src/gromacs/gmxlib/nonbonded/nb_generic.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * Copyright (c) 2012,2014, by the GROMACS development team, led by
   7  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   8  * and including many others, as listed in the AUTHORS file in the
   9  * top-level source directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37 #include "gmxpre.h"
  38
  39 #include "config.h"
  40
  41 #include <math.h>
  42
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/typedefs.h"
  46 #include "nb_generic.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/utility/fatalerror.h"
  50
  51 #include "gromacs/legacyheaders/nonbonded.h"
  52 #include "nb_kernel.h"
  53
  54 void
  55 gmx_nb_generic_kernel(t_nblist *                nlist,
  56                       rvec *                    xx,
  57                       rvec *                    ff,
  58                       t_forcerec *              fr,
  59                       t_mdatoms *               mdatoms,
  60                       nb_kernel_data_t *        kernel_data,
  61                       t_nrnb *                  nrnb)
  62 {
  63     int           nri, ntype, table_nelements, ielec, ivdw;
  64     real          facel, gbtabscale;
  65     int           n, ii, is3, ii3, k, nj0, nj1, jnr, j3, ggid, nnn, n0;
  66     real          shX, shY, shZ;
  67     real          fscal, felec, fvdw, velec, vvdw, tx, ty, tz;
  68     real          rinvsq;
  69     real          iq;
  70     real          qq, vctot;
  71     int           nti, nvdwparam;
  72     int           tj;
  73     real          rt, r, eps, eps2, Y, F, Geps, Heps2, VV, FF, Fp, fijD, fijR;
  74     real          rinvsix;
  75     real          vvdwtot;
  76     real          vvdw_rep, vvdw_disp;
  77     real          ix, iy, iz, fix, fiy, fiz;
  78     real          jx, jy, jz;
  79     real          dx, dy, dz, rsq, rinv;
  80     real          c6, c12, c6grid, cexp1, cexp2, br;
  81     real *        charge;
  82     real *        shiftvec;
  83     real *        vdwparam, *vdwgridparam;
  84     int *         shift;
  85     int *         type;
  86     real *        fshift;
  87     real *        velecgrp;
  88     real *        vvdwgrp;
  89     real          tabscale;
  90     real *        VFtab;
  91     real *        x;
  92     real *        f;
  93     int           ewitab;
  94     real          ewtabscale, eweps, sh_ewald, ewrt, ewtabhalfspace;
  95     real *        ewtab;
  96     real          rcoulomb2, rvdw, rvdw2, sh_dispersion, sh_repulsion;
  97     real          rcutoff, rcutoff2;
  98     real          rswitch_elec, rswitch_vdw, d, d2, sw, dsw, rinvcorr;
  99     real          elec_swV3, elec_swV4, elec_swV5, elec_swF2, elec_swF3, elec_swF4;
 100     real          vdw_swV3, vdw_swV4, vdw_swV5, vdw_swF2, vdw_swF3, vdw_swF4;
 101     real          ewclj, ewclj2, ewclj6, ewcljrsq, poly, exponent, sh_lj_ewald;
 102     gmx_bool      bExactElecCutoff, bExactVdwCutoff, bExactCutoff;
 103
 104     x                   = xx[0];
 105     f                   = ff[0];
 106     ielec               = nlist->ielec;
 107     ivdw                = nlist->ivdw;
 108
 109     fshift              = fr->fshift[0];
 110     velecgrp            = kernel_data->energygrp_elec;
 111     vvdwgrp             = kernel_data->energygrp_vdw;
 112     tabscale            = kernel_data->table_elec_vdw->scale;
 113     VFtab               = kernel_data->table_elec_vdw->data;
 114
 115     sh_ewald            = fr->ic->sh_ewald;
 116     ewtab               = fr->ic->tabq_coul_FDV0;
 117     ewtabscale          = fr->ic->tabq_scale;
 118     ewtabhalfspace      = 0.5/ewtabscale;
 119
 120     rcoulomb2           = fr->rcoulomb*fr->rcoulomb;
 121     rvdw                = fr->rvdw;
 122     rvdw2               = rvdw*rvdw;
 123     sh_dispersion       = fr->ic->dispersion_shift.cpot;
 124     sh_repulsion        = fr->ic->repulsion_shift.cpot;
 125     sh_lj_ewald         = fr->ic->sh_lj_ewald;
 126
 127     ewclj               = fr->ewaldcoeff_lj;
 128     ewclj2              = ewclj*ewclj;
 129     ewclj6              = ewclj2*ewclj2*ewclj2;
 130
 131     if (fr->coulomb_modifier == eintmodPOTSWITCH)
