src/gromacs/gmxlib/nonbonded/nb_free_energy.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
   5  * Copyright (c) 2001-2004, The GROMACS development team.
   6  * Copyright (c) 2013,2014, by the GROMACS development team, led by
   7  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   8  * and including many others, as listed in the AUTHORS file in the
   9  * top-level source directory and at http://www.gromacs.org.
  10  *
  11  * GROMACS is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public License
  13  * as published by the Free Software Foundation; either version 2.1
  14  * of the License, or (at your option) any later version.
  15  *
  16  * GROMACS is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with GROMACS; if not, see
  23  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  24  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  25  *
  26  * If you want to redistribute modifications to GROMACS, please
  27  * consider that scientific software is very special. Version
  28  * control is crucial - bugs must be traceable. We will be happy to
  29  * consider code for inclusion in the official distribution, but
  30  * derived work must not be called official GROMACS. Details are found
  31  * in the README & COPYING files - if they are missing, get the
  32  * official version at http://www.gromacs.org.
  33  *
  34  * To help us fund GROMACS development, we humbly ask that you cite
  35  * the research papers on the package. Check out http://www.gromacs.org.
  36  */
  37 #ifdef HAVE_CONFIG_H
  38 #include <config.h>
  39 #endif
  40
  41 #include <math.h>
  42
  43 #include "vec.h"
  44 #include "typedefs.h"
  45 #include "nonbonded.h"
  46 #include "nb_kernel.h"
  47 #include "nrnb.h"
  48 #include "macros.h"
  49 #include "nb_free_energy.h"
  50
  51 #include "gmx_fatal.h"
  52
  53 void
  54 gmx_nb_free_energy_kernel(const t_nblist * gmx_restrict    nlist,
  55                           rvec * gmx_restrict              xx,
  56                           rvec * gmx_restrict              ff,
  57                           t_forcerec * gmx_restrict        fr,
  58                           const t_mdatoms * gmx_restrict   mdatoms,
  59                           nb_kernel_data_t * gmx_restrict  kernel_data,
  60                           t_nrnb * gmx_restrict            nrnb)
  61 {
  62
  63 #define  STATE_A  0
  64 #define  STATE_B  1
  65 #define  NSTATES  2
  66     int           i, j, n, ii, is3, ii3, k, nj0, nj1, jnr, j3, ggid;
  67     real          shX, shY, shZ;
  68     real          tx, ty, tz, Fscal;
  69     double        FscalC[NSTATES], FscalV[NSTATES];  /* Needs double for sc_power==48 */
  70     double        Vcoul[NSTATES], Vvdw[NSTATES];     /* Needs double for sc_power==48 */
  71     real          rinv6, r, rt, rtC, rtV;
  72     real          iqA, iqB;
  73     real          qq[NSTATES], vctot, krsq;
  74     int           ntiA, ntiB, tj[NSTATES];
  75     real          Vvdw6, Vvdw12, vvtot;
  76     real          ix, iy, iz, fix, fiy, fiz;
  77     real          dx, dy, dz, rsq, rinv;
  78     real          c6[NSTATES], c12[NSTATES], c6grid;
  79     real          LFC[NSTATES], LFV[NSTATES], DLF[NSTATES];
  80     double        dvdl_coul, dvdl_vdw;
  81     real          lfac_coul[NSTATES], dlfac_coul[NSTATES], lfac_vdw[NSTATES], dlfac_vdw[NSTATES];
  82     real          sigma6[NSTATES], alpha_vdw_eff, alpha_coul_eff, sigma2_def, sigma2_min;
  83     double        rp, rpm2, rC, rV, rinvC, rpinvC, rinvV, rpinvV; /* Needs double for sc_power==48 */
  84     real          sigma2[NSTATES], sigma_pow[NSTATES], sigma_powm2[NSTATES], rs, rs2;
  85     int           do_tab, tab_elemsize;
  86     int           n0, n1C, n1V, nnn;
  87     real          Y, F, G, H, Fp, Geps, Heps2, epsC, eps2C, epsV, eps2V, VV, FF;
  88     int           icoul, ivdw;
  89     int           nri;
  90     const int *   iinr;
  91     const int *   jindex;
  92     const int *   jjnr;
  93     const int *   shift;
  94     const int *   gid;
  95     const int *   typeA;
  96     const int *   typeB;
  97     int           ntype;
  98     const real *  shiftvec;
  99     real          dvdl_part;
 100     real *        fshift;
 101     real          tabscale = 0;
 102     const real *  VFtab    = NULL;
 103     const real *  x;
 104     real *        f;
 105     real          facel, krf, crf;
 106     const real *  chargeA;
 107     const real *  chargeB;
 108     real          sigma6_min, sigma6_def, lam_power, sc_power, sc_r_power;
 109     real          alpha_coul, alpha_vdw, lambda_coul, lambda_vdw, ewc_lj;
 110     real          ewcljrsq, ewclj, ewclj2, exponent, poly, vvdw_disp, vvdw_rep, sh_lj_ewald;
 111     real          ewclj6;
 112     const real *  nbfp, *nbfp_grid;
 113     real *        dvdl;
 114     real *        Vv;
 115     real *        Vc;
 116     gmx_bool      bDoForces, bDoShiftForces, bDoPotential;
 117     real          rcoulomb, rvdw, sh_invrc6;
 118     gmx_bool      bExactElecCutoff, bExactVdwCutoff, bExactCutoffAll;
 119     gmx_bool      bEwald, bEwaldLJ;
 120     real          rcutoff_max2;
 121     const real *  tab_ewald_F_lj;
 122     const real *  tab_ewald_V_lj;
 123     real          d, d2, sw, dsw, rinvcorr;
 124     real          elec_swV3, elec_swV4, elec_swV5, elec_swF2, elec_swF3, elec_swF4;
 125     real          vdw_swV3, vdw_swV4, vdw_swV5, vdw_swF2, vdw_swF3, vdw_swF4;
 126     gmx_bool      bConvertEwaldToCoulomb, bConvertLJEwaldToLJ6;
 127     gmx_bool      bComputeVdwInteraction, bComputeElecInteraction;
 128     const real *  ewtab;
 129     int           ewitab;
