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