 132     {
 133         d               = fr->rcoulomb-fr->rcoulomb_switch;
 134         elec_swV3       = -10.0/(d*d*d);
 135         elec_swV4       =  15.0/(d*d*d*d);
 136         elec_swV5       =  -6.0/(d*d*d*d*d);
 137         elec_swF2       = -30.0/(d*d*d);
 138         elec_swF3       =  60.0/(d*d*d*d);
 139         elec_swF4       = -30.0/(d*d*d*d*d);
 140     }
 141     else
 142     {
 143         /* Avoid warnings from stupid compilers (looking at you, Clang!) */
 144         elec_swV3 = elec_swV4 = elec_swV5 = elec_swF2 = elec_swF3 = elec_swF4 = 0.0;
 145     }
 146     if (fr->vdw_modifier == eintmodPOTSWITCH)
 147     {
 148         d               = fr->rvdw-fr->rvdw_switch;
 149         vdw_swV3        = -10.0/(d*d*d);
 150         vdw_swV4        =  15.0/(d*d*d*d);
 151         vdw_swV5        =  -6.0/(d*d*d*d*d);
 152         vdw_swF2        = -30.0/(d*d*d);
 153         vdw_swF3        =  60.0/(d*d*d*d);
 154         vdw_swF4        = -30.0/(d*d*d*d*d);
 155     }
 156     else
 157     {
 158         /* Avoid warnings from stupid compilers (looking at you, Clang!) */
 159         vdw_swV3 = vdw_swV4 = vdw_swV5 = vdw_swF2 = vdw_swF3 = vdw_swF4 = 0.0;
 160     }
 161
 162     bExactElecCutoff    = (fr->coulomb_modifier != eintmodNONE) || fr->eeltype == eelRF_ZERO;
 163     bExactVdwCutoff     = (fr->vdw_modifier != eintmodNONE);
 164     bExactCutoff        = bExactElecCutoff && bExactVdwCutoff;
 165
 166     if (bExactCutoff)
 167     {
 168         rcutoff  = ( fr->rcoulomb > fr->rvdw ) ? fr->rcoulomb : fr->rvdw;
 169         rcutoff2 = rcutoff*rcutoff;
 170     }
 171     else
 172     {
 173         /* Fix warnings for stupid compilers */
 174         rcutoff = rcutoff2 = 1e30;
 175     }
 176
 177     /* avoid compiler warnings for cases that cannot happen */
 178     nnn                 = 0;
 179     eps                 = 0.0;
 180     eps2                = 0.0;
 181
 182     /* 3 VdW parameters for Buckingham, otherwise 2 */
 183     nvdwparam           = (ivdw == GMX_NBKERNEL_VDW_BUCKINGHAM) ? 3 : 2;
 184     table_nelements     = 12;
 185
 186     charge              = mdatoms->chargeA;
 187     type                = mdatoms->typeA;
 188     facel               = fr->epsfac;
 189     shiftvec            = fr->shift_vec[0];
 190     vdwparam            = fr->nbfp;
 191     ntype               = fr->ntype;
 192     vdwgridparam        = fr->ljpme_c6grid;
 193
 194     for (n = 0; (n < nlist->nri); n++)
 195     {
 196         is3              = 3*nlist->shift[n];
 197         shX              = shiftvec[is3];
 198         shY              = shiftvec[is3+1];
 199         shZ              = shiftvec[is3+2];
 200         nj0              = nlist->jindex[n];
 201         nj1              = nlist->jindex[n+1];
 202         ii               = nlist->iinr[n];
 203         ii3              = 3*ii;
 204         ix               = shX + x[ii3+0];
 205         iy               = shY + x[ii3+1];
 206         iz               = shZ + x[ii3+2];
 207         iq               = facel*charge[ii];
 208         nti              = nvdwparam*ntype*type[ii];
 209         vctot            = 0;
 210         vvdwtot          = 0;
 211         fix              = 0;
 212         fiy              = 0;
 213         fiz              = 0;
 214
 215         for (k = nj0; (k < nj1); k++)
 216         {
 217             jnr              = nlist->jjnr[k];
 218             j3               = 3*jnr;
 219             jx               = x[j3+0];
 220             jy               = x[j3+1];
 221             jz               = x[j3+2];
 222             dx               = ix - jx;
 223             dy               = iy - jy;
 224             dz               = iz - jz;
 225             rsq              = dx*dx+dy*dy+dz*dz;
 226             rinv             = gmx_invsqrt(rsq);
 227             rinvsq           = rinv*rinv;
 228             felec            = 0;
 229             fvdw             = 0;
 230             velec            = 0;
 231             vvdw             = 0;
 232
 233             if (bExactCutoff && rsq >= rcutoff2)
 234             {
 235                 continue;
 236             }
 237
