src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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 sparc64_hpc_ace_double 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/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct 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     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_elec_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* Avoid stupid compiler warnings */
 137     jnrA = jnrB = 0;
 138     j_coord_offsetA = 0;
 139     j_coord_offsetB = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     /* Start outer loop over neighborlists */
 145     for(iidx=0; iidx<nri; iidx++)
 146     {
 147         /* Load shift vector for this list */
 148         i_shift_offset   = DIM*shiftidx[iidx];
 149
 150         /* Load limits for loop over neighbors */
 151         j_index_start    = jindex[iidx];
 152         j_index_end      = jindex[iidx+1];
 153
 154         /* Get outer coordinate index */
 155         inr              = iinr[iidx];
 156         i_coord_offset   = DIM*inr;
 157
 158         /* Load i particle coords and add shift vector */
 159         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 160                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165         fix1             = _fjsp_setzero_v2r8();
 166         fiy1             = _fjsp_setzero_v2r8();
 167         fiz1             = _fjsp_setzero_v2r8();
 168         fix2             = _fjsp_setzero_v2r8();
 169         fiy2             = _fjsp_setzero_v2r8();
 170         fiz2             = _fjsp_setzero_v2r8();
 171         fix3             = _fjsp_setzero_v2r8();
 172         fiy3             = _fjsp_setzero_v2r8();
 173         fiz3             = _fjsp_setzero_v2r8();
 174
 175         /* Reset potential sums */
 176         velecsum         = _fjsp_setzero_v2r8();
 177         vvdwsum          = _fjsp_setzero_v2r8();
 178
 179         /* Start inner kernel loop */
 180         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 181         {
 182
 183             /* Get j neighbor index, and coordinate index */
 184             jnrA             = jjnr[jidx];
 185             jnrB             = jjnr[jidx+1];
 186             j_coord_offsetA  = DIM*jnrA;
 187             j_coord_offsetB  = DIM*jnrB;
 188
 189             /* load j atom coordinates */
 190             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 191                                               &jx0,&jy0,&jz0);
 192
 193             /* Calculate displacement vector */
 194             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 195             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 196             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 197             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 198             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 199             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 200             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 201             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 202             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 203             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 204             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 205             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 217
 218             /* Load parameters for j particles */
 219             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 220             vdwjidx0A        = 2*vdwtype[jnrA+0];
 221             vdwjidx0B        = 2*vdwtype[jnrB+0];
 222
 223             fjx0             = _fjsp_setzero_v2r8();
 224             fjy0             = _fjsp_setzero_v2r8();
 225             fjz0             = _fjsp_setzero_v2r8();
 226
 227             /**************************
 228              * CALCULATE INTERACTIONS *
 229              **************************/
 230
 231             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 232
 233             /* Compute parameters for interactions between i and j atoms */
 234             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 235                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 236
 237             /* Calculate table index by multiplying r with table scale and truncate to integer */
 238             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 239             itab_tmp         = _fjsp_dtox_v2r8(rt);
 240             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 241             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 242             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 243
 244             vfconv.i[0]     *= 12;
 245             vfconv.i[1]     *= 12;
 246
 247             /* CUBIC SPLINE TABLE DISPERSION */
 248             vfconv.i[0]       += 4;
 249             vfconv.i[1]       += 4;
 250             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 251             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 252             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 253             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 254             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 255             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 256             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 257             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 258             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 259             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 260             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 261
 262             /* CUBIC SPLINE TABLE REPULSION */
 263             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 264             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 265             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 266             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 267             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 268             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 269             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 270             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 271             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 272             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 273             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 274             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 275             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 276
 277             /* Update potential sum for this i atom from the interaction with this j atom. */
 278             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 279
 280             fscal            = fvdw;
 281
 282             /* Update vectorial force */
 283             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 284             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 285             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 286
 287             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 288             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 289             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 290
 291             /**************************
 292              * CALCULATE INTERACTIONS *
 293              **************************/
 294
 295             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 296
 297             /* Compute parameters for interactions between i and j atoms */
 298             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 299
 300             /* Calculate table index by multiplying r with table scale and truncate to integer */
 301             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 302             itab_tmp         = _fjsp_dtox_v2r8(rt);
 303             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 304             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 305             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 306
 307             vfconv.i[0]     *= 12;
 308             vfconv.i[1]     *= 12;
 309
 310             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 311             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 312             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 313             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 314             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 315             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 316             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 317             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 318             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 319             velec            = _fjsp_mul_v2r8(qq10,VV);
 320             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 321             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 322
 323             /* Update potential sum for this i atom from the interaction with this j atom. */
 324             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 325
 326             fscal            = felec;
 327
 328             /* Update vectorial force */
 329             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 330             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 331             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 332
 333             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 334             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 335             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 336
 337             /**************************
 338              * CALCULATE INTERACTIONS *
 339              **************************/
 340
 341             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 342
