src/gromacs/gmxlib/nonbonded/nb_kernel_c/nb_kernel_ElecEwSw_VdwNone_GeomW3P1_c.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS c kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_c
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            None
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_c
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     int              i_shift_offset,i_coord_offset,j_coord_offset;
  67     int              j_index_start,j_index_end;
  68     int              nri,inr,ggid,iidx,jidx,jnr,outeriter,inneriter;
  69     real             shX,shY,shZ,tx,ty,tz,fscal,rcutoff,rcutoff2;
  70     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  71     real             *shiftvec,*fshift,*x,*f;
  72     int              vdwioffset0;
  73     real             ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  74     int              vdwioffset1;
  75     real             ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  76     int              vdwioffset2;
  77     real             ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  78     int              vdwjidx0;
  79     real             jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     real             dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00,cexp1_00,cexp2_00;
  81     real             dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10,cexp1_10,cexp2_10;
  82     real             dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20,cexp1_20,cexp2_20;
  83     real             velec,felec,velecsum,facel,crf,krf,krf2;
  84     real             *charge;
  85     int              ewitab;
  86     real             ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace;
  87     real             *ewtab;
  88     real             rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  89
  90     x                = xx[0];
  91     f                = ff[0];
  92
  93     nri              = nlist->nri;
  94     iinr             = nlist->iinr;
  95     jindex           = nlist->jindex;
  96     jjnr             = nlist->jjnr;
  97     shiftidx         = nlist->shift;
  98     gid              = nlist->gid;
  99     shiftvec         = fr->shift_vec[0];
 100     fshift           = fr->fshift[0];
 101     facel            = fr->epsfac;
 102     charge           = mdatoms->chargeA;
 103
 104     sh_ewald         = fr->ic->sh_ewald;
 105     ewtab            = fr->ic->tabq_coul_FDV0;
 106     ewtabscale       = fr->ic->tabq_scale;
 107     ewtabhalfspace   = 0.5/ewtabscale;
 108
 109     /* Setup water-specific parameters */
 110     inr              = nlist->iinr[0];
 111     iq0              = facel*charge[inr+0];
 112     iq1              = facel*charge[inr+1];
 113     iq2              = facel*charge[inr+2];
 114
 115     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 116     rcutoff          = fr->rcoulomb;
 117     rcutoff2         = rcutoff*rcutoff;
 118
 119     rswitch          = fr->rcoulomb_switch;
 120     /* Setup switch parameters */
 121     d                = rcutoff-rswitch;
 122     swV3             = -10.0/(d*d*d);
 123     swV4             =  15.0/(d*d*d*d);
 124     swV5             =  -6.0/(d*d*d*d*d);
 125     swF2             = -30.0/(d*d*d);
 126     swF3             =  60.0/(d*d*d*d);
 127     swF4             = -30.0/(d*d*d*d*d);
 128
 129     outeriter        = 0;
 130     inneriter        = 0;
 131
 132     /* Start outer loop over neighborlists */
 133     for(iidx=0; iidx<nri; iidx++)
 134     {
 135         /* Load shift vector for this list */
 136         i_shift_offset   = DIM*shiftidx[iidx];
 137         shX              = shiftvec[i_shift_offset+XX];
 138         shY              = shiftvec[i_shift_offset+YY];
 139         shZ              = shiftvec[i_shift_offset+ZZ];
 140
 141         /* Load limits for loop over neighbors */
 142         j_index_start    = jindex[iidx];
 143         j_index_end      = jindex[iidx+1];
 144
 145         /* Get outer coordinate index */
 146         inr              = iinr[iidx];
 147         i_coord_offset   = DIM*inr;
 148
 149         /* Load i particle coords and add shift vector */
 150         ix0              = shX + x[i_coord_offset+DIM*0+XX];
