src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3P1_sparc64_hpc_ace_double.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 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/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LJEwald
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  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       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8           c6grid_00;
 101     _fjsp_v2r8           c6grid_10;
 102     _fjsp_v2r8           c6grid_20;
 103     real                 *vdwgridparam;
 104     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 105     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 106     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 107     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 108     real             *ewtab;
 109     _fjsp_v2r8       itab_tmp;
 110     _fjsp_v2r8       dummy_mask,cutoff_mask;
 111     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 112     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 113     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 114
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128     nvdwtype         = fr->ntype;
 129     vdwparam         = fr->nbfp;
 130     vdwtype          = mdatoms->typeA;
 131     vdwgridparam     = fr->ljpme_c6grid;
 132     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 133     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 134     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 135
 136     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 137     ewtab            = fr->ic->tabq_coul_FDV0;
 138     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 139     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 140
 141     /* Setup water-specific parameters */
 142     inr              = nlist->iinr[0];
 143     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 144     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 145     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 146     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 147
 148     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 149     rcutoff_scalar   = fr->rcoulomb;
 150     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 151     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 152
 153     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 154     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 155
 156     /* Avoid stupid compiler warnings */
 157     jnrA = jnrB = 0;
 158     j_coord_offsetA = 0;
 159     j_coord_offsetB = 0;
 160
 161     outeriter        = 0;
 162     inneriter        = 0;
 163
 164     /* Start outer loop over neighborlists */
 165     for(iidx=0; iidx<nri; iidx++)
 166     {
 167         /* Load shift vector for this list */
 168         i_shift_offset   = DIM*shiftidx[iidx];
 169
 170         /* Load limits for loop over neighbors */
 171         j_index_start    = jindex[iidx];
 172         j_index_end      = jindex[iidx+1];
 173
 174         /* Get outer coordinate index */
 175         inr              = iinr[iidx];
 176         i_coord_offset   = DIM*inr;
 177
 178         /* Load i particle coords and add shift vector */
 179         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 180                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 181
 182         fix0             = _fjsp_setzero_v2r8();
 183         fiy0             = _fjsp_setzero_v2r8();
 184         fiz0             = _fjsp_setzero_v2r8();
 185         fix1             = _fjsp_setzero_v2r8();
 186         fiy1             = _fjsp_setzero_v2r8();
 187         fiz1             = _fjsp_setzero_v2r8();
 188         fix2             = _fjsp_setzero_v2r8();
 189         fiy2             = _fjsp_setzero_v2r8();
 190         fiz2             = _fjsp_setzero_v2r8();
 191
 192         /* Reset potential sums */
 193         velecsum         = _fjsp_setzero_v2r8();
 194         vvdwsum          = _fjsp_setzero_v2r8();
 195
 196         /* Start inner kernel loop */
 197         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 198         {
 199
 200             /* Get j neighbor index, and coordinate index */
 201             jnrA             = jjnr[jidx];
 202             jnrB             = jjnr[jidx+1];
 203             j_coord_offsetA  = DIM*jnrA;
 204             j_coord_offsetB  = DIM*jnrB;
 205
 206             /* load j atom coordinates */
 207             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 208                                               &jx0,&jy0,&jz0);
 209
 210             /* Calculate displacement vector */
 211             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 212             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 213             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 214             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 215             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 216             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 217             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 218             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 219             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 220
 221             /* Calculate squared distance and things based on it */
 222             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 223             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 224             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 225
 226             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 227             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 228             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 229
 230             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 231             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 232             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 233
 234             /* Load parameters for j particles */
 235             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 236             vdwjidx0A        = 2*vdwtype[jnrA+0];
 237             vdwjidx0B        = 2*vdwtype[jnrB+0];
 238
 239             fjx0             = _fjsp_setzero_v2r8();
 240             fjy0             = _fjsp_setzero_v2r8();
 241             fjz0             = _fjsp_setzero_v2r8();
 242
 243             /**************************
 244              * CALCULATE INTERACTIONS *
 245              **************************/
 246
 247             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 248             {
 249
 250             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 251
 252             /* Compute parameters for interactions between i and j atoms */
 253             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 254             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 255                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 256
 257             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 258                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 259
 260             /* EWALD ELECTROSTATICS */
 261
 262             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 263             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 264             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 265             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 266             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 267
 268             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 269             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 270             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 271             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 272             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 273             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 274             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 275             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 276             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 277             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 278
 279             /* Analytical LJ-PME */
 280             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 281             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 282             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 283             exponent         = gmx_simd_exp_d(-ewcljrsq);
 284             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 285             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 286             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 287             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 288             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 289             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 290                                _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
 291             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 292             fvdw             = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
 293
 294             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 295
 296             /* Update potential sum for this i atom from the interaction with this j atom. */
 297             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 298             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 299             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 300             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 301
 302             fscal            = _fjsp_add_v2r8(felec,fvdw);
 303
 304             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 305
 306             /* Update vectorial force */
 307             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 308             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 309             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 310
 311             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 312             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 313             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 314
 315             }
 316
 317             /**************************
 318              * CALCULATE INTERACTIONS *
 319              **************************/
 320
 321             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 322             {