 343             /* Compute parameters for interactions between i and j atoms */
 344             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 345
 346             /* Calculate table index by multiplying r with table scale and truncate to integer */
 347             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 348             itab_tmp         = _fjsp_dtox_v2r8(rt);
 349             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 350             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 351             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 352
 353             vfconv.i[0]     *= 12;
 354             vfconv.i[1]     *= 12;
 355
 356             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 357             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 358             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 359             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 360             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 361             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 362             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 363             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 364             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 365             velec            = _fjsp_mul_v2r8(qq20,VV);
 366             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 367             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 368
 369             /* Update potential sum for this i atom from the interaction with this j atom. */
 370             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 371
 372             fscal            = felec;
 373
 374             /* Update vectorial force */
 375             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 376             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 377             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 378
 379             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 380             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 381             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 382
 383             /**************************
 384              * CALCULATE INTERACTIONS *
 385              **************************/
 386
 387             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 388
 389             /* Compute parameters for interactions between i and j atoms */
 390             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 391
 392             /* Calculate table index by multiplying r with table scale and truncate to integer */
 393             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 394             itab_tmp         = _fjsp_dtox_v2r8(rt);
 395             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 396             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 397             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 398
 399             vfconv.i[0]     *= 12;
 400             vfconv.i[1]     *= 12;
 401
 402             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 403             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 404             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 405             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 406             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 407             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 408             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 409             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 410             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 411             velec            = _fjsp_mul_v2r8(qq30,VV);
 412             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 413             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 414
 415             /* Update potential sum for this i atom from the interaction with this j atom. */
 416             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 417
 418             fscal            = felec;
 419
 420             /* Update vectorial force */
 421             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 422             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 423             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 424
 425             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 426             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 427             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 428
 429             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 430
 431             /* Inner loop uses 200 flops */
 432         }
 433
 434         if(jidx<j_index_end)
 435         {
 436
 437             jnrA             = jjnr[jidx];
 438             j_coord_offsetA  = DIM*jnrA;
 439
 440             /* load j atom coordinates */
 441             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 442                                               &jx0,&jy0,&jz0);
 443
 444             /* Calculate displacement vector */
 445             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 446             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 447             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 448             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 449             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 450             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 451             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 452             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 453             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 454             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 455             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 456             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 457
 458             /* Calculate squared distance and things based on it */
 459             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 460             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 461             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 462             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 463
 464             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 465             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 466             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 467             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 468
 469             /* Load parameters for j particles */
 470             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 471             vdwjidx0A        = 2*vdwtype[jnrA+0];
 472
 473             fjx0             = _fjsp_setzero_v2r8();
 474             fjy0             = _fjsp_setzero_v2r8();
 475             fjz0             = _fjsp_setzero_v2r8();
 476
 477             /**************************
 478              * CALCULATE INTERACTIONS *
 479              **************************/
 480
 481             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 482
 483             /* Compute parameters for interactions between i and j atoms */
 484             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 485                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 486
 487             /* Calculate table index by multiplying r with table scale and truncate to integer */
 488             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 489             itab_tmp         = _fjsp_dtox_v2r8(rt);
 490             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 491             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 492             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 493
 494             vfconv.i[0]     *= 12;
 495             vfconv.i[1]     *= 12;
 496
 497             /* CUBIC SPLINE TABLE DISPERSION */
 498             vfconv.i[0]       += 4;
 499             vfconv.i[1]       += 4;
 500             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 501             F                = _fjsp_setzero_v2r8();
 502             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 503             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 504             H                = _fjsp_setzero_v2r8();
 505             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 506             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 507             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 508             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 509             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 510             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 511
 512             /* CUBIC SPLINE TABLE REPULSION */
 513             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 514             F                = _fjsp_setzero_v2r8();
 515             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 516             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 517             H                = _fjsp_setzero_v2r8();
 518             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 519             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 520             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 521             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 522             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 523             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 524             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 525             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 529             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 530
 531             fscal            = fvdw;
 532