 151         iy0              = shY + x[i_coord_offset+DIM*0+YY];
 152         iz0              = shZ + x[i_coord_offset+DIM*0+ZZ];
 153         ix1              = shX + x[i_coord_offset+DIM*1+XX];
 154         iy1              = shY + x[i_coord_offset+DIM*1+YY];
 155         iz1              = shZ + x[i_coord_offset+DIM*1+ZZ];
 156         ix2              = shX + x[i_coord_offset+DIM*2+XX];
 157         iy2              = shY + x[i_coord_offset+DIM*2+YY];
 158         iz2              = shZ + x[i_coord_offset+DIM*2+ZZ];
 159
 160         fix0             = 0.0;
 161         fiy0             = 0.0;
 162         fiz0             = 0.0;
 163         fix1             = 0.0;
 164         fiy1             = 0.0;
 165         fiz1             = 0.0;
 166         fix2             = 0.0;
 167         fiy2             = 0.0;
 168         fiz2             = 0.0;
 169
 170         /* Reset potential sums */
 171         velecsum         = 0.0;
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end; jidx++)
 175         {
 176             /* Get j neighbor index, and coordinate index */
 177             jnr              = jjnr[jidx];
 178             j_coord_offset   = DIM*jnr;
 179
 180             /* load j atom coordinates */
 181             jx0              = x[j_coord_offset+DIM*0+XX];
 182             jy0              = x[j_coord_offset+DIM*0+YY];
 183             jz0              = x[j_coord_offset+DIM*0+ZZ];
 184
 185             /* Calculate displacement vector */
 186             dx00             = ix0 - jx0;
 187             dy00             = iy0 - jy0;
 188             dz00             = iz0 - jz0;
 189             dx10             = ix1 - jx0;
 190             dy10             = iy1 - jy0;
 191             dz10             = iz1 - jz0;
 192             dx20             = ix2 - jx0;
 193             dy20             = iy2 - jy0;
 194             dz20             = iz2 - jz0;
 195
 196             /* Calculate squared distance and things based on it */
 197             rsq00            = dx00*dx00+dy00*dy00+dz00*dz00;
 198             rsq10            = dx10*dx10+dy10*dy10+dz10*dz10;
 199             rsq20            = dx20*dx20+dy20*dy20+dz20*dz20;
 200
 201             rinv00           = gmx_invsqrt(rsq00);
 202             rinv10           = gmx_invsqrt(rsq10);
 203             rinv20           = gmx_invsqrt(rsq20);
 204
 205             rinvsq00         = rinv00*rinv00;
 206             rinvsq10         = rinv10*rinv10;
 207             rinvsq20         = rinv20*rinv20;
 208
 209             /* Load parameters for j particles */
 210             jq0              = charge[jnr+0];
 211
 212             /**************************
 213              * CALCULATE INTERACTIONS *
 214              **************************/
 215
 216             if (rsq00<rcutoff2)
 217             {
 218
 219             r00              = rsq00*rinv00;
 220
 221             qq00             = iq0*jq0;
 222
 223             /* EWALD ELECTROSTATICS */
 224
 225             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 226             ewrt             = r00*ewtabscale;
 227             ewitab           = ewrt;
 228             eweps            = ewrt-ewitab;
 229             ewitab           = 4*ewitab;
 230             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 231             velec            = qq00*(rinv00-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 232             felec            = qq00*rinv00*(rinvsq00-felec);
 233
 234             d                = r00-rswitch;
 235             d                = (d>0.0) ? d : 0.0;
 236             d2               = d*d;
 237             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 238
 239             dsw              = d2*(swF2+d*(swF3+d*swF4));
 240
 241             /* Evaluate switch function */
 242             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 243             felec            = felec*sw - rinv00*velec*dsw;
 244             velec           *= sw;
 245
 246             /* Update potential sums from outer loop */
 247             velecsum        += velec;
 248
 249             fscal            = felec;
 250
 251             /* Calculate temporary vectorial force */