 323
 324             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 325
 326             /* Compute parameters for interactions between i and j atoms */
 327             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 328
 329             /* EWALD ELECTROSTATICS */
 330
 331             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 332             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 333             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 334             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 335             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 336
 337             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 338             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 339             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 340             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 341             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 342             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 343             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 344             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 345             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 346             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 347
 348             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 349
 350             /* Update potential sum for this i atom from the interaction with this j atom. */
 351             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 352             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 353
 354             fscal            = felec;
 355
 356             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 357
 358             /* Update vectorial force */
 359             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 360             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 361             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 362
 363             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 364             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 365             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 366
 367             }
 368
 369             /**************************
 370              * CALCULATE INTERACTIONS *
 371              **************************/
 372
 373             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 374             {
 375
 376             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 377
 378             /* Compute parameters for interactions between i and j atoms */
 379             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 380
 381             /* EWALD ELECTROSTATICS */
 382
 383             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 384             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 385             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 386             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 387             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 388
 389             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 390             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 391             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 392             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 393             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 394             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 395             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 396             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 397             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 398             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 399
 400             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 401
 402             /* Update potential sum for this i atom from the interaction with this j atom. */
 403             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 404             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 405
 406             fscal            = felec;
 407
 408             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 409
 410             /* Update vectorial force */
 411             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 412             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 413             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 414
 415             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 416             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 417             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 418
 419             }
 420
 421             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 422
 423             /* Inner loop uses 180 flops */
 424         }
 425
 426         if(jidx<j_index_end)
 427         {
 428
 429             jnrA             = jjnr[jidx];
 430             j_coord_offsetA  = DIM*jnrA;
 431
 432             /* load j atom coordinates */
 433             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 434                                               &jx0,&jy0,&jz0);
 435
 436             /* Calculate displacement vector */
 437             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 438             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 439             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 440             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 441             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 442             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 443             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 444             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 445             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 446
 447             /* Calculate squared distance and things based on it */
 448             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 449             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 450             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 451
 452             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 453             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 454             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 455
 456             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 457             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 458             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 459
 460             /* Load parameters for j particles */
 461             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 462             vdwjidx0A        = 2*vdwtype[jnrA+0];
 463
 464             fjx0             = _fjsp_setzero_v2r8();
 465             fjy0             = _fjsp_setzero_v2r8();
 466             fjz0             = _fjsp_setzero_v2r8();
 467
 468             /**************************
 469              * CALCULATE INTERACTIONS *
 470              **************************/
 471
 472             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 473             {
 474
 475             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 476
 477             /* Compute parameters for interactions between i and j atoms */
 478             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 479             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 480
 481             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
 482
 483             /* EWALD ELECTROSTATICS */
 484
 485             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 486             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 487             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 488             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 489             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 490
 491             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 492             ewtabD           = _fjsp_setzero_v2r8();
 493             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 494             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 495             ewtabFn          = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 497             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 498             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 499             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 500             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 501
 502             /* Analytical LJ-PME */
 503             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 504             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 505             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 506             exponent         = gmx_simd_exp_d(-ewcljrsq);
 507             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 508             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 509             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 510             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 511             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 512             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 513                                _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
 514             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 515             fvdw             = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
 516
 517             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 518
 519             /* Update potential sum for this i atom from the interaction with this j atom. */
 520             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 521             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 522             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 523             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 524             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 525             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 526
 527             fscal            = _fjsp_add_v2r8(felec,fvdw);
 528
 529             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 530
 531             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 532
 533             /* Update vectorial force */
