if (DOMAINDECOMP(cr))
2593         {
2594             end = cgs->nr;
2595             fill_grid(log,dd_zones,grid,end,-1,end,fr->cg_cm);
2596             grid->icg0 = 0;
2597             grid->icg1 = dd_zones->izone[dd_zones->nizone-1].cg1;
2598         }
2599         else
2600         {
2601             fill_grid(log,NULL,grid,cgs->nr,fr->cg0,fr->hcg,fr->cg_cm);
2602             grid->icg0 = fr->cg0;
2603             grid->icg1 = fr->hcg;
2604             debug_gmx();
2605
2606             if (PARTDECOMP(cr))
2607                 mv_grid(cr,grid);
2608             debug_gmx();
2609         }
2610
2611         calc_elemnr(log,grid,start,end,cgs->nr);
2612         calc_ptrs(grid);
2613         grid_last(log,grid,start,end,cgs->nr);
2614
2615         if (gmx_debug_at)
2616         {
2617             check_grid(debug,grid);
2618             print_grid(debug,grid);
2619         }
2620     }
2621     else if (fr->n_tpi)
2622     {
2623         /* Set the grid cell index for the test particle only.
2624          * The cell to cg index is not corrected, but that does not matter.
2625          */
2626         fill_grid(log,NULL,ns->grid,fr->hcg,fr->hcg-1,fr->hcg,fr->cg_cm);
2627     }
2628     debug_gmx();
2629
2630     if (!fr->ns.bCGlist)
2631     {
2632         put_in_list = put_in_list_at;
2633     }
2634     else
2635     {
2636         put_in_list = put_in_list_cg;
2637     }
2638
2639     /* Do the core! */
2640     if (bGrid)
2641     {
2642         grid = ns->grid;
2643         nsearch = nsgrid_core(log,cr,fr,box,box_size,ngid,top,
2644                               grid,x,ns->bexcl,ns->bExcludeAlleg,
2645                               nrnb,md,lambda,dvdlambda,grppener,
2646                               put_in_list,ns->bHaveVdW,
2647                               bDoLongRange,bDoForces,f,
2648                               FALSE);
2649
2650         /* neighbour searching withouth QMMM! QM atoms have zero charge in
2651          * the classical calculation. The charge-charge interaction
2652          * between QM and MM atoms is handled in the QMMM core calculation
2653          * (see QMMM.c). The VDW however, we'd like to compute classically
2654          * and the QM MM atom pairs have just been put in the
2655          * corresponding neighbourlists. in case of QMMM we still need to
2656          * fill a special QMMM neighbourlist that contains all neighbours
2657          * of the QM atoms. If bQMMM is true, this list will now be made:
2658          */
2659         if (fr->bQMMM && fr->qr->QMMMscheme!=eQMMMschemeoniom)
2660         {
2661             nsearch += nsgrid_core(log,cr,fr,box,box_size,ngid,top,
2662                                    grid,x,ns->bexcl,ns->bExcludeAlleg,
2663                                    nrnb,md,lambda,dvdlambda,grppener,
2664                                    put_in_list_qmmm,ns->bHaveVdW,
2665                                    bDoLongRange,bDoForces,f,
2666                                    TRUE);
2667         }
2668     }
2669     else
2670     {
2671         nsearch = ns_simple_core(fr,top,md,box,box_size,
2672                                  ns->bexcl,ns->simple_aaj,
2673                                  ngid,ns->ns_buf,put_in_list,ns->bHaveVdW);
2674     }
2675     debug_gmx();
2676
2677 #ifdef DEBUG
2678     pr_nsblock(log);
2679 #endif
2680
2681     inc_nrnb(nrnb,eNR_NS,nsearch);
2682     /* inc_nrnb(nrnb,eNR_LR,fr->nlr); */
2683
2684     return nsearch;
2685 }
2686
2687 int natoms_beyond_ns_buffer(t_inputrec *ir,t_forcerec *fr,t_block *cgs,
2688                             matrix scale_tot,rvec *x)
2689 {
2690     int  cg0,cg1,cg,a0,a1,a,i,j;
2691     real rint,hbuf2,scale;
2692     rvec *cg_cm,cgsc;
2693     gmx_bool bIsotropic;
2694     int  nBeyond;
2695
2696     nBeyond = 0;
2697
2698     rint = max(ir->rcoulomb,ir->rvdw);
2699     if (ir->rlist < rint)
2700     {
2701         gmx_fatal(FARGS,"The neighbor search buffer has negative size: %f nm",
2702                   ir->rlist - rint);
2703     }
2704     cg_cm = fr->cg_cm;
2705
2706     cg0 = fr->cg0;
2707     cg1 = fr->hcg;
2708
2709     if (!EI_DYNAMICS(ir->eI) || !DYNAMIC_BOX(*ir))
2710     {
2711         hbuf2 = sqr(0.5*(ir->rlist - rint));
2712         for(cg=cg0; cg<cg1; cg++)
2713         {
2714             a0 = cgs->index[cg];
2715             a1 = cgs->index[cg+1];
2716             for(a=a0; a<a1; a++)
2717             {
2718                 if (distance2(cg_cm[cg],x[a]) > hbuf2)
2719                 {
2720                     nBeyond++;
2721                 }
2722             }
2723         }
2724     }
2725     else
2726     {
2727         bIsotropic = TRUE;
2728         scale = scale_tot[0][0];
2729         for(i=1; i<DIM; i++)
2730         {
2731             /* With anisotropic scaling, the original spherical ns volumes become
2732              * ellipsoids. To avoid costly transformations we use the minimum
2733              * eigenvalue of the scaling matrix for determining the buffer size.
2734              * Since the lower half is 0, the eigenvalues are the diagonal elements.
2735              */
2736             scale = min(scale,scale_tot[i][i]);
2737             if (scale_tot[i][i] != scale_tot[i-1][i-1])
2738             {
2739                 bIsotropic = FALSE;
2740             }
2741             for(j=0; j<i; j++)
2742             {
2743                 if (scale_tot[i][j] != 0)
2744                 {
2745                     bIsotropic = FALSE;
2746                 }
2747             }
2748         }
2749         hbuf2 = sqr(0.5*(scale*ir->rlist - rint));
2750         if (bIsotropic)
2751         {
2752             for(cg=cg0; cg<cg1; cg++)
2753             {
2754                 svmul(scale,cg_cm[cg],cgsc);
2755                 a0 = cgs->index[cg];
2756                 a1 = cgs->index[cg+1];
2757                 for(a=a0; a<a1; a++)
2758                 {
2759                     if (distance2(cgsc,x[a]) > hbuf2)
2760                     {
2761                         nBeyond++;
2762                     }
2763                 }
2764             }
2765         }
2766         else
2767         {
2768             /* Anistropic scaling */
2769             for(cg=cg0; cg<cg1; cg++)
2770             {
2771                 /* Since scale_tot contains the transpose of the scaling matrix,
2772                  * we need to multiply with the transpose.
2773                  */
2774                 tmvmul_ur0(scale_tot,cg_cm[cg],cgsc);
2775                 a0 = cgs->index[cg];
2776                 a1 = cgs->index[cg+1];
2777                 for(a=a0; a<a1; a++)
2778                 {
2779                     if (distance2(cgsc,x[a]) > hbuf2)
2780                     {
2781                         nBeyond++;
2782                     }
2783                 }
2784             }
2785         }
2786     }
2787
2788     return nBeyond;
2789 }