 130     real          ewrt, eweps, ewtabscale, ewtabhalfspace, sh_ewald;
 131
 132     sh_ewald            = fr->ic->sh_ewald;
 133     ewtab               = fr->ic->tabq_coul_FDV0;
 134     ewtabscale          = fr->ic->tabq_scale;
 135     ewtabhalfspace      = 0.5/ewtabscale;
 136     tab_ewald_F_lj      = fr->ic->tabq_vdw_F;
 137     tab_ewald_V_lj      = fr->ic->tabq_vdw_V;
 138
 139     x                   = xx[0];
 140     f                   = ff[0];
 141
 142     fshift              = fr->fshift[0];
 143
 144     nri                 = nlist->nri;
 145     iinr                = nlist->iinr;
 146     jindex              = nlist->jindex;
 147     jjnr                = nlist->jjnr;
 148     icoul               = nlist->ielec;
 149     ivdw                = nlist->ivdw;
 150     shift               = nlist->shift;
 151     gid                 = nlist->gid;
 152
 153     shiftvec            = fr->shift_vec[0];
 154     chargeA             = mdatoms->chargeA;
 155     chargeB             = mdatoms->chargeB;
 156     facel               = fr->epsfac;
 157     krf                 = fr->k_rf;
 158     crf                 = fr->c_rf;
 159     ewc_lj              = fr->ewaldcoeff_lj;
 160     Vc                  = kernel_data->energygrp_elec;
 161     typeA               = mdatoms->typeA;
 162     typeB               = mdatoms->typeB;
 163     ntype               = fr->ntype;
 164     nbfp                = fr->nbfp;
 165     nbfp_grid           = fr->ljpme_c6grid;
 166     Vv                  = kernel_data->energygrp_vdw;
 167     lambda_coul         = kernel_data->lambda[efptCOUL];
 168     lambda_vdw          = kernel_data->lambda[efptVDW];
 169     dvdl                = kernel_data->dvdl;
 170     alpha_coul          = fr->sc_alphacoul;
 171     alpha_vdw           = fr->sc_alphavdw;
 172     lam_power           = fr->sc_power;
 173     sc_r_power          = fr->sc_r_power;
 174     sigma6_def          = fr->sc_sigma6_def;
 175     sigma6_min          = fr->sc_sigma6_min;
 176     bDoForces           = kernel_data->flags & GMX_NONBONDED_DO_FORCE;
 177     bDoShiftForces      = kernel_data->flags & GMX_NONBONDED_DO_SHIFTFORCE;
 178     bDoPotential        = kernel_data->flags & GMX_NONBONDED_DO_POTENTIAL;
 179
 180     rcoulomb            = fr->rcoulomb;
 181     sh_ewald            = fr->ic->sh_ewald;
 182     rvdw                = fr->rvdw;
 183     sh_invrc6           = fr->ic->sh_invrc6;
 184     sh_lj_ewald         = fr->ic->sh_lj_ewald;
 185     ewclj               = fr->ewaldcoeff_lj;
 186     ewclj2              = ewclj*ewclj;
 187     ewclj6              = ewclj2*ewclj2*ewclj2;
 188
 189     if (fr->coulomb_modifier == eintmodPOTSWITCH)
 190     {
 191         d               = fr->rcoulomb-fr->rcoulomb_switch;
 192         elec_swV3       = -10.0/(d*d*d);
 193         elec_swV4       =  15.0/(d*d*d*d);
 194         elec_swV5       =  -6.0/(d*d*d*d*d);
 195         elec_swF2       = -30.0/(d*d*d);
 196         elec_swF3       =  60.0/(d*d*d*d);
 197         elec_swF4       = -30.0/(d*d*d*d*d);
 198     }
 199     else
 200     {
 201         /* Avoid warnings from stupid compilers (looking at you, Clang!) */
 202         elec_swV3 = elec_swV4 = elec_swV5 = elec_swF2 = elec_swF3 = elec_swF4 = 0.0;
 203     }
 204
 205     if (fr->vdw_modifier == eintmodPOTSWITCH)
 206     {
 207         d               = fr->rvdw-fr->rvdw_switch;
 208         vdw_swV3        = -10.0/(d*d*d);
 209         vdw_swV4        =  15.0/(d*d*d*d);
 210         vdw_swV5        =  -6.0/(d*d*d*d*d);
 211         vdw_swF2        = -30.0/(d*d*d);
 212         vdw_swF3        =  60.0/(d*d*d*d);
 213         vdw_swF4        = -30.0/(d*d*d*d*d);
 214     }
 215     else
 216     {
 217         /* Avoid warnings from stupid compilers (looking at you, Clang!) */
 218         vdw_swV3 = vdw_swV4 = vdw_swV5 = vdw_swF2 = vdw_swF3 = vdw_swF4 = 0.0;
 219     }
 220
 221     if (fr->cutoff_scheme == ecutsVERLET)
 222     {
 223         const interaction_const_t *ic;
 224
 225         ic = fr->ic;
 226         if (EVDW_PME(ic->vdwtype))
 227         {
 228             ivdw         = GMX_NBKERNEL_VDW_LJEWALD;
 229         }
 230         else
 231         {
 232             ivdw         = GMX_NBKERNEL_VDW_LENNARDJONES;
 233         }
 234
 235         if (ic->eeltype == eelCUT || EEL_RF(ic->eeltype))
 236         {
 237             icoul        = GMX_NBKERNEL_ELEC_REACTIONFIELD;
 238         }
 239         else if (EEL_PME_EWALD(ic->eeltype))
 240         {
 241             icoul        = GMX_NBKERNEL_ELEC_EWALD;
 242         }
 243         else
 244         {
 245             gmx_incons("Unsupported eeltype with Verlet and free-energy");
 246         }
 247
 248         bExactElecCutoff = TRUE;
 249         bExactVdwCutoff  = TRUE;
 250     }
 251     else
 252     {
 253         bExactElecCutoff = (fr->coulomb_modifier != eintmodNONE) || fr->eeltype == eelRF_ZERO;
 254         bExactVdwCutoff  = (fr->vdw_modifier != eintmodNONE);
 255     }
 256
 257     bExactCutoffAll = (bExactElecCutoff && bExactVdwCutoff);
 258     rcutoff_max2    = max(fr->rcoulomb, fr->rvdw);
 259     rcutoff_max2    = rcutoff_max2*rcutoff_max2;
 260
 261     bEwald          = (icoul == GMX_NBKERNEL_ELEC_EWALD);
 262     bEwaldLJ        = (ivdw == GMX_NBKERNEL_VDW_LJEWALD);
 263
 264     /* For Ewald/PME interactions we cannot easily apply the soft-core component to
 265      * reciprocal space. When we use vanilla (not switch/shift) Ewald interactions, we
 266      * can apply the small trick of subtracting the _reciprocal_ space contribution
 267      * in this kernel, and instead apply the free energy interaction to the 1/r
 268      * (standard coulomb) interaction.