 238             if (ielec == GMX_NBKERNEL_ELEC_CUBICSPLINETABLE || ivdw == GMX_NBKERNEL_VDW_CUBICSPLINETABLE)
 239             {
 240                 r                = rsq*rinv;
 241                 rt               = r*tabscale;
 242                 n0               = rt;
 243                 eps              = rt-n0;
 244                 eps2             = eps*eps;
 245                 nnn              = table_nelements*n0;
 246             }
 247
 248             /* Coulomb interaction. ielec==0 means no interaction */
 249             if (ielec != GMX_NBKERNEL_ELEC_NONE)
 250             {
 251                 qq               = iq*charge[jnr];
 252
 253                 switch (ielec)
 254                 {
 255                     case GMX_NBKERNEL_ELEC_NONE:
 256                         break;
 257
 258                     case GMX_NBKERNEL_ELEC_COULOMB:
 259                         /* Vanilla cutoff coulomb */
 260                         velec            = qq*rinv;
 261                         felec            = velec*rinvsq;
 262                         /* The shift for the Coulomb potential is stored in
 263                          * the RF parameter c_rf, which is 0 without shift
 264                          */
 265                         velec           -= qq*fr->ic->c_rf;
 266                         break;
 267
 268                     case GMX_NBKERNEL_ELEC_REACTIONFIELD:
 269                         /* Reaction-field */
 270                         velec            = qq*(rinv+fr->k_rf*rsq-fr->c_rf);
 271                         felec            = qq*(rinv*rinvsq-2.0*fr->k_rf);
 272                         break;
 273
 274                     case GMX_NBKERNEL_ELEC_CUBICSPLINETABLE:
 275                         /* Tabulated coulomb */
 276                         Y                = VFtab[nnn];
 277                         F                = VFtab[nnn+1];
 278                         Geps             = eps*VFtab[nnn+2];
 279                         Heps2            = eps2*VFtab[nnn+3];
 280                         Fp               = F+Geps+Heps2;
 281                         VV               = Y+eps*Fp;
 282                         FF               = Fp+Geps+2.0*Heps2;
 283                         velec            = qq*VV;
 284                         felec            = -qq*FF*tabscale*rinv;
 285                         break;
 286
 287                     case GMX_NBKERNEL_ELEC_GENERALIZEDBORN:
 288                         /* GB */
 289                         gmx_fatal(FARGS, "Death & horror! GB generic interaction not implemented.\n");
 290                         break;
 291
 292                     case GMX_NBKERNEL_ELEC_EWALD:
 293                         ewrt             = rsq*rinv*ewtabscale;
 294                         ewitab           = ewrt;
 295                         eweps            = ewrt-ewitab;
 296                         ewitab           = 4*ewitab;
 297                         felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 298                         rinvcorr         = (fr->coulomb_modifier == eintmodPOTSHIFT) ? rinv-fr->ic->sh_ewald : rinv;
 299                         velec            = qq*(rinvcorr-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 300                         felec            = qq*rinv*(rinvsq-felec);
 301                         break;
 302
 303                     default:
 304                         gmx_fatal(FARGS, "Death & horror! No generic coulomb interaction for ielec=%d.\n", ielec);
 305                         break;
 306                 }
 307                 if (fr->coulomb_modifier == eintmodPOTSWITCH)
 308                 {
 309                     d                = rsq*rinv-fr->rcoulomb_switch;
 310                     d                = (d > 0.0) ? d : 0.0;
 311                     d2               = d*d;
 312                     sw               = 1.0+d2*d*(elec_swV3+d*(elec_swV4+d*elec_swV5));
 313                     dsw              = d2*(elec_swF2+d*(elec_swF3+d*elec_swF4));
 314                     /* Apply switch function. Note that felec=f/r since it will be multiplied
 315                      * by the i-j displacement vector. This means felec'=f'/r=-(v*sw)'/r=