 533             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 534
 535             /* Update vectorial force */
 536             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 537             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 538             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 539
 540             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 541             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 542             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 549
 550             /* Compute parameters for interactions between i and j atoms */
 551             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 552
 553             /* Calculate table index by multiplying r with table scale and truncate to integer */
 554             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 555             itab_tmp         = _fjsp_dtox_v2r8(rt);
 556             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 557             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 558             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 559
 560             vfconv.i[0]     *= 12;
 561             vfconv.i[1]     *= 12;
 562
 563             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 564             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 565             F                = _fjsp_setzero_v2r8();
 566             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 567             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 568             H                = _fjsp_setzero_v2r8();
 569             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 570             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 571             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 572             velec            = _fjsp_mul_v2r8(qq10,VV);
 573             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 574             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 575
 576             /* Update potential sum for this i atom from the interaction with this j atom. */
 577             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 578             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 579
 580             fscal            = felec;
 581
 582             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 583
 584             /* Update vectorial force */
 585             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 586             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 587             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 588
 589             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 590             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 591             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 592
 593             /**************************
 594              * CALCULATE INTERACTIONS *
 595              **************************/
 596
 597             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 598
 599             /* Compute parameters for interactions between i and j atoms */
 600             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 601
 602             /* Calculate table index by multiplying r with table scale and truncate to integer */
 603             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 604             itab_tmp         = _fjsp_dtox_v2r8(rt);
 605             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 606             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 607             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 608
 609             vfconv.i[0]     *= 12;
 610             vfconv.i[1]     *= 12;
 611
 612             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 613             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 614             F                = _fjsp_setzero_v2r8();
 615             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 616             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 617             H                = _fjsp_setzero_v2r8();
 618             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 619             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 620             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 621             velec            = _fjsp_mul_v2r8(qq20,VV);
 622             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 623             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 624
 625             /* Update potential sum for this i atom from the interaction with this j atom. */
 626             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 627             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 628
 629             fscal            = felec;
 630
 631             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 632
 633             /* Update vectorial force */
 634             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 635             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 636             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 637
 638             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 639             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 640             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 641
 642             /**************************
 643              * CALCULATE INTERACTIONS *
 644              **************************/
 645
 646             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 647
 648             /* Compute parameters for interactions between i and j atoms */
 649             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 650
 651             /* Calculate table index by multiplying r with table scale and truncate to integer */
 652             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 653             itab_tmp         = _fjsp_dtox_v2r8(rt);
 654             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 655             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 656             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 657
 658             vfconv.i[0]     *= 12;
 659             vfconv.i[1]     *= 12;
 660
 661             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 662             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 663             F                = _fjsp_setzero_v2r8();
 664             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 665             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 666             H                = _fjsp_setzero_v2r8();
 667             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 668             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 669             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 670             velec            = _fjsp_mul_v2r8(qq30,VV);
 671             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 672             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 673
 674             /* Update potential sum for this i atom from the interaction with this j atom. */
 675             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 676             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 677
 678             fscal            = felec;
 679
 680             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 681
 682             /* Update vectorial force */
 683             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 684             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 685             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 686
 687             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 688             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 689             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 690
 691             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 692
 693             /* Inner loop uses 200 flops */
 694         }
 695
 696         /* End of innermost loop */
 697
 698         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 699                                               f+i_coord_offset,fshift+i_shift_offset);
 700
 701         ggid                        = gid[iidx];
 702         /* Update potential energies */
 703         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 704         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 705
 706         /* Increment number of inner iterations */
 707         inneriter                  += j_index_end - j_index_start;
 708
 709         /* Outer loop uses 26 flops */
 710     }
 711
 712     /* Increment number of outer iterations */
 713     outeriter        += nri;
 714
 715     /* Update outer/inner flops */
 716
 717     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*200);
 718 }
 719 /*
 720  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 721  * Electrostatics interaction: CubicSplineTable
 722  * VdW interaction:            CubicSplineTable
 723  * Geometry:                   Water4-Particle
 724  * Calculate force/pot:        Force
 725  */
 726 void
 727 nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 728                     (t_nblist                    * gmx_restrict       nlist,
 729                      rvec                        * gmx_restrict          xx,
 730                      rvec                        * gmx_restrict          ff,
 731                      struct t_forcerec           * gmx_restrict          fr,
 732                      t_mdatoms                   * gmx_restrict     mdatoms,
 733                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 734                      t_nrnb                      * gmx_restrict        nrnb)
 735 {
 736     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 737      * just 0 for non-waters.