 269      *
 270      * However, we cannot use this approach for switch-modified since we would then
 271      * effectively end up evaluating a significantly different interaction here compared to the
 272      * normal (non-free-energy) kernels, either by applying a cutoff at a different
 273      * position than what the user requested, or by switching different
 274      * things (1/r rather than short-range Ewald). For these settings, we just
 275      * use the traditional short-range Ewald interaction in that case.
 276      */
 277     bConvertEwaldToCoulomb = (bEwald && (fr->coulomb_modifier != eintmodPOTSWITCH));
 278     /* For now the below will always be true (since LJ-PME only works with Shift in Gromacs-5.0),
 279      * but writing it this way means we stay in sync with coulomb, and it avoids future bugs.
 280      */
 281     bConvertLJEwaldToLJ6   = (bEwaldLJ && (fr->vdw_modifier   != eintmodPOTSWITCH));
 282
 283     /* We currently don't implement exclusion correction, needed with the Verlet cut-off scheme, without conversion */
 284     if (fr->cutoff_scheme == ecutsVERLET &&
 285         ((bEwald   && !bConvertEwaldToCoulomb) ||
 286          (bEwaldLJ && !bConvertLJEwaldToLJ6)))
 287     {
 288         gmx_incons("Unimplemented non-bonded setup");
 289     }
 290
 291     /* fix compiler warnings */
 292     nj1   = 0;
 293     n1C   = n1V   = 0;
 294     epsC  = epsV  = 0;
 295     eps2C = eps2V = 0;
 296
 297     dvdl_coul  = 0;
 298     dvdl_vdw   = 0;
 299
 300     /* Lambda factor for state A, 1-lambda*/
 301     LFC[STATE_A] = 1.0 - lambda_coul;
 302     LFV[STATE_A] = 1.0 - lambda_vdw;
 303
 304     /* Lambda factor for state B, lambda*/
 305     LFC[STATE_B] = lambda_coul;
 306     LFV[STATE_B] = lambda_vdw;
 307
 308     /*derivative of the lambda factor for state A and B */
 309     DLF[STATE_A] = -1;
 310     DLF[STATE_B] = 1;
 311
 312     for (i = 0; i < NSTATES; i++)
 313     {
 314         lfac_coul[i]  = (lam_power == 2 ? (1-LFC[i])*(1-LFC[i]) : (1-LFC[i]));
 315         dlfac_coul[i] = DLF[i]*lam_power/sc_r_power*(lam_power == 2 ? (1-LFC[i]) : 1);
 316         lfac_vdw[i]   = (lam_power == 2 ? (1-LFV[i])*(1-LFV[i]) : (1-LFV[i]));
 317         dlfac_vdw[i]  = DLF[i]*lam_power/sc_r_power*(lam_power == 2 ? (1-LFV[i]) : 1);
 318     }
 319     /* precalculate */
 320     sigma2_def = pow(sigma6_def, 1.0/3.0);
 321     sigma2_min = pow(sigma6_min, 1.0/3.0);
 322
 323     /* Ewald (not PME) table is special (icoul==enbcoulFEWALD) */
 324
 325     do_tab = (icoul == GMX_NBKERNEL_ELEC_CUBICSPLINETABLE ||
 326               ivdw == GMX_NBKERNEL_VDW_CUBICSPLINETABLE);
 327     if (do_tab)
 328     {
 329         tabscale         = kernel_data->table_elec_vdw->scale;
 330         VFtab            = kernel_data->table_elec_vdw->data;
 331         /* we always use the combined table here */
 332         tab_elemsize     = 12;
 333     }
 334
 335     for (n = 0; (n < nri); n++)
 336     {
 337         int npair_within_cutoff;
 338
 339         npair_within_cutoff = 0;
 340
 341         is3              = 3*shift[n];
 342         shX              = shiftvec[is3];
 343         shY              = shiftvec[is3+1];
 344         shZ              = shiftvec[is3+2];
 345         nj0              = jindex[n];
 346         nj1              = jindex[n+1];
 347         ii               = iinr[n];
 348         ii3              = 3*ii;
 349         ix               = shX + x[ii3+0];
 350         iy               = shY + x[ii3+1];
 351         iz               = shZ + x[ii3+2];
 352         iqA              = facel*chargeA[ii];
 353         iqB              = facel*chargeB[ii];
 354         ntiA             = 2*ntype*typeA[ii];
 355         ntiB             = 2*ntype*typeB[ii];
 356         vctot            = 0;
 357         vvtot            = 0;
 358         fix              = 0;
 359         fiy              = 0;
 360         fiz              = 0;
 361
 362         for (k = nj0; (k < nj1); k++)
 363         {
 364             jnr              = jjnr[k];
 365             j3               = 3*jnr;
 366             dx               = ix - x[j3];
 367             dy               = iy - x[j3+1];
 368             dz               = iz - x[j3+2];
 369             rsq              = dx*dx + dy*dy + dz*dz;
 370
 371             if (bExactCutoffAll && rsq >= rcutoff_max2)
 372             {
 373                 /* We save significant time by skipping all code below.
 374                  * Note that with soft-core interactions, the actual cut-off
 375                  * check might be different. But since the soft-core distance
 376                  * is always larger than r, checking on r here is safe.
 377                  */
 378                 continue;
 379             }
 380             npair_within_cutoff++;
 381
 382             if (rsq > 0)
 383             {
 384                 rinv         = gmx_invsqrt(rsq);
 385                 r            = rsq*rinv;
 386             }
 387             else
 388             {
 389                 /* The force at r=0 is zero, because of symmetry.
 390                  * But note that the potential is in general non-zero,
 391                  * since the soft-cored r will be non-zero.
