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