 738      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 739      * jnr indices corresponding to data put in the four positions in the SIMD register.
 740      */
 741     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 742     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 743     int              jnrA,jnrB;
 744     int              j_coord_offsetA,j_coord_offsetB;
 745     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 746     real             rcutoff_scalar;
 747     real             *shiftvec,*fshift,*x,*f;
 748     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 749     int              vdwioffset0;
 750     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 751     int              vdwioffset1;
 752     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 753     int              vdwioffset2;
 754     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 755     int              vdwioffset3;
 756     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 757     int              vdwjidx0A,vdwjidx0B;
 758     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 759     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 760     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 761     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 762     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 763     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 764     real             *charge;
 765     int              nvdwtype;
 766     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 767     int              *vdwtype;
 768     real             *vdwparam;
 769     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 770     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 771     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 772     real             *vftab;
 773     _fjsp_v2r8       itab_tmp;
 774     _fjsp_v2r8       dummy_mask,cutoff_mask;
 775     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 776     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 777     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 778
 779     x                = xx[0];
 780     f                = ff[0];
 781
 782     nri              = nlist->nri;
 783     iinr             = nlist->iinr;
 784     jindex           = nlist->jindex;
 785     jjnr             = nlist->jjnr;
 786     shiftidx         = nlist->shift;
 787     gid              = nlist->gid;
 788     shiftvec         = fr->shift_vec[0];
 789     fshift           = fr->fshift[0];
 790     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 791     charge           = mdatoms->chargeA;
 792     nvdwtype         = fr->ntype;
 793     vdwparam         = fr->nbfp;
 794     vdwtype          = mdatoms->typeA;
 795
 796     vftab            = kernel_data->table_elec_vdw->data;
 797     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 798
 799     /* Setup water-specific parameters */
 800     inr              = nlist->iinr[0];
 801     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 802     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 803     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 804     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 805
 806     /* Avoid stupid compiler warnings */
 807     jnrA = jnrB = 0;
 808     j_coord_offsetA = 0;
 809     j_coord_offsetB = 0;
 810
 811     outeriter        = 0;
 812     inneriter        = 0;
 813
 814     /* Start outer loop over neighborlists */
 815     for(iidx=0; iidx<nri; iidx++)
 816     {
 817         /* Load shift vector for this list */
 818         i_shift_offset   = DIM*shiftidx[iidx];
 819
 820         /* Load limits for loop over neighbors */
 821         j_index_start    = jindex[iidx];
 822         j_index_end      = jindex[iidx+1];
 823
 824         /* Get outer coordinate index */
 825         inr              = iinr[iidx];
 826         i_coord_offset   = DIM*inr;
 827
 828         /* Load i particle coords and add shift vector */
 829         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 830                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 831
 832         fix0             = _fjsp_setzero_v2r8();
 833         fiy0             = _fjsp_setzero_v2r8();
 834         fiz0             = _fjsp_setzero_v2r8();
 835         fix1             = _fjsp_setzero_v2r8();
 836         fiy1             = _fjsp_setzero_v2r8();
 837         fiz1             = _fjsp_setzero_v2r8();
 838         fix2             = _fjsp_setzero_v2r8();
 839         fiy2             = _fjsp_setzero_v2r8();
 840         fiz2             = _fjsp_setzero_v2r8();
 841         fix3             = _fjsp_setzero_v2r8();
 842         fiy3             = _fjsp_setzero_v2r8();
 843         fiz3             = _fjsp_setzero_v2r8();
 844
 845         /* Start inner kernel loop */
 846         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 847         {
 848
 849             /* Get j neighbor index, and coordinate index */
 850             jnrA             = jjnr[jidx];
 851             jnrB             = jjnr[jidx+1];
 852             j_coord_offsetA  = DIM*jnrA;
 853             j_coord_offsetB  = DIM*jnrB;
 854
 855             /* load j atom coordinates */
 856             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 857                                               &jx0,&jy0,&jz0);
 858
 859             /* Calculate displacement vector */
 860             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 861             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 862             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 863             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 864             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 865             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 866             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 867             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 868             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 869             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 870             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 871             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 872
 873             /* Calculate squared distance and things based on it */
 874             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 875             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 876             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 877             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 878
 879             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 880             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 881             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 882             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 883
 884             /* Load parameters for j particles */
 885             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 886             vdwjidx0A        = 2*vdwtype[jnrA+0];
 887             vdwjidx0B        = 2*vdwtype[jnrB+0];
 888
 889             fjx0             = _fjsp_setzero_v2r8();
 890             fjy0             = _fjsp_setzero_v2r8();
 891             fjz0             = _fjsp_setzero_v2r8();
 892
 893             /**************************
 894              * CALCULATE INTERACTIONS *
 895              **************************/
 896
 897             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 898
 899             /* Compute parameters for interactions between i and j atoms */
 900             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 901                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 902
 903             /* Calculate table index by multiplying r with table scale and truncate to integer */
 904             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 905             itab_tmp         = _fjsp_dtox_v2r8(rt);
 906             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 907             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 908             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 909
 910             vfconv.i[0]     *= 12;
 911             vfconv.i[1]     *= 12;
 912
 913             /* CUBIC SPLINE TABLE DISPERSION */
 914             vfconv.i[0]       += 4;
 915             vfconv.i[1]       += 4;
 916             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 917             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 918             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 919             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 920             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 921             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 922             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 923             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 924             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 925
 926             /* CUBIC SPLINE TABLE REPULSION */
 927             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 928             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 929             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 930             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 931             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 932             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 933             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 934             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 935             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 936             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 937