 534             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 535             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 536             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 537
 538             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 539             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 540             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 541
 542             }
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 549             {
 550
 551             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 552
 553             /* Compute parameters for interactions between i and j atoms */
 554             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 555
 556             /* EWALD ELECTROSTATICS */
 557
 558             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 559             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 560             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 561             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 562             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 563
 564             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 565             ewtabD           = _fjsp_setzero_v2r8();
 566             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 567             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 568             ewtabFn          = _fjsp_setzero_v2r8();
 569             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 570             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 571             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 572             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 573             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 574
 575             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 576
 577             /* Update potential sum for this i atom from the interaction with this j atom. */
 578             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 579             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 580             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 581
 582             fscal            = felec;
 583
 584             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 585
 586             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 587
 588             /* Update vectorial force */
 589             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 590             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 591             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 592
 593             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 594             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 595             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 596
 597             }
 598
 599             /**************************
 600              * CALCULATE INTERACTIONS *
 601              **************************/
 602
 603             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 604             {
 605
 606             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 607
 608             /* Compute parameters for interactions between i and j atoms */
 609             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 610
 611             /* EWALD ELECTROSTATICS */
 612
 613             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 614             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 615             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 616             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 617             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 618
 619             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 620             ewtabD           = _fjsp_setzero_v2r8();
 621             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 622             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 623             ewtabFn          = _fjsp_setzero_v2r8();
 624             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 625             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 626             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 627             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 628             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 629
 630             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 631
 632             /* Update potential sum for this i atom from the interaction with this j atom. */
 633             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 634             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 635             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 636
 637             fscal            = felec;
 638
 639             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 640
 641             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 642
 643             /* Update vectorial force */
 644             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 645             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 646             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 647
 648             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 649             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 650             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 651
 652             }
 653
 654             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 655
 656             /* Inner loop uses 180 flops */
 657         }
 658
 659         /* End of innermost loop */
 660
 661         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 662                                               f+i_coord_offset,fshift+i_shift_offset);
 663
 664         ggid                        = gid[iidx];
 665         /* Update potential energies */
 666         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 667         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 668
 669         /* Increment number of inner iterations */
 670         inneriter                  += j_index_end - j_index_start;
 671
 672         /* Outer loop uses 20 flops */
 673     }
 674
 675     /* Increment number of outer iterations */
 676     outeriter        += nri;
 677
 678     /* Update outer/inner flops */
 679
 680     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*180);
 681 }
 682 /*
 683  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3P1_F_sparc64_hpc_ace_double
 684  * Electrostatics interaction: Ewald
 685  * VdW interaction:            LJEwald
 686  * Geometry:                   Water3-Particle
 687  * Calculate force/pot:        Force
 688  */
 689 void
 690 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3P1_F_sparc64_hpc_ace_double
 691                     (t_nblist                    * gmx_restrict       nlist,
 692                      rvec                        * gmx_restrict          xx,
 693                      rvec                        * gmx_restrict          ff,
 694                      t_forcerec                  * gmx_restrict          fr,
 695                      t_mdatoms                   * gmx_restrict     mdatoms,
 696                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 697                      t_nrnb                      * gmx_restrict        nrnb)
 698 {
 699     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 700      * just 0 for non-waters.
 701      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 702      * jnr indices corresponding to data put in the four positions in the SIMD register.
 703      */
 704     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 705     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 706     int              jnrA,jnrB;
 707     int              j_coord_offsetA,j_coord_offsetB;
 708     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 709     real             rcutoff_scalar;
 710     real             *shiftvec,*fshift,*x,*f;
 711     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 712     int              vdwioffset0;
 713     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 714     int              vdwioffset1;
 715     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 716     int              vdwioffset2;
 717     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 718     int              vdwjidx0A,vdwjidx0B;
 719     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 720     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 721     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 722     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 723     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 724     real             *charge;
 725     int              nvdwtype;
 726     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 727     int              *vdwtype;
 728     real             *vdwparam;
 729     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 730     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 731     _fjsp_v2r8           c6grid_00;
 732     _fjsp_v2r8           c6grid_10;
 733     _fjsp_v2r8           c6grid_20;
 734     real                 *vdwgridparam;
 735     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 736     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 737     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 738     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 739     real             *ewtab;
 740     _fjsp_v2r8       itab_tmp;
 741     _fjsp_v2r8       dummy_mask,cutoff_mask;
 742     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 743     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 744     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 745
 746     x                = xx[0];
 747     f                = ff[0];
 748
 749     nri              = nlist->nri;
 750     iinr             = nlist->iinr;
 751     jindex           = nlist->jindex;
 752     jjnr             = nlist->jjnr;
 753     shiftidx         = nlist->shift;
 754     gid              = nlist->gid;
 755     shiftvec         = fr->shift_vec[0];
 756     fshift           = fr->fshift[0];
 757     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 758     charge           = mdatoms->chargeA;
 759     nvdwtype         = fr->ntype;
 760     vdwparam         = fr->nbfp;
 761     vdwtype          = mdatoms->typeA;
 762     vdwgridparam     = fr->ljpme_c6grid;