 252             tx               = fscal*dx00;
 253             ty               = fscal*dy00;
 254             tz               = fscal*dz00;
 255
 256             /* Update vectorial force */
 257             fix0            += tx;
 258             fiy0            += ty;
 259             fiz0            += tz;
 260             f[j_coord_offset+DIM*0+XX] -= tx;
 261             f[j_coord_offset+DIM*0+YY] -= ty;
 262             f[j_coord_offset+DIM*0+ZZ] -= tz;
 263
 264             }
 265
 266             /**************************
 267              * CALCULATE INTERACTIONS *
 268              **************************/
 269
 270             if (rsq10<rcutoff2)
 271             {
 272
 273             r10              = rsq10*rinv10;
 274
 275             qq10             = iq1*jq0;
 276
 277             /* EWALD ELECTROSTATICS */
 278
 279             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 280             ewrt             = r10*ewtabscale;
 281             ewitab           = ewrt;
 282             eweps            = ewrt-ewitab;
 283             ewitab           = 4*ewitab;
 284             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 285             velec            = qq10*(rinv10-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 286             felec            = qq10*rinv10*(rinvsq10-felec);
 287
 288             d                = r10-rswitch;
 289             d                = (d>0.0) ? d : 0.0;
 290             d2               = d*d;
 291             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 292
 293             dsw              = d2*(swF2+d*(swF3+d*swF4));
 294
 295             /* Evaluate switch function */
 296             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 297             felec            = felec*sw - rinv10*velec*dsw;
 298             velec           *= sw;
 299
 300             /* Update potential sums from outer loop */
 301             velecsum        += velec;
 302
 303             fscal            = felec;
 304
 305             /* Calculate temporary vectorial force */
 306             tx               = fscal*dx10;
 307             ty               = fscal*dy10;
 308             tz               = fscal*dz10;
 309
 310             /* Update vectorial force */
 311             fix1            += tx;
 312             fiy1            += ty;
 313             fiz1            += tz;
 314             f[j_coord_offset+DIM*0+XX] -= tx;
 315             f[j_coord_offset+DIM*0+YY] -= ty;
 316             f[j_coord_offset+DIM*0+ZZ] -= tz;
 317
 318             }
 319
 320             /**************************
 321              * CALCULATE INTERACTIONS *
 322              **************************/
 323
 324             if (rsq20<rcutoff2)
 325             {
 326
 327             r20              = rsq20*rinv20;
 328
 329             qq20             = iq2*jq0;
 330
 331             /* EWALD ELECTROSTATICS */
 332
 333             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 334             ewrt             = r20*ewtabscale;
 335             ewitab           = ewrt;
 336             eweps            = ewrt-ewitab;
 337             ewitab           = 4*ewitab;
 338             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 339             velec            = qq20*(rinv20-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 340             felec            = qq20*rinv20*(rinvsq20-felec);
 341
 342             d                = r20-rswitch;
 343             d                = (d>0.0) ? d : 0.0;
 344             d2               = d*d;
 345             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 346
 347             dsw              = d2*(swF2+d*(swF3+d*swF4));
 348
 349             /* Evaluate switch function */
 350             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 351             felec            = felec*sw - rinv20*velec*dsw;
 352             velec           *= sw;
 353
 354             /* Update potential sums from outer loop */
 355             velecsum        += velec;
 356
 357             fscal            = felec;
 358
 359             /* Calculate temporary vectorial force */
 360             tx               = fscal*dx20;
 361             ty               = fscal*dy20;