 938             fscal            = fvdw;
 939
 940             /* Update vectorial force */
 941             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 942             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 943             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 944
 945             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 946             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 947             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 948
 949             /**************************
 950              * CALCULATE INTERACTIONS *
 951              **************************/
 952
 953             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 954
 955             /* Compute parameters for interactions between i and j atoms */
 956             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 957
 958             /* Calculate table index by multiplying r with table scale and truncate to integer */
 959             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 960             itab_tmp         = _fjsp_dtox_v2r8(rt);
 961             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 962             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 963             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 964
 965             vfconv.i[0]     *= 12;
 966             vfconv.i[1]     *= 12;
 967
 968             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 969             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 970             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 971             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 972             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 973             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 974             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 975             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 976             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 977             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 978
 979             fscal            = felec;
 980
 981             /* Update vectorial force */
 982             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 983             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 984             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 985
 986             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 987             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 988             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 989
 990             /**************************
 991              * CALCULATE INTERACTIONS *
 992              **************************/
 993
 994             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 995
 996             /* Compute parameters for interactions between i and j atoms */
 997             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 998
 999             /* Calculate table index by multiplying r with table scale and truncate to integer */
1000             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1001             itab_tmp         = _fjsp_dtox_v2r8(rt);
1002             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1003             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1004             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1005
1006             vfconv.i[0]     *= 12;
1007             vfconv.i[1]     *= 12;
1008
1009             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1010             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1011             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
1012             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1013             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1014             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
1015             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1016             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1017             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1018             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1019
1020             fscal            = felec;
1021
1022             /* Update vectorial force */
1023             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1024             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1025             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1026
1027             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1028             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1029             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1030
1031             /**************************
1032              * CALCULATE INTERACTIONS *
1033              **************************/
1034
1035             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1036
1037             /* Compute parameters for interactions between i and j atoms */
1038             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1039
1040             /* Calculate table index by multiplying r with table scale and truncate to integer */
1041             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1042             itab_tmp         = _fjsp_dtox_v2r8(rt);
1043             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1044             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1045             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1046
1047             vfconv.i[0]     *= 12;
1048             vfconv.i[1]     *= 12;
1049
1050             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1051             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1052             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
1053             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1054             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1055             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
1056             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1057             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1058             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1059             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1060
1061             fscal            = felec;
1062
1063             /* Update vectorial force */
1064             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1065             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1066             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1067
1068             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1069             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1070             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1071
1072             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1073
1074             /* Inner loop uses 180 flops */
1075         }
1076
1077         if(jidx<j_index_end)
1078         {
1079
1080             jnrA             = jjnr[jidx];
1081             j_coord_offsetA  = DIM*jnrA;
1082
1083             /* load j atom coordinates */
1084             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1085                                               &jx0,&jy0,&jz0);
1086
1087             /* Calculate displacement vector */
1088             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1089             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1090             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1091             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1092             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1093             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1094             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1095             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1096             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1097             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1098             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1099             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1100
1101             /* Calculate squared distance and things based on it */
1102             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1103             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1104             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1105             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1106
1107             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1108             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1109             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1110             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1111
1112             /* Load parameters for j particles */
1113             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1114             vdwjidx0A        = 2*vdwtype[jnrA+0];
1115
1116             fjx0             = _fjsp_setzero_v2r8();
1117             fjy0             = _fjsp_setzero_v2r8();
1118             fjz0             = _fjsp_setzero_v2r8();
1119
1120             /**************************
1121              * CALCULATE INTERACTIONS *
1122              **************************/
1123
1124             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1125
1126             /* Compute parameters for interactions between i and j atoms */
1127             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1128                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1129
1130             /* Calculate table index by multiplying r with table scale and truncate to integer */