 392                  */
 393                 rinv         = 0;
 394                 r            = 0;
 395             }
 396
 397             if (sc_r_power == 6.0)
 398             {
 399                 rpm2             = rsq*rsq;  /* r4 */
 400                 rp               = rpm2*rsq; /* r6 */
 401             }
 402             else if (sc_r_power == 48.0)
 403             {
 404                 rp               = rsq*rsq*rsq; /* r6 */
 405                 rp               = rp*rp;       /* r12 */
 406                 rp               = rp*rp;       /* r24 */
 407                 rp               = rp*rp;       /* r48 */
 408                 rpm2             = rp/rsq;      /* r46 */
 409             }
 410             else
 411             {
 412                 rp             = pow(r, sc_r_power);  /* not currently supported as input, but can handle it */
 413                 rpm2           = rp/rsq;
 414             }
 415
 416             Fscal = 0;
 417
 418             qq[STATE_A]      = iqA*chargeA[jnr];
 419             qq[STATE_B]      = iqB*chargeB[jnr];
 420
 421             tj[STATE_A]      = ntiA+2*typeA[jnr];
 422             tj[STATE_B]      = ntiB+2*typeB[jnr];
 423
 424             if (nlist->excl_fep == NULL || nlist->excl_fep[k])
 425             {
 426                 c6[STATE_A]      = nbfp[tj[STATE_A]];
 427                 c6[STATE_B]      = nbfp[tj[STATE_B]];
 428
 429                 for (i = 0; i < NSTATES; i++)
 430                 {
 431                     c12[i]             = nbfp[tj[i]+1];
 432                     if ((c6[i] > 0) && (c12[i] > 0))
 433                     {
 434                         /* c12 is stored scaled with 12.0 and c6 is scaled with 6.0 - correct for this */
 435                         sigma6[i]       = 0.5*c12[i]/c6[i];
 436                         sigma2[i]       = pow(sigma6[i], 1.0/3.0);
 437                         /* should be able to get rid of this ^^^ internal pow call eventually.  Will require agreement on
 438                            what data to store externally.  Can't be fixed without larger scale changes, so not 4.6 */
 439                         if (sigma6[i] < sigma6_min)   /* for disappearing coul and vdw with soft core at the same time */
 440                         {
 441                             sigma6[i] = sigma6_min;
 442                             sigma2[i] = sigma2_min;
 443                         }
 444                     }
 445                     else
 446                     {
 447                         sigma6[i]       = sigma6_def;
 448                         sigma2[i]       = sigma2_def;
 449                     }
 450                     if (sc_r_power == 6.0)
 451                     {
 452                         sigma_pow[i]    = sigma6[i];
 453                         sigma_powm2[i]  = sigma6[i]/sigma2[i];
 454                     }
 455                     else if (sc_r_power == 48.0)
 456                     {
 457                         sigma_pow[i]    = sigma6[i]*sigma6[i];       /* sigma^12 */
 458                         sigma_pow[i]    = sigma_pow[i]*sigma_pow[i]; /* sigma^24 */
 459                         sigma_pow[i]    = sigma_pow[i]*sigma_pow[i]; /* sigma^48 */
 460                         sigma_powm2[i]  = sigma_pow[i]/sigma2[i];
 461                     }
 462                     else
 463                     {    /* not really supported as input, but in here for testing the general case*/
 464                         sigma_pow[i]    = pow(sigma2[i], sc_r_power/2);
 465                         sigma_powm2[i]  = sigma_pow[i]/(sigma2[i]);
 466                     }
 467                 }
 468
 469                 /* only use softcore if one of the states has a zero endstate - softcore is for avoiding infinities!*/
 470                 if ((c12[STATE_A] > 0) && (c12[STATE_B] > 0))
 471                 {
 472                     alpha_vdw_eff    = 0;
 473                     alpha_coul_eff   = 0;
 474                 }
 475                 else
 476                 {
 477                     alpha_vdw_eff    = alpha_vdw;
 478                     alpha_coul_eff   = alpha_coul;
 479                 }
 480
 481                 for (i = 0; i < NSTATES; i++)
 482                 {
 483                     FscalC[i]    = 0;
 484                     FscalV[i]    = 0;
 485                     Vcoul[i]     = 0;
 486                     Vvdw[i]      = 0;
 487
 488                     /* Only spend time on A or B state if it is non-zero */
 489                     if ( (qq[i] != 0) || (c6[i] != 0) || (c12[i] != 0) )
 490                     {
 491                         /* this section has to be inside the loop because of the dependence on sigma_pow */
 492                         rpinvC         = 1.0/(alpha_coul_eff*lfac_coul[i]*sigma_pow[i]+rp);
 493                         rinvC          = pow(rpinvC, 1.0/sc_r_power);
 494                         rC             = 1.0/rinvC;
 495
 496                         rpinvV         = 1.0/(alpha_vdw_eff*lfac_vdw[i]*sigma_pow[i]+rp);
 497                         rinvV          = pow(rpinvV, 1.0/sc_r_power);
 498                         rV             = 1.0/rinvV;
 499
 500                         if (do_tab)
 501                         {
 502                             rtC        = rC*tabscale;
 503                             n0         = rtC;
 504                             epsC       = rtC-n0;
 505                             eps2C      = epsC*epsC;
 506                             n1C        = tab_elemsize*n0;
 507
 508                             rtV        = rV*tabscale;
 509                             n0         = rtV;
 510                             epsV       = rtV-n0;
 511                             eps2V      = epsV*epsV;
 512                             n1V        = tab_elemsize*n0;
 513                         }
 514
 515                         /* Only process the coulomb interactions if we have charges,
 516                          * and if we either include all entries in the list (no cutoff
 517                          * used in the kernel), or if we are within the cutoff.
 518                          */
 519                         bComputeElecInteraction = !bExactElecCutoff ||
 520                             ( bConvertEwaldToCoulomb && r < rcoulomb) ||
 521                             (!bConvertEwaldToCoulomb && rC < rcoulomb);
 522
 523                         if ( (qq[i] != 0) && bComputeElecInteraction)
 524                         {
 525                             switch (icoul)
 526                             {
 527                                 case GMX_NBKERNEL_ELEC_COULOMB:
 528                                     /* simple cutoff */
 529                                     Vcoul[i]   = qq[i]*rinvC;
 530                                     FscalC[i]  = Vcoul[i];
 531                                     /* The shift for the Coulomb potential is stored in
 532                                      * the RF parameter c_rf, which is 0 without shift.