 362             tz               = fscal*dz20;
 363
 364             /* Update vectorial force */
 365             fix2            += tx;
 366             fiy2            += ty;
 367             fiz2            += tz;
 368             f[j_coord_offset+DIM*0+XX] -= tx;
 369             f[j_coord_offset+DIM*0+YY] -= ty;
 370             f[j_coord_offset+DIM*0+ZZ] -= tz;
 371
 372             }
 373
 374             /* Inner loop uses 177 flops */
 375         }
 376         /* End of innermost loop */
 377
 378         tx = ty = tz = 0;
 379         f[i_coord_offset+DIM*0+XX] += fix0;
 380         f[i_coord_offset+DIM*0+YY] += fiy0;
 381         f[i_coord_offset+DIM*0+ZZ] += fiz0;
 382         tx                         += fix0;
 383         ty                         += fiy0;
 384         tz                         += fiz0;
 385         f[i_coord_offset+DIM*1+XX] += fix1;
 386         f[i_coord_offset+DIM*1+YY] += fiy1;
 387         f[i_coord_offset+DIM*1+ZZ] += fiz1;
 388         tx                         += fix1;
 389         ty                         += fiy1;
 390         tz                         += fiz1;
 391         f[i_coord_offset+DIM*2+XX] += fix2;
 392         f[i_coord_offset+DIM*2+YY] += fiy2;
 393         f[i_coord_offset+DIM*2+ZZ] += fiz2;
 394         tx                         += fix2;
 395         ty                         += fiy2;
 396         tz                         += fiz2;
 397         fshift[i_shift_offset+XX]  += tx;
 398         fshift[i_shift_offset+YY]  += ty;
 399         fshift[i_shift_offset+ZZ]  += tz;
 400
 401         ggid                        = gid[iidx];
 402         /* Update potential energies */
 403         kernel_data->energygrp_elec[ggid] += velecsum;
 404
 405         /* Increment number of inner iterations */
 406         inneriter                  += j_index_end - j_index_start;
 407
 408         /* Outer loop uses 31 flops */
 409     }
 410
 411     /* Increment number of outer iterations */
 412     outeriter        += nri;
 413
 414     /* Update outer/inner flops */
 415
 416     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*31 + inneriter*177);
 417 }
 418 /*
 419  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_c
 420  * Electrostatics interaction: Ewald
 421  * VdW interaction:            None
 422  * Geometry:                   Water3-Particle
 423  * Calculate force/pot:        Force
 424  */
 425 void
 426 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_c
 427                     (t_nblist                    * gmx_restrict       nlist,
 428                      rvec                        * gmx_restrict          xx,
 429                      rvec                        * gmx_restrict          ff,
 430                      t_forcerec                  * gmx_restrict          fr,
 431                      t_mdatoms                   * gmx_restrict     mdatoms,
 432                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 433                      t_nrnb                      * gmx_restrict        nrnb)
 434 {
 435     int              i_shift_offset,i_coord_offset,j_coord_offset;
 436     int              j_index_start,j_index_end;
 437     int              nri,inr,ggid,iidx,jidx,jnr,outeriter,inneriter;
 438     real             shX,shY,shZ,tx,ty,tz,fscal,rcutoff,rcutoff2;
 439     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 440     real             *shiftvec,*fshift,*x,*f;
 441     int              vdwioffset0;
 442     real             ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 443     int              vdwioffset1;
 444     real             ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 445     int              vdwioffset2;
 446     real             ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 447     int              vdwjidx0;
 448     real             jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 449     real             dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00,cexp1_00,cexp2_00;
 450     real             dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10,cexp1_10,cexp2_10;
 451     real             dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20,cexp1_20,cexp2_20;