1131             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1132             itab_tmp         = _fjsp_dtox_v2r8(rt);
1133             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1134             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1135             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1136
1137             vfconv.i[0]     *= 12;
1138             vfconv.i[1]     *= 12;
1139
1140             /* CUBIC SPLINE TABLE DISPERSION */
1141             vfconv.i[0]       += 4;
1142             vfconv.i[1]       += 4;
1143             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1144             F                = _fjsp_setzero_v2r8();
1145             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1146             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1147             H                = _fjsp_setzero_v2r8();
1148             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1149             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1150             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1151             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1152
1153             /* CUBIC SPLINE TABLE REPULSION */
1154             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1155             F                = _fjsp_setzero_v2r8();
1156             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1157             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1158             H                = _fjsp_setzero_v2r8();
1159             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1160             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1161             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1162             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1163             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1164
1165             fscal            = fvdw;
1166
1167             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1168
1169             /* Update vectorial force */
1170             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1171             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1172             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1173
1174             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1175             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1176             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1177
1178             /**************************
1179              * CALCULATE INTERACTIONS *
1180              **************************/
1181
1182             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1183
1184             /* Compute parameters for interactions between i and j atoms */
1185             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1186
1187             /* Calculate table index by multiplying r with table scale and truncate to integer */
1188             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1189             itab_tmp         = _fjsp_dtox_v2r8(rt);
1190             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1191             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1192             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1193
1194             vfconv.i[0]     *= 12;
1195             vfconv.i[1]     *= 12;
1196
1197             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1198             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1199             F                = _fjsp_setzero_v2r8();
1200             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1201             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1202             H                = _fjsp_setzero_v2r8();
1203             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1204             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1205             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1206             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1207
1208             fscal            = felec;
1209
1210             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1211
1212             /* Update vectorial force */
1213             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1214             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1215             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1216
1217             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1218             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1219             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1220
1221             /**************************
1222              * CALCULATE INTERACTIONS *
1223              **************************/
1224
1225             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1226
1227             /* Compute parameters for interactions between i and j atoms */
1228             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1229
1230             /* Calculate table index by multiplying r with table scale and truncate to integer */
1231             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1232             itab_tmp         = _fjsp_dtox_v2r8(rt);
1233             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1234             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1235             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1236
1237             vfconv.i[0]     *= 12;
1238             vfconv.i[1]     *= 12;
1239
1240             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1241             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1242             F                = _fjsp_setzero_v2r8();
1243             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1244             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1245             H                = _fjsp_setzero_v2r8();
1246             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1247             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1248             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1249             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1250
1251             fscal            = felec;
1252
1253             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1254
1255             /* Update vectorial force */
1256             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1257             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1258             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1259
1260             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1261             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1262             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1263
1264             /**************************
1265              * CALCULATE INTERACTIONS *
1266              **************************/
1267
1268             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1269
1270             /* Compute parameters for interactions between i and j atoms */
1271             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1272
1273             /* Calculate table index by multiplying r with table scale and truncate to integer */
1274             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1275             itab_tmp         = _fjsp_dtox_v2r8(rt);
1276             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1277             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1278             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1279
1280             vfconv.i[0]     *= 12;
1281             vfconv.i[1]     *= 12;
1282
1283             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1284             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1285             F                = _fjsp_setzero_v2r8();
1286             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1287             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1288             H                = _fjsp_setzero_v2r8();
1289             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1290             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1291             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1292             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1293
1294             fscal            = felec;
1295
1296             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1297
1298             /* Update vectorial force */
1299             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1300             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1301             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1302
1303             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1304             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1305             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1306
1307             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1308
1309             /* Inner loop uses 180 flops */
1310         }
1311
1312         /* End of innermost loop */
1313
1314         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1315                                               f+i_coord_offset,fshift+i_shift_offset);
1316
1317         /* Increment number of inner iterations */
1318         inneriter                  += j_index_end - j_index_start;
1319
1320         /* Outer loop uses 24 flops */
1321     }
1322
1323     /* Increment number of outer iterations */
1324     outeriter        += nri;
1325
1326     /* Update outer/inner flops */
1327
1328     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*180);
1329 }