 763     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 764     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 765     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 766
 767     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 768     ewtab            = fr->ic->tabq_coul_F;
 769     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 770     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 771
 772     /* Setup water-specific parameters */
 773     inr              = nlist->iinr[0];
 774     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 775     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 776     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 777     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 778
 779     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 780     rcutoff_scalar   = fr->rcoulomb;
 781     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 782     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 783
 784     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 785     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 786
 787     /* Avoid stupid compiler warnings */
 788     jnrA = jnrB = 0;
 789     j_coord_offsetA = 0;
 790     j_coord_offsetB = 0;
 791
 792     outeriter        = 0;
 793     inneriter        = 0;
 794
 795     /* Start outer loop over neighborlists */
 796     for(iidx=0; iidx<nri; iidx++)
 797     {
 798         /* Load shift vector for this list */
 799         i_shift_offset   = DIM*shiftidx[iidx];
 800
 801         /* Load limits for loop over neighbors */
 802         j_index_start    = jindex[iidx];
 803         j_index_end      = jindex[iidx+1];
 804
 805         /* Get outer coordinate index */
 806         inr              = iinr[iidx];
 807         i_coord_offset   = DIM*inr;
 808
 809         /* Load i particle coords and add shift vector */
 810         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 811                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 812
 813         fix0             = _fjsp_setzero_v2r8();
 814         fiy0             = _fjsp_setzero_v2r8();
 815         fiz0             = _fjsp_setzero_v2r8();
 816         fix1             = _fjsp_setzero_v2r8();
 817         fiy1             = _fjsp_setzero_v2r8();
 818         fiz1             = _fjsp_setzero_v2r8();
 819         fix2             = _fjsp_setzero_v2r8();
 820         fiy2             = _fjsp_setzero_v2r8();
 821         fiz2             = _fjsp_setzero_v2r8();
 822
 823         /* Start inner kernel loop */
 824         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 825         {
 826
 827             /* Get j neighbor index, and coordinate index */
 828             jnrA             = jjnr[jidx];
 829             jnrB             = jjnr[jidx+1];
 830             j_coord_offsetA  = DIM*jnrA;
 831             j_coord_offsetB  = DIM*jnrB;
 832
 833             /* load j atom coordinates */
 834             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 835                                               &jx0,&jy0,&jz0);
 836
 837             /* Calculate displacement vector */
 838             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 839             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 840             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 841             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 842             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 843             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 844             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 845             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 846             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 847
 848             /* Calculate squared distance and things based on it */
 849             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 850             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 851             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 852
 853             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 854             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 855             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 856
 857             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 858             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 859             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 860
 861             /* Load parameters for j particles */
 862             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 863             vdwjidx0A        = 2*vdwtype[jnrA+0];
 864             vdwjidx0B        = 2*vdwtype[jnrB+0];
 865
 866             fjx0             = _fjsp_setzero_v2r8();
 867             fjy0             = _fjsp_setzero_v2r8();
 868             fjz0             = _fjsp_setzero_v2r8();
 869
 870             /**************************
 871              * CALCULATE INTERACTIONS *
 872              **************************/
 873
 874             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 875             {
 876
 877             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 878
 879             /* Compute parameters for interactions between i and j atoms */
 880             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 881             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 882                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 883
 884             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 885                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 886
 887             /* EWALD ELECTROSTATICS */
 888
 889             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 890             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 891             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 892             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 893             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 894
 895             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 896                                          &ewtabF,&ewtabFn);
 897             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 898             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 899
 900             /* Analytical LJ-PME */
 901             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 902             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 903             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 904             exponent         = gmx_simd_exp_d(-ewcljrsq);
 905             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 906             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 907             /* f6A = 6 * C6grid * (1 - poly) */
 908             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
 909             /* f6B = C6grid * exponent * beta^6 */
 910             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
 911             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 912             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 913
 914             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 915
 916             fscal            = _fjsp_add_v2r8(felec,fvdw);
 917
 918             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 919
 920             /* Update vectorial force */
 921             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 922             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 923             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 924
 925             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 926             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 927             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 928
 929             }
 930
 931             /**************************
 932              * CALCULATE INTERACTIONS *
 933              **************************/
 934
 935             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 936             {
 937
 938             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 939
 940             /* Compute parameters for interactions between i and j atoms */
 941             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 942
 943             /* EWALD ELECTROSTATICS */
 944
 945             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 946             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 947             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 948             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 949             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 950
 951             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 952                                          &ewtabF,&ewtabFn);
 953             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 954             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 955
 956             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 957
 958             fscal            = felec;
 959
 960             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 961
 962             /* Update vectorial force */
 963             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 964             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 965             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 966
 967             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 968             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 969             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 970
 971             }
 972
 973             /**************************
 974              * CALCULATE INTERACTIONS *
 975              **************************/
 976
 977             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 978             {
 979
 980             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 981
 982             /* Compute parameters for interactions between i and j atoms */
 983             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 984
 985             /* EWALD ELECTROSTATICS */
 986
 987             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 988             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 989             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 990             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 991             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 992