 533                                      */
 534                                     Vcoul[i]  -= qq[i]*fr->ic->c_rf;
 535                                     break;
 536
 537                                 case GMX_NBKERNEL_ELEC_REACTIONFIELD:
 538                                     /* reaction-field */
 539                                     Vcoul[i]   = qq[i]*(rinvC + krf*rC*rC-crf);
 540                                     FscalC[i]  = qq[i]*(rinvC - 2.0*krf*rC*rC);
 541                                     break;
 542
 543                                 case GMX_NBKERNEL_ELEC_CUBICSPLINETABLE:
 544                                     /* non-Ewald tabulated coulomb */
 545                                     nnn        = n1C;
 546                                     Y          = VFtab[nnn];
 547                                     F          = VFtab[nnn+1];
 548                                     Geps       = epsC*VFtab[nnn+2];
 549                                     Heps2      = eps2C*VFtab[nnn+3];
 550                                     Fp         = F+Geps+Heps2;
 551                                     VV         = Y+epsC*Fp;
 552                                     FF         = Fp+Geps+2.0*Heps2;
 553                                     Vcoul[i]   = qq[i]*VV;
 554                                     FscalC[i]  = -qq[i]*tabscale*FF*rC;
 555                                     break;
 556
 557                                 case GMX_NBKERNEL_ELEC_GENERALIZEDBORN:
 558                                     gmx_fatal(FARGS, "Free energy and GB not implemented.\n");
 559                                     break;
 560
 561                                 case GMX_NBKERNEL_ELEC_EWALD:
 562                                     if (bConvertEwaldToCoulomb)
 563                                     {
 564                                         /* Ewald FEP is done only on the 1/r part */
 565                                         Vcoul[i]   = qq[i]*(rinvC-sh_ewald);
 566                                         FscalC[i]  = qq[i]*rinvC;
 567                                     }
 568                                     else
 569                                     {
 570                                         ewrt      = rC*ewtabscale;
 571                                         ewitab    = (int) ewrt;
 572                                         eweps     = ewrt-ewitab;
 573                                         ewitab    = 4*ewitab;
 574                                         FscalC[i] = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 575                                         rinvcorr  = rinvC-sh_ewald;
 576                                         Vcoul[i]  = qq[i]*(rinvcorr-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+FscalC[i])));
 577                                         FscalC[i] = qq[i]*(rinvC-rC*FscalC[i]);
 578                                     }
 579                                     break;
 580
 581                                 case GMX_NBKERNEL_ELEC_NONE:
 582                                     FscalC[i]  = 0.0;
 583                                     Vcoul[i]   = 0.0;
 584                                     break;
 585
 586                                 default:
 587                                     gmx_incons("Invalid icoul in free energy kernel");
 588                                     break;
 589                             }
 590
 591                             if (fr->coulomb_modifier == eintmodPOTSWITCH)
 592                             {
 593                                 d                = rC-fr->rcoulomb_switch;
 594                                 d                = (d > 0.0) ? d : 0.0;
 595                                 d2               = d*d;
 596                                 sw               = 1.0+d2*d*(elec_swV3+d*(elec_swV4+d*elec_swV5));
 597                                 dsw              = d2*(elec_swF2+d*(elec_swF3+d*elec_swF4));
 598
 599                                 FscalC[i]        = FscalC[i]*sw - rC*Vcoul[i]*dsw;
 600                                 Vcoul[i]        *= sw;
 601
 602                                 FscalC[i]        = (rC < rcoulomb) ? FscalC[i] : 0.0;
 603                                 Vcoul[i]         = (rC < rcoulomb) ? Vcoul[i] : 0.0;
 604                             }
 605                         }
 606
 607                         /* Only process the VDW interactions if we have
 608                          * some non-zero parameters, and if we either
 609                          * include all entries in the list (no cutoff used
 610                          * in the kernel), or if we are within the cutoff.
 611                          */
 612                         bComputeVdwInteraction = !bExactVdwCutoff ||
 613                             ( bConvertLJEwaldToLJ6 && r < rvdw) ||
 614                             (!bConvertLJEwaldToLJ6 && rV < rvdw);
 615                         if ((c6[i] != 0 || c12[i] != 0) && bComputeVdwInteraction)
 616                         {
 617                             switch (ivdw)
 618                             {
 619                                 case GMX_NBKERNEL_VDW_LENNARDJONES:
 620                                     /* cutoff LJ */
 621                                     if (sc_r_power == 6.0)
 622                                     {
 623                                         rinv6            = rpinvV;
 624                                     }
 625                                     else
 626                                     {
 627                                         rinv6            = rinvV*rinvV;
 628                                         rinv6            = rinv6*rinv6*rinv6;
 629                                     }
 630                                     Vvdw6            = c6[i]*rinv6;
 631                                     Vvdw12           = c12[i]*rinv6*rinv6;
 632
 633                                     Vvdw[i]          = ( (Vvdw12 - c12[i]*sh_invrc6*sh_invrc6)*(1.0/12.0)
 634                                                          - (Vvdw6 - c6[i]*sh_invrc6)*(1.0/6.0));
 635                                     FscalV[i]        = Vvdw12 - Vvdw6;
 636                                     break;
 637
 638                                 case GMX_NBKERNEL_VDW_BUCKINGHAM:
 639                                     gmx_fatal(FARGS, "Buckingham free energy not supported.");
 640                                     break;
 641
 642                                 case GMX_NBKERNEL_VDW_CUBICSPLINETABLE:
 643                                     /* Table LJ */
 644                                     nnn = n1V+4;
 645                                     /* dispersion */
 646                                     Y          = VFtab[nnn];
 647                                     F          = VFtab[nnn+1];
 648                                     Geps       = epsV*VFtab[nnn+2];
 649                                     Heps2      = eps2V*VFtab[nnn+3];
 650                                     Fp         = F+Geps+Heps2;
 651                                     VV         = Y+epsV*Fp;
 652                                     FF         = Fp+Geps+2.0*Heps2;
 653                                     Vvdw[i]   += c6[i]*VV;
 654                                     FscalV[i] -= c6[i]*tabscale*FF*rV;
 655
 656                                     /* repulsion */
 657                                     Y          = VFtab[nnn+4];
 658                                     F          = VFtab[nnn+5];
 659                                     Geps       = epsV*VFtab[nnn+6];
 660                                     Heps2      = eps2V*VFtab[nnn+7];
 661                                     Fp         = F+Geps+Heps2;
 662                                     VV         = Y+epsV*Fp;
 663                                     FF         = Fp+Geps+2.0*Heps2;
 664                                     Vvdw[i]   += c12[i]*VV;
 665                                     FscalV[i] -= c12[i]*tabscale*FF*rV;
 666                                     break;
 667
 668                                 case GMX_NBKERNEL_VDW_LJEWALD:
 669                                     if (sc_r_power == 6.0)
 670                                     {
 671                                         rinv6            = rpinvV;