 452     real             velec,felec,velecsum,facel,crf,krf,krf2;
 453     real             *charge;
 454     int              ewitab;
 455     real             ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace;
 456     real             *ewtab;
 457     real             rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 458
 459     x                = xx[0];
 460     f                = ff[0];
 461
 462     nri              = nlist->nri;
 463     iinr             = nlist->iinr;
 464     jindex           = nlist->jindex;
 465     jjnr             = nlist->jjnr;
 466     shiftidx         = nlist->shift;
 467     gid              = nlist->gid;
 468     shiftvec         = fr->shift_vec[0];
 469     fshift           = fr->fshift[0];
 470     facel            = fr->epsfac;
 471     charge           = mdatoms->chargeA;
 472
 473     sh_ewald         = fr->ic->sh_ewald;
 474     ewtab            = fr->ic->tabq_coul_FDV0;
 475     ewtabscale       = fr->ic->tabq_scale;
 476     ewtabhalfspace   = 0.5/ewtabscale;
 477
 478     /* Setup water-specific parameters */
 479     inr              = nlist->iinr[0];
 480     iq0              = facel*charge[inr+0];
 481     iq1              = facel*charge[inr+1];
 482     iq2              = facel*charge[inr+2];
 483
 484     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 485     rcutoff          = fr->rcoulomb;
 486     rcutoff2         = rcutoff*rcutoff;
 487
 488     rswitch          = fr->rcoulomb_switch;
 489     /* Setup switch parameters */
 490     d                = rcutoff-rswitch;
 491     swV3             = -10.0/(d*d*d);
 492     swV4             =  15.0/(d*d*d*d);
 493     swV5             =  -6.0/(d*d*d*d*d);
 494     swF2             = -30.0/(d*d*d);
 495     swF3             =  60.0/(d*d*d*d);
 496     swF4             = -30.0/(d*d*d*d*d);
 497
 498     outeriter        = 0;
 499     inneriter        = 0;
 500
 501     /* Start outer loop over neighborlists */
 502     for(iidx=0; iidx<nri; iidx++)
 503     {
 504         /* Load shift vector for this list */
 505         i_shift_offset   = DIM*shiftidx[iidx];
 506         shX              = shiftvec[i_shift_offset+XX];
 507         shY              = shiftvec[i_shift_offset+YY];
 508         shZ              = shiftvec[i_shift_offset+ZZ];
 509
 510         /* Load limits for loop over neighbors */
 511         j_index_start    = jindex[iidx];
 512         j_index_end      = jindex[iidx+1];
 513
 514         /* Get outer coordinate index */
 515         inr              = iinr[iidx];
 516         i_coord_offset   = DIM*inr;
 517
 518         /* Load i particle coords and add shift vector */
 519         ix0              = shX + x[i_coord_offset+DIM*0+XX];
 520         iy0              = shY + x[i_coord_offset+DIM*0+YY];
 521         iz0              = shZ + x[i_coord_offset+DIM*0+ZZ];
 522         ix1              = shX + x[i_coord_offset+DIM*1+XX];
 523         iy1              = shY + x[i_coord_offset+DIM*1+YY];
 524         iz1              = shZ + x[i_coord_offset+DIM*1+ZZ];
 525         ix2              = shX + x[i_coord_offset+DIM*2+XX];
 526         iy2              = shY + x[i_coord_offset+DIM*2+YY];
 527         iz2              = shZ + x[i_coord_offset+DIM*2+ZZ];
 528
 529         fix0             = 0.0;
 530         fiy0             = 0.0;
 531         fiz0             = 0.0;
 532         fix1             = 0.0;
 533         fiy1             = 0.0;
 534         fiz1             = 0.0;
 535         fix2             = 0.0;
 536         fiy2             = 0.0;
 537         fiz2             = 0.0;
 538
 539         /* Start inner kernel loop */
 540         for(jidx=j_index_start; jidx<j_index_end; jidx++)
 541         {
 542             /* Get j neighbor index, and coordinate index */
 543             jnr              = jjnr[jidx];
 544             j_coord_offset   = DIM*jnr;
 545
 546             /* load j atom coordinates */
 547             jx0              = x[j_coord_offset+DIM*0+XX];
 548             jy0              = x[j_coord_offset+DIM*0+YY];
 549             jz0              = x[j_coord_offset+DIM*0+ZZ];
 550
 551             /* Calculate displacement vector */
 552             dx00             = ix0 - jx0;
 553             dy00             = iy0 - jy0;