 993             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 994                                          &ewtabF,&ewtabFn);
 995             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 996             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 997
 998             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 999
1000             fscal            = felec;
1001
1002             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1003
1004             /* Update vectorial force */
1005             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1006             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1007             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1008
1009             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1010             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1011             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1012
1013             }
1014
1015             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1016
1017             /* Inner loop uses 151 flops */
1018         }
1019
1020         if(jidx<j_index_end)
1021         {
1022
1023             jnrA             = jjnr[jidx];
1024             j_coord_offsetA  = DIM*jnrA;
1025
1026             /* load j atom coordinates */
1027             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1028                                               &jx0,&jy0,&jz0);
1029
1030             /* Calculate displacement vector */
1031             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1032             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1033             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1034             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1035             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1036             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1037             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1038             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1039             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1040
1041             /* Calculate squared distance and things based on it */
1042             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1043             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1044             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1045
1046             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1047             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1048             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1049
1050             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1051             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1052             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1053
1054             /* Load parameters for j particles */
1055             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1056             vdwjidx0A        = 2*vdwtype[jnrA+0];
1057
1058             fjx0             = _fjsp_setzero_v2r8();
1059             fjy0             = _fjsp_setzero_v2r8();
1060             fjz0             = _fjsp_setzero_v2r8();
1061
1062             /**************************
1063              * CALCULATE INTERACTIONS *
1064              **************************/
1065
1066             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1067             {
1068
1069             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1070
1071             /* Compute parameters for interactions between i and j atoms */
1072             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1073             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1074
1075             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
1076
1077             /* EWALD ELECTROSTATICS */
1078
1079             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1080             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1081             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1082             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1083             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1084
1085             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1086             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1087             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1088
1089             /* Analytical LJ-PME */
1090             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1091             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
1092             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
1093             exponent         = gmx_simd_exp_d(-ewcljrsq);
1094             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1095             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
1096             /* f6A = 6 * C6grid * (1 - poly) */
1097             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
1098             /* f6B = C6grid * exponent * beta^6 */
1099             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
1100             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1101             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1102
1103             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1104
1105             fscal            = _fjsp_add_v2r8(felec,fvdw);
1106
1107             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1108
1109             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1110
1111             /* Update vectorial force */
1112             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1113             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1114             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1115
1116             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1117             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1118             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1119
1120             }
1121
1122             /**************************
1123              * CALCULATE INTERACTIONS *
1124              **************************/
1125
1126             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1127             {
1128
1129             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1130
1131             /* Compute parameters for interactions between i and j atoms */
1132             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1133
1134             /* EWALD ELECTROSTATICS */
1135
1136             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1137             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1138             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1139             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1140             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1141
1142             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1143             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1144             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1145
1146             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1147
1148             fscal            = felec;
1149
1150             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1151
1152             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1153
1154             /* Update vectorial force */
1155             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1156             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1157             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1158
1159             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1160             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1161             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1162
1163             }
1164
1165             /**************************
1166              * CALCULATE INTERACTIONS *
1167              **************************/
1168
1169             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1170             {
1171
1172             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1173
1174             /* Compute parameters for interactions between i and j atoms */
1175             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1176
1177             /* EWALD ELECTROSTATICS */
1178
1179             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1180             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1181             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1182             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1183             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1184
1185             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1186             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1187             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1188
1189             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1190
1191             fscal            = felec;
1192
1193             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1194
1195             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1196
1197             /* Update vectorial force */
1198             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1199             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1200             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1201
1202             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1203             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1204             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1205
1206             }
1207
1208             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1209
1210             /* Inner loop uses 151 flops */
1211         }
1212
1213         /* End of innermost loop */
1214
1215         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1216                                               f+i_coord_offset,fshift+i_shift_offset);
1217
1218         /* Increment number of inner iterations */
1219         inneriter                  += j_index_end - j_index_start;
1220
1221         /* Outer loop uses 18 flops */
1222     }
1223
1224     /* Increment number of outer iterations */
1225     outeriter        += nri;
1226
1227     /* Update outer/inner flops */
1228
1229     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*151);
1230 }