 672                                     }
 673                                     else
 674                                     {
 675                                         rinv6            = rinvV*rinvV;
 676                                         rinv6            = rinv6*rinv6*rinv6;
 677                                     }
 678                                     c6grid           = nbfp_grid[tj[i]];
 679
 680                                     if (bConvertLJEwaldToLJ6)
 681                                     {
 682                                         /* cutoff LJ */
 683                                         Vvdw6            = c6[i]*rinv6;
 684                                         Vvdw12           = c12[i]*rinv6*rinv6;
 685
 686                                         Vvdw[i]          = ( (Vvdw12 - c12[i]*sh_invrc6*sh_invrc6)*(1.0/12.0)
 687                                                              - (Vvdw6 - c6[i]*sh_invrc6 - c6grid*sh_lj_ewald)*(1.0/6.0));
 688                                         FscalV[i]        = Vvdw12 - Vvdw6;
 689                                     }
 690                                     else
 691                                     {
 692                                         /* Normal LJ-PME */
 693                                         ewcljrsq         = ewclj2*rV*rV;
 694                                         exponent         = exp(-ewcljrsq);
 695                                         poly             = exponent*(1.0 + ewcljrsq + ewcljrsq*ewcljrsq*0.5);
 696                                         vvdw_disp        = (c6[i]-c6grid*(1.0-poly))*rinv6;
 697                                         vvdw_rep         = c12[i]*rinv6*rinv6;
 698                                         FscalV[i]        = vvdw_rep - vvdw_disp - c6grid*(1.0/6.0)*exponent*ewclj6;
 699                                         Vvdw[i]          = (vvdw_rep - c12[i]*sh_invrc6*sh_invrc6)/12.0 - (vvdw_disp - c6[i]*sh_invrc6 - c6grid*sh_lj_ewald)/6.0;
 700                                     }
 701                                     break;
 702
 703                                 case GMX_NBKERNEL_VDW_NONE:
 704                                     Vvdw[i]    = 0.0;
 705                                     FscalV[i]  = 0.0;
 706                                     break;
 707
 708                                 default:
 709                                     gmx_incons("Invalid ivdw in free energy kernel");
 710                                     break;
 711                             }
 712
 713                             if (fr->vdw_modifier == eintmodPOTSWITCH)
 714                             {
 715                                 d                = rV-fr->rvdw_switch;
 716                                 d                = (d > 0.0) ? d : 0.0;
 717                                 d2               = d*d;
 718                                 sw               = 1.0+d2*d*(vdw_swV3+d*(vdw_swV4+d*vdw_swV5));
 719                                 dsw              = d2*(vdw_swF2+d*(vdw_swF3+d*vdw_swF4));
 720
 721                                 FscalV[i]        = FscalV[i]*sw - rV*Vvdw[i]*dsw;
 722                                 Vvdw[i]         *= sw;
 723
 724                                 FscalV[i]  = (rV < rvdw) ? FscalV[i] : 0.0;
 725                                 Vvdw[i]    = (rV < rvdw) ? Vvdw[i] : 0.0;
 726                             }
 727                         }
 728
 729                         /* FscalC (and FscalV) now contain: dV/drC * rC
 730                          * Now we multiply by rC^-p, so it will be: dV/drC * rC^1-p
 731                          * Further down we first multiply by r^p-2 and then by
 732                          * the vector r, which in total gives: dV/drC * (r/rC)^1-p
 733                          */
 734                         FscalC[i] *= rpinvC;
 735                         FscalV[i] *= rpinvV;
 736                     }
 737                 }
 738
 739                 /* Assemble A and B states */
 740                 for (i = 0; i < NSTATES; i++)
 741                 {
 742                     vctot         += LFC[i]*Vcoul[i];
 743                     vvtot         += LFV[i]*Vvdw[i];
 744
 745                     Fscal         += LFC[i]*FscalC[i]*rpm2;
 746                     Fscal         += LFV[i]*FscalV[i]*rpm2;
 747
 748                     dvdl_coul     += Vcoul[i]*DLF[i] + LFC[i]*alpha_coul_eff*dlfac_coul[i]*FscalC[i]*sigma_pow[i];
 749                     dvdl_vdw      += Vvdw[i]*DLF[i] + LFV[i]*alpha_vdw_eff*dlfac_vdw[i]*FscalV[i]*sigma_pow[i];
 750                 }
 751             }
 752             else if (icoul == GMX_NBKERNEL_ELEC_REACTIONFIELD)
 753             {
 754                 /* For excluded pairs, which are only in this pair list when
 755                  * using the Verlet scheme, we don't use soft-core.
 756                  * The group scheme also doesn't soft-core for these.
 757                  * As there is no singularity, there is no need for soft-core.
 758                  */
 759                 VV = krf*rsq - crf;
 760                 FF = -2.0*krf;
 761
 762                 if (ii == jnr)
 763                 {
 764                     VV *= 0.5;
 765                 }
 766
 767                 for (i = 0; i < NSTATES; i++)
 768                 {
 769                     vctot      += LFC[i]*qq[i]*VV;
 770                     Fscal      += LFC[i]*qq[i]*FF;
 771                     dvdl_coul  += DLF[i]*qq[i]*VV;
 772                 }
 773             }
 774
 775             if (bConvertEwaldToCoulomb && ( !bExactElecCutoff || r < rcoulomb ) )
 776             {
 777                 /* See comment in the preamble. When using Ewald interactions
 778                  * (unless we use a switch modifier) we subtract the reciprocal-space
 779                  * Ewald component here which made it possible to apply the free
 780                  * energy interaction to 1/r (vanilla coulomb short-range part)
 781                  * above. This gets us closer to the ideal case of applying
 782                  * the softcore to the entire electrostatic interaction,
 783                  * including the reciprocal-space component.
 784                  */
 785                 real v_lr, f_lr;
 786
 787                 ewrt      = r*ewtabscale;
 788                 ewitab    = (int) ewrt;
 789                 eweps     = ewrt-ewitab;
 790                 ewitab    = 4*ewitab;
 791                 f_lr      = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 792                 v_lr      = (ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+f_lr));
 793                 f_lr     *= rinv;
 794
 795                 /* Note that any possible Ewald shift has already been applied in
 796                  * the normal interaction part above.
 797                  */
 798
 799                 if (ii == jnr)
 800                 {
 801                     /* If we get here, the i particle (ii) has itself (jnr)
 802                      * in its neighborlist. This can only happen with the Verlet
 803                      * scheme, and corresponds to a self-interaction that will
 804                      * occur twice. Scale it down by 50% to only include it once.
 805                      */
 806                     v_lr *= 0.5;
 807                 }
 808
 809                 for (i = 0; i < NSTATES; i++)
 810                 {
 811                     vctot      -= LFC[i]*qq[i]*v_lr;
 812                     Fscal      -= LFC[i]*qq[i]*f_lr;
 813                     dvdl_coul  -= (DLF[i]*qq[i])*v_lr;
 814                 }
 815             }
 816
 817             if (bConvertLJEwaldToLJ6 && (!bExactVdwCutoff || r < rvdw))
 818             {
 819                 /* See comment in the preamble. When using LJ-Ewald interactions
 820                  * (unless we use a switch modifier) we subtract the reciprocal-space
 821                  * Ewald component here which made it possible to apply the free
 822                  * energy interaction to r^-6 (vanilla LJ6 short-range part)
 823                  * above. This gets us closer to the ideal case of applying
 824                  * the softcore to the entire VdW interaction,
 825                  * including the reciprocal-space component.