 554             dz00             = iz0 - jz0;
 555             dx10             = ix1 - jx0;
 556             dy10             = iy1 - jy0;
 557             dz10             = iz1 - jz0;
 558             dx20             = ix2 - jx0;
 559             dy20             = iy2 - jy0;
 560             dz20             = iz2 - jz0;
 561
 562             /* Calculate squared distance and things based on it */
 563             rsq00            = dx00*dx00+dy00*dy00+dz00*dz00;
 564             rsq10            = dx10*dx10+dy10*dy10+dz10*dz10;
 565             rsq20            = dx20*dx20+dy20*dy20+dz20*dz20;
 566
 567             rinv00           = gmx_invsqrt(rsq00);
 568             rinv10           = gmx_invsqrt(rsq10);
 569             rinv20           = gmx_invsqrt(rsq20);
 570
 571             rinvsq00         = rinv00*rinv00;
 572             rinvsq10         = rinv10*rinv10;
 573             rinvsq20         = rinv20*rinv20;
 574
 575             /* Load parameters for j particles */
 576             jq0              = charge[jnr+0];
 577
 578             /**************************
 579              * CALCULATE INTERACTIONS *
 580              **************************/
 581
 582             if (rsq00<rcutoff2)
 583             {
 584
 585             r00              = rsq00*rinv00;
 586
 587             qq00             = iq0*jq0;
 588
 589             /* EWALD ELECTROSTATICS */
 590
 591             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 592             ewrt             = r00*ewtabscale;
 593             ewitab           = ewrt;
 594             eweps            = ewrt-ewitab;
 595             ewitab           = 4*ewitab;
 596             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 597             velec            = qq00*(rinv00-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 598             felec            = qq00*rinv00*(rinvsq00-felec);
 599
 600             d                = r00-rswitch;
 601             d                = (d>0.0) ? d : 0.0;
 602             d2               = d*d;
 603             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 604
 605             dsw              = d2*(swF2+d*(swF3+d*swF4));
 606
 607             /* Evaluate switch function */
 608             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 609             felec            = felec*sw - rinv00*velec*dsw;
 610
 611             fscal            = felec;
 612
 613             /* Calculate temporary vectorial force */
 614             tx               = fscal*dx00;
 615             ty               = fscal*dy00;
 616             tz               = fscal*dz00;
 617
 618             /* Update vectorial force */
 619             fix0            += tx;
 620             fiy0            += ty;
 621             fiz0            += tz;
 622             f[j_coord_offset+DIM*0+XX] -= tx;
 623             f[j_coord_offset+DIM*0+YY] -= ty;
 624             f[j_coord_offset+DIM*0+ZZ] -= tz;
 625
 626             }
 627
 628             /**************************
 629              * CALCULATE INTERACTIONS *
 630              **************************/
 631
 632             if (rsq10<rcutoff2)
 633             {
 634
 635             r10              = rsq10*rinv10;
 636
 637             qq10             = iq1*jq0;
 638
 639             /* EWALD ELECTROSTATICS */
 640
 641             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 642             ewrt             = r10*ewtabscale;
 643             ewitab           = ewrt;
 644             eweps            = ewrt-ewitab;
 645             ewitab           = 4*ewitab;
 646             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 647             velec            = qq10*(rinv10-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 648             felec            = qq10*rinv10*(rinvsq10-felec);
 649
 650             d                = r10-rswitch;
 651             d                = (d>0.0) ? d : 0.0;
 652             d2               = d*d;
 653             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 654
 655             dsw              = d2*(swF2+d*(swF3+d*swF4));
 656
 657             /* Evaluate switch function */