 316                      * -(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=felec*sw-v*dsw/r
 317                      */
 318                     felec            = felec*sw - rinv*velec*dsw;
 319                     /* Once we have used velec to update felec we can modify velec too */
 320                     velec           *= sw;
 321                 }
 322                 if (bExactElecCutoff)
 323                 {
 324                     felec            = (rsq < rcoulomb2) ? felec : 0.0;
 325                     velec            = (rsq < rcoulomb2) ? velec : 0.0;
 326                 }
 327                 vctot           += velec;
 328             } /* End of coulomb interactions */
 329
 330
 331             /* VdW interaction. ivdw==0 means no interaction */
 332             if (ivdw != GMX_NBKERNEL_VDW_NONE)
 333             {
 334                 tj               = nti+nvdwparam*type[jnr];
 335
 336                 switch (ivdw)
 337                 {
 338                     case GMX_NBKERNEL_VDW_NONE:
 339                         break;
 340
 341                     case GMX_NBKERNEL_VDW_LENNARDJONES:
 342                         /* Vanilla Lennard-Jones cutoff */
 343                         c6               = vdwparam[tj];
 344                         c12              = vdwparam[tj+1];
 345                         rinvsix          = rinvsq*rinvsq*rinvsq;
 346                         vvdw_disp        = c6*rinvsix;
 347                         vvdw_rep         = c12*rinvsix*rinvsix;
 348                         fvdw             = (vvdw_rep-vvdw_disp)*rinvsq;
 349                         if (fr->vdw_modifier == eintmodPOTSHIFT)
 350                         {
 351                             vvdw             = (vvdw_rep + c12*sh_repulsion)/12.0 - (vvdw_disp + c6*sh_dispersion)/6.0;
 352                         }
 353                         else
 354                         {
 355                             vvdw             = vvdw_rep/12.0-vvdw_disp/6.0;
 356                         }
 357                         break;
 358
 359                     case GMX_NBKERNEL_VDW_BUCKINGHAM:
 360                         /* Buckingham */
 361                         c6               = vdwparam[tj];
 362                         cexp1            = vdwparam[tj+1];
 363                         cexp2            = vdwparam[tj+2];
 364
 365                         rinvsix          = rinvsq*rinvsq*rinvsq;
 366                         vvdw_disp        = c6*rinvsix;
 367                         br               = cexp2*rsq*rinv;
 368                         vvdw_rep         = cexp1*exp(-br);
 369                         fvdw             = (br*vvdw_rep-vvdw_disp)*rinvsq;
 370                         if (fr->vdw_modifier == eintmodPOTSHIFT)
 371                         {
 372                             vvdw             = (vvdw_rep-cexp1*exp(-cexp2*rvdw))-(vvdw_disp + c6*sh_dispersion)/6.0;
 373                         }
 374                         else
 375                         {
 376                             vvdw             = vvdw_rep-vvdw_disp/6.0;
 377                         }
 378                         break;
 379
 380                     case GMX_NBKERNEL_VDW_CUBICSPLINETABLE:
 381                         /* Tabulated VdW */
 382                         c6               = vdwparam[tj];
 383                         c12              = vdwparam[tj+1];
 384                         Y                = VFtab[nnn+4];
 385                         F                = VFtab[nnn+5];
 386                         Geps             = eps*VFtab[nnn+6];
 387                         Heps2            = eps2*VFtab[nnn+7];
 388                         Fp               = F+Geps+Heps2;
 389                         VV               = Y+eps*Fp;
 390                         FF               = Fp+Geps+2.0*Heps2;
 391                         vvdw_disp        = c6*VV;
 392                         fijD             = c6*FF;
 393                         Y                = VFtab[nnn+8];
 394                         F                = VFtab[nnn+9];
 395                         Geps             = eps*VFtab[nnn+10];
 396                         Heps2            = eps2*VFtab[nnn+11];