 826                  */
 827                 /* We could also use the analytical form here
 828                  * iso a table, but that can cause issues for
 829                  * r close to 0 for non-interacting pairs.
 830                  */
 831                 real rs, frac, f_lr;
 832                 int  ri;
 833
 834                 rs     = rsq*rinv*ewtabscale;
 835                 ri     = (int)rs;
 836                 frac   = rs - ri;
 837                 f_lr   = (1 - frac)*tab_ewald_F_lj[ri] + frac*tab_ewald_F_lj[ri+1];
 838                 /* TODO: Currently the Ewald LJ table does not contain
 839                  * the factor 1/6, we should add this.
 840                  */
 841                 FF     = f_lr*rinv/6.0;
 842                 VV     = (tab_ewald_V_lj[ri] - ewtabhalfspace*frac*(tab_ewald_F_lj[ri] + f_lr))/6.0;
 843
 844                 if (ii == jnr)
 845                 {
 846                     /* If we get here, the i particle (ii) has itself (jnr)
 847                      * in its neighborlist. This can only happen with the Verlet
 848                      * scheme, and corresponds to a self-interaction that will
 849                      * occur twice. Scale it down by 50% to only include it once.
 850                      */
 851                     VV *= 0.5;
 852                 }
 853
 854                 for (i = 0; i < NSTATES; i++)
 855                 {
 856                     c6grid      = nbfp_grid[tj[i]];
 857                     vvtot      += LFV[i]*c6grid*VV;
 858                     Fscal      += LFV[i]*c6grid*FF;
 859                     dvdl_vdw   += (DLF[i]*c6grid)*VV;
 860                 }
 861             }
 862
 863             if (bDoForces)
 864             {
 865                 tx         = Fscal*dx;
 866                 ty         = Fscal*dy;
 867                 tz         = Fscal*dz;
 868                 fix        = fix + tx;
 869                 fiy        = fiy + ty;
 870                 fiz        = fiz + tz;
 871                 /* OpenMP atomics are expensive, but this kernels is also
 872                  * expensive, so we can take this hit, instead of using
 873                  * thread-local output buffers and extra reduction.
 874                  */
 875 #pragma omp atomic
 876                 f[j3]     -= tx;
 877 #pragma omp atomic
 878                 f[j3+1]   -= ty;
 879 #pragma omp atomic
 880                 f[j3+2]   -= tz;
 881             }
 882         }
 883
 884         /* The atomics below are expensive with many OpenMP threads.
 885          * Here unperturbed i-particles will usually only have a few
 886          * (perturbed) j-particles in the list. Thus with a buffered list
 887          * we can skip a significant number of i-reductions with a check.
 888          */
 889         if (npair_within_cutoff > 0)
 890         {
 891             if (bDoForces)
 892             {
 893 #pragma omp atomic
 894                 f[ii3]        += fix;
 895 #pragma omp atomic
 896                 f[ii3+1]      += fiy;
 897 #pragma omp atomic
 898                 f[ii3+2]      += fiz;
 899             }
 900             if (bDoShiftForces)
 901             {
 902 #pragma omp atomic
 903                 fshift[is3]   += fix;
 904 #pragma omp atomic
 905                 fshift[is3+1] += fiy;
 906 #pragma omp atomic
 907                 fshift[is3+2] += fiz;
 908             }
 909             if (bDoPotential)
 910             {
 911                 ggid               = gid[n];
 912 #pragma omp atomic
 913                 Vc[ggid]          += vctot;
 914 #pragma omp atomic
 915                 Vv[ggid]          += vvtot;
 916             }
 917         }
 918     }
 919
 920 #pragma omp atomic
 921     dvdl[efptCOUL]     += dvdl_coul;
 922  #pragma omp atomic
 923     dvdl[efptVDW]      += dvdl_vdw;
 924
 925     /* Estimate flops, average for free energy stuff:
 926      * 12  flops per outer iteration
 927      * 150 flops per inner iteration
 928      */
 929 #pragma omp atomic
 930     inc_nrnb(nrnb, eNR_NBKERNEL_FREE_ENERGY, nlist->nri*12 + nlist->jindex[n]*150);
 931 }
 932
 933 real
 934 nb_free_energy_evaluate_single(real r2, real sc_r_power, real alpha_coul, real alpha_vdw,
 935                                real tabscale, real *vftab,
 936                                real qqA, real c6A, real c12A, real qqB, real c6B, real c12B,
 937                                real LFC[2], real LFV[2], real DLF[2],
 938                                real lfac_coul[2], real lfac_vdw[2], real dlfac_coul[2], real dlfac_vdw[2],
 939                                real sigma6_def, real sigma6_min, real sigma2_def, real sigma2_min,
 940                                real *velectot, real *vvdwtot, real *dvdl)
 941 {
 942     real       r, rp, rpm2, rtab, eps, eps2, Y, F, Geps, Heps2, Fp, VV, FF, fscal;
 943     real       qq[2], c6[2], c12[2], sigma6[2], sigma2[2], sigma_pow[2], sigma_powm2[2];
 944     real       alpha_coul_eff, alpha_vdw_eff, dvdl_coul, dvdl_vdw;
 945     real       rpinv, r_coul, r_vdw, velecsum, vvdwsum;
 946     real       fscal_vdw[2], fscal_elec[2];
 947     real       velec[2], vvdw[2];
 948     int        i, ntab;
 949
 950     qq[0]    = qqA;
 951     qq[1]    = qqB;
 952     c6[0]    = c6A;
 953     c6[1]    = c6B;
 954     c12[0]   = c12A;
 955     c12[1]   = c12B;
 956
 957     if (sc_r_power == 6.0)
 958     {
 959         rpm2             = r2*r2;   /* r4 */
 960         rp               = rpm2*r2; /* r6 */
 961     }
 962     else if (sc_r_power == 48.0)
 963     {
 964         rp               = r2*r2*r2; /* r6 */
 965         rp               = rp*rp;    /* r12 */
 966         rp               = rp*rp;    /* r24 */
 967         rp               = rp*rp;    /* r48 */
 968         rpm2             = rp/r2;    /* r46 */
 969     }
 970     else
 971     {
 972         rp             = pow(r2, 0.5*sc_r_power);  /* not currently supported as input, but can handle it */
 973         rpm2           = rp/r2;
 974     }
 975
 976     /* Loop over state A(0) and B(1) */
 977     for (i = 0; i < 2; i++)
 978     {
 979         if ((c6[i] > 0) && (c12[i] > 0))
 980         {
 981             /* The c6 & c12 coefficients now contain the constants 6.0 and 12.0, respectively.