 658             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 659             felec            = felec*sw - rinv10*velec*dsw;
 660
 661             fscal            = felec;
 662
 663             /* Calculate temporary vectorial force */
 664             tx               = fscal*dx10;
 665             ty               = fscal*dy10;
 666             tz               = fscal*dz10;
 667
 668             /* Update vectorial force */
 669             fix1            += tx;
 670             fiy1            += ty;
 671             fiz1            += tz;
 672             f[j_coord_offset+DIM*0+XX] -= tx;
 673             f[j_coord_offset+DIM*0+YY] -= ty;
 674             f[j_coord_offset+DIM*0+ZZ] -= tz;
 675
 676             }
 677
 678             /**************************
 679              * CALCULATE INTERACTIONS *
 680              **************************/
 681
 682             if (rsq20<rcutoff2)
 683             {
 684
 685             r20              = rsq20*rinv20;
 686
 687             qq20             = iq2*jq0;
 688
 689             /* EWALD ELECTROSTATICS */
 690
 691             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 692             ewrt             = r20*ewtabscale;
 693             ewitab           = ewrt;
 694             eweps            = ewrt-ewitab;
 695             ewitab           = 4*ewitab;
 696             felec            = ewtab[ewitab]+eweps*ewtab[ewitab+1];
 697             velec            = qq20*(rinv20-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
 698             felec            = qq20*rinv20*(rinvsq20-felec);
 699
 700             d                = r20-rswitch;
 701             d                = (d>0.0) ? d : 0.0;
 702             d2               = d*d;
 703             sw               = 1.0+d2*d*(swV3+d*(swV4+d*swV5));
 704
 705             dsw              = d2*(swF2+d*(swF3+d*swF4));
 706
 707             /* Evaluate switch function */
 708             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 709             felec            = felec*sw - rinv20*velec*dsw;
 710
 711             fscal            = felec;
 712
 713             /* Calculate temporary vectorial force */
 714             tx               = fscal*dx20;
 715             ty               = fscal*dy20;
 716             tz               = fscal*dz20;
 717
 718             /* Update vectorial force */
 719             fix2            += tx;
 720             fiy2            += ty;
 721             fiz2            += tz;
 722             f[j_coord_offset+DIM*0+XX] -= tx;
 723             f[j_coord_offset+DIM*0+YY] -= ty;
 724             f[j_coord_offset+DIM*0+ZZ] -= tz;
 725
 726             }
 727
 728             /* Inner loop uses 171 flops */
 729         }
 730         /* End of innermost loop */
 731
 732         tx = ty = tz = 0;
 733         f[i_coord_offset+DIM*0+XX] += fix0;
 734         f[i_coord_offset+DIM*0+YY] += fiy0;
 735         f[i_coord_offset+DIM*0+ZZ] += fiz0;
 736         tx                         += fix0;
 737         ty                         += fiy0;
 738         tz                         += fiz0;
 739         f[i_coord_offset+DIM*1+XX] += fix1;
 740         f[i_coord_offset+DIM*1+YY] += fiy1;
 741         f[i_coord_offset+DIM*1+ZZ] += fiz1;
 742         tx                         += fix1;
 743         ty                         += fiy1;
 744         tz                         += fiz1;
 745         f[i_coord_offset+DIM*2+XX] += fix2;
 746         f[i_coord_offset+DIM*2+YY] += fiy2;
 747         f[i_coord_offset+DIM*2+ZZ] += fiz2;
 748         tx                         += fix2;
 749         ty                         += fiy2;
 750         tz                         += fiz2;
 751         fshift[i_shift_offset+XX]  += tx;
 752         fshift[i_shift_offset+YY]  += ty;
 753         fshift[i_shift_offset+ZZ]  += tz;
 754
 755         /* Increment number of inner iterations */
 756         inneriter                  += j_index_end - j_index_start;
 757
 758         /* Outer loop uses 30 flops */
 759     }
 760
 761     /* Increment number of outer iterations */
 762     outeriter        += nri;
 763
 764     /* Update outer/inner flops */
 765
 766     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*30 + inneriter*171);
 767 }