 397                         Fp               = F+Geps+Heps2;
 398                         VV               = Y+eps*Fp;
 399                         FF               = Fp+Geps+2.0*Heps2;
 400                         vvdw_rep         = c12*VV;
 401                         fijR             = c12*FF;
 402                         fvdw             = -(fijD+fijR)*tabscale*rinv;
 403                         vvdw             = vvdw_disp + vvdw_rep;
 404                         break;
 405
 406
 407                     case GMX_NBKERNEL_VDW_LJEWALD:
 408                         /* LJ-PME */
 409                         rinvsix          = rinvsq*rinvsq*rinvsq;
 410                         ewcljrsq         = ewclj2*rsq;
 411                         exponent         = exp(-ewcljrsq);
 412                         poly             = exponent*(1.0 + ewcljrsq + ewcljrsq*ewcljrsq*0.5);
 413                         c6               = vdwparam[tj];
 414                         c12              = vdwparam[tj+1];
 415                         c6grid           = vdwgridparam[tj];
 416                         vvdw_disp        = (c6-c6grid*(1.0-poly))*rinvsix;
 417                         vvdw_rep         = c12*rinvsix*rinvsix;
 418                         fvdw             = (vvdw_rep - vvdw_disp - c6grid*(1.0/6.0)*exponent*ewclj6)*rinvsq;
 419                         if (fr->vdw_modifier == eintmodPOTSHIFT)
 420                         {
 421                             vvdw             = (vvdw_rep + c12*sh_repulsion)/12.0 - (vvdw_disp + c6*sh_dispersion - c6grid*sh_lj_ewald)/6.0;
 422                         }
 423                         else
 424                         {
 425                             vvdw             = vvdw_rep/12.0-vvdw_disp/6.0;
 426                         }
 427                         break;
 428
 429                     default:
 430                         gmx_fatal(FARGS, "Death & horror! No generic VdW interaction for ivdw=%d.\n", ivdw);
 431                         break;
 432                 }
 433                 if (fr->vdw_modifier == eintmodPOTSWITCH)
 434                 {
 435                     d                = rsq*rinv-fr->rvdw_switch;
 436                     d                = (d > 0.0) ? d : 0.0;
 437                     d2               = d*d;
 438                     sw               = 1.0+d2*d*(vdw_swV3+d*(vdw_swV4+d*vdw_swV5));
 439                     dsw              = d2*(vdw_swF2+d*(vdw_swF3+d*vdw_swF4));
 440                     /* See coulomb interaction for the force-switch formula */
 441                     fvdw             = fvdw*sw - rinv*vvdw*dsw;
 442                     vvdw            *= sw;
 443                 }
 444                 if (bExactVdwCutoff)
 445                 {
 446                     fvdw             = (rsq < rvdw2) ? fvdw : 0.0;
 447                     vvdw             = (rsq < rvdw2) ? vvdw : 0.0;
 448                 }
 449                 vvdwtot         += vvdw;
 450             } /* end VdW interactions */
 451
 452             fscal            = felec+fvdw;
 453
 454             tx               = fscal*dx;
 455             ty               = fscal*dy;
 456             tz               = fscal*dz;
 457             fix              = fix + tx;
 458             fiy              = fiy + ty;
 459             fiz              = fiz + tz;
 460             f[j3+0]          = f[j3+0] - tx;
 461             f[j3+1]          = f[j3+1] - ty;
 462             f[j3+2]          = f[j3+2] - tz;
 463         }
 464
 465         f[ii3+0]         = f[ii3+0] + fix;
 466         f[ii3+1]         = f[ii3+1] + fiy;
 467         f[ii3+2]         = f[ii3+2] + fiz;
 468         fshift[is3]      = fshift[is3]+fix;
 469         fshift[is3+1]    = fshift[is3+1]+fiy;
 470         fshift[is3+2]    = fshift[is3+2]+fiz;
 471         ggid             = nlist->gid[n];
 472         velecgrp[ggid]  += vctot;
 473         vvdwgrp[ggid]   += vvdwtot;
 474     }
 475     /* Estimate flops, average for generic kernel:
 476      * 12 flops per outer iteration
 477      * 50 flops per inner iteration
 478      */
 479     inc_nrnb(nrnb, eNR_NBKERNEL_GENERIC, nlist->nri*12 + nlist->jindex[n]*50);
 480 }