 982              * Correct for this by multiplying with (1/12.0)/(1/6.0)=6.0/12.0=0.5.
 983              */
 984             sigma6[i]       = 0.5*c12[i]/c6[i];
 985             sigma2[i]       = pow(0.5*c12[i]/c6[i], 1.0/3.0);
 986             /* should be able to get rid of this ^^^ internal pow call eventually.  Will require agreement on
 987                what data to store externally.  Can't be fixed without larger scale changes, so not 5.0 */
 988             if (sigma6[i] < sigma6_min)   /* for disappearing coul and vdw with soft core at the same time */
 989             {
 990                 sigma6[i] = sigma6_min;
 991                 sigma2[i] = sigma2_min;
 992             }
 993         }
 994         else
 995         {
 996             sigma6[i]       = sigma6_def;
 997             sigma2[i]       = sigma2_def;
 998         }
 999         if (sc_r_power == 6.0)
1000         {
1001             sigma_pow[i]    = sigma6[i];
1002             sigma_powm2[i]  = sigma6[i]/sigma2[i];
1003         }
1004         else if (sc_r_power == 48.0)
1005         {
1006             sigma_pow[i]    = sigma6[i]*sigma6[i];       /* sigma^12 */
1007             sigma_pow[i]    = sigma_pow[i]*sigma_pow[i]; /* sigma^24 */
1008             sigma_pow[i]    = sigma_pow[i]*sigma_pow[i]; /* sigma^48 */
1009             sigma_powm2[i]  = sigma_pow[i]/sigma2[i];
1010         }
1011         else
1012         {    /* not really supported as input, but in here for testing the general case*/
1013             sigma_pow[i]    = pow(sigma2[i], sc_r_power/2);
1014             sigma_powm2[i]  = sigma_pow[i]/(sigma2[i]);
1015         }
1016     }
1017
1018     /* only use softcore if one of the states has a zero endstate - softcore is for avoiding infinities!*/
1019     if ((c12[0] > 0) && (c12[1] > 0))
1020     {
1021         alpha_vdw_eff    = 0;
1022         alpha_coul_eff   = 0;
1023     }
1024     else
1025     {
1026         alpha_vdw_eff    = alpha_vdw;
1027         alpha_coul_eff   = alpha_coul;
1028     }
1029
1030     /* Loop over A and B states again */
1031     for (i = 0; i < 2; i++)
1032     {
1033         fscal_elec[i] = 0;
1034         fscal_vdw[i]  = 0;
1035         velec[i]      = 0;
1036         vvdw[i]       = 0;
1037
1038         /* Only spend time on A or B state if it is non-zero */
1039         if ( (qq[i] != 0) || (c6[i] != 0) || (c12[i] != 0) )
1040         {
1041             /* Coulomb */
1042             rpinv            = 1.0/(alpha_coul_eff*lfac_coul[i]*sigma_pow[i]+rp);
1043             r_coul           = pow(rpinv, -1.0/sc_r_power);
1044
1045             /* Electrostatics table lookup data */
1046             rtab             = r_coul*tabscale;
1047             ntab             = rtab;
1048             eps              = rtab-ntab;
1049             eps2             = eps*eps;
1050             ntab             = 12*ntab;
1051             /* Electrostatics */
1052             Y                = vftab[ntab];
1053             F                = vftab[ntab+1];
1054             Geps             = eps*vftab[ntab+2];
1055             Heps2            = eps2*vftab[ntab+3];
1056             Fp               = F+Geps+Heps2;
1057             VV               = Y+eps*Fp;
1058             FF               = Fp+Geps+2.0*Heps2;
1059             velec[i]         = qq[i]*VV;
1060             fscal_elec[i]    = -qq[i]*FF*r_coul*rpinv*tabscale;
1061
1062             /* Vdw */
1063             rpinv            = 1.0/(alpha_vdw_eff*lfac_vdw[i]*sigma_pow[i]+rp);
1064             r_vdw            = pow(rpinv, -1.0/sc_r_power);
1065             /* Vdw table lookup data */
1066             rtab             = r_vdw*tabscale;
1067             ntab             = rtab;
1068             eps              = rtab-ntab;
1069             eps2             = eps*eps;
1070             ntab             = 12*ntab;
1071             /* Dispersion */
1072             Y                = vftab[ntab+4];
1073             F                = vftab[ntab+5];
1074             Geps             = eps*vftab[ntab+6];
1075             Heps2            = eps2*vftab[ntab+7];
1076             Fp               = F+Geps+Heps2;
1077             VV               = Y+eps*Fp;
1078             FF               = Fp+Geps+2.0*Heps2;
1079             vvdw[i]          = c6[i]*VV;
1080             fscal_vdw[i]     = -c6[i]*FF;
1081
1082             /* Repulsion */
1083             Y                = vftab[ntab+8];
1084             F                = vftab[ntab+9];
1085             Geps             = eps*vftab[ntab+10];
1086             Heps2            = eps2*vftab[ntab+11];
1087             Fp               = F+Geps+Heps2;
1088             VV               = Y+eps*Fp;
1089             FF               = Fp+Geps+2.0*Heps2;
1090             vvdw[i]         += c12[i]*VV;
1091             fscal_vdw[i]    -= c12[i]*FF;
1092             fscal_vdw[i]    *= r_vdw*rpinv*tabscale;
1093         }
1094     }
1095     /* Now we have velec[i], vvdw[i], and fscal[i] for both states */
1096     /* Assemble A and B states */
1097     velecsum  = 0;
1098     vvdwsum   = 0;
1099     dvdl_coul = 0;
1100     dvdl_vdw  = 0;
1101     fscal     = 0;
1102     for (i = 0; i < 2; i++)
1103     {
1104         velecsum      += LFC[i]*velec[i];
1105         vvdwsum       += LFV[i]*vvdw[i];
1106
1107         fscal         += (LFC[i]*fscal_elec[i]+LFV[i]*fscal_vdw[i])*rpm2;
1108
1109         dvdl_coul     += velec[i]*DLF[i] + LFC[i]*alpha_coul_eff*dlfac_coul[i]*fscal_elec[i]*sigma_pow[i];
1110         dvdl_vdw      += vvdw[i]*DLF[i] + LFV[i]*alpha_vdw_eff*dlfac_vdw[i]*fscal_vdw[i]*sigma_pow[i];
1111     }
1112
1113     dvdl[efptCOUL]     += dvdl_coul;
1114     dvdl[efptVDW]      += dvdl_vdw;
1115
1116     *velectot           = velecsum;
1117     *vvdwtot            = vvdwsum;
1118
1119     return fscal;
1120 }