src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_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_ElecEwSw_VdwLJSw_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_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              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 104     real             *ewtab;
 105     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 106     real             rswitch_scalar,d_scalar;
 107     _fjsp_v2r8       itab_tmp;
 108     _fjsp_v2r8       dummy_mask,cutoff_mask;
 109     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 110     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 111     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 112
 113     x                = xx[0];
 114     f                = ff[0];
 115
 116     nri              = nlist->nri;
 117     iinr             = nlist->iinr;
 118     jindex           = nlist->jindex;
 119     jjnr             = nlist->jjnr;
 120     shiftidx         = nlist->shift;
 121     gid              = nlist->gid;
 122     shiftvec         = fr->shift_vec[0];
 123     fshift           = fr->fshift[0];
 124     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 125     charge           = mdatoms->chargeA;
 126     nvdwtype         = fr->ntype;
 127     vdwparam         = fr->nbfp;
 128     vdwtype          = mdatoms->typeA;
 129
 130     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 131     ewtab            = fr->ic->tabq_coul_FDV0;
 132     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 133     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 134
 135     /* Setup water-specific parameters */
 136     inr              = nlist->iinr[0];
 137     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 138     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 139     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 140     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 141
 142     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 143     rcutoff_scalar   = fr->rcoulomb;
 144     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 145     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 146
 147     rswitch_scalar   = fr->rcoulomb_switch;
 148     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 149     /* Setup switch parameters */
 150     d_scalar         = rcutoff_scalar-rswitch_scalar;
 151     d                = gmx_fjsp_set1_v2r8(d_scalar);
 152     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 153     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 154     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 155     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 156     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 157     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 158
 159     /* Avoid stupid compiler warnings */
 160     jnrA = jnrB = 0;
 161     j_coord_offsetA = 0;
 162     j_coord_offsetB = 0;
 163
 164     outeriter        = 0;
 165     inneriter        = 0;
 166
 167     /* Start outer loop over neighborlists */
 168     for(iidx=0; iidx<nri; iidx++)
 169     {
 170         /* Load shift vector for this list */
 171         i_shift_offset   = DIM*shiftidx[iidx];
 172
 173         /* Load limits for loop over neighbors */
 174         j_index_start    = jindex[iidx];
 175         j_index_end      = jindex[iidx+1];
 176
 177         /* Get outer coordinate index */
 178         inr              = iinr[iidx];
 179         i_coord_offset   = DIM*inr;
 180
 181         /* Load i particle coords and add shift vector */
 182         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 183                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 184
 185         fix0             = _fjsp_setzero_v2r8();
 186         fiy0             = _fjsp_setzero_v2r8();
 187         fiz0             = _fjsp_setzero_v2r8();
 188         fix1             = _fjsp_setzero_v2r8();
 189         fiy1             = _fjsp_setzero_v2r8();
 190         fiz1             = _fjsp_setzero_v2r8();
 191         fix2             = _fjsp_setzero_v2r8();
 192         fiy2             = _fjsp_setzero_v2r8();
 193         fiz2             = _fjsp_setzero_v2r8();
 194         fix3             = _fjsp_setzero_v2r8();
 195         fiy3             = _fjsp_setzero_v2r8();
 196         fiz3             = _fjsp_setzero_v2r8();
 197
 198         /* Reset potential sums */
 199         velecsum         = _fjsp_setzero_v2r8();
 200         vvdwsum          = _fjsp_setzero_v2r8();
 201
 202         /* Start inner kernel loop */
 203         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 204         {
 205
 206             /* Get j neighbor index, and coordinate index */
 207             jnrA             = jjnr[jidx];
 208             jnrB             = jjnr[jidx+1];
 209             j_coord_offsetA  = DIM*jnrA;
 210             j_coord_offsetB  = DIM*jnrB;
 211
 212             /* load j atom coordinates */
 213             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 214                                               &jx0,&jy0,&jz0);
 215
 216             /* Calculate displacement vector */
 217             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 218             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 219             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 220             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 221             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 222             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 223             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 224             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 225             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 226             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 227             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 228             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 229
 230             /* Calculate squared distance and things based on it */
 231             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 232             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 233             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 234             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 235
 236             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 237             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 238             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 239             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 240
 241             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 242             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 243             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 244             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 245
 246             /* Load parameters for j particles */
 247             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 248             vdwjidx0A        = 2*vdwtype[jnrA+0];
 249             vdwjidx0B        = 2*vdwtype[jnrB+0];
 250
 251             fjx0             = _fjsp_setzero_v2r8();
 252             fjy0             = _fjsp_setzero_v2r8();
 253             fjz0             = _fjsp_setzero_v2r8();
 254
 255             /**************************
 256              * CALCULATE INTERACTIONS *
 257              **************************/
 258
 259             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 260             {
 261
 262             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 263
 264             /* Compute parameters for interactions between i and j atoms */
 265             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 266                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 267
 268             /* LENNARD-JONES DISPERSION/REPULSION */
 269
 270             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 271             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 272             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 273             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 274             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 275
 276             d                = _fjsp_sub_v2r8(r00,rswitch);
 277             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 278             d2               = _fjsp_mul_v2r8(d,d);
 279             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 280
 281             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 282
 283             /* Evaluate switch function */
 284             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 285             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 286             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 287             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 291             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 292
 293             fscal            = fvdw;
 294
 295             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 296
 297             /* Update vectorial force */
 298             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 299             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 300             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 301
 302             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 303             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 304             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 305
 306             }
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 313             {
 314
 315             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 319
 320             /* EWALD ELECTROSTATICS */
 321
 322             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 323             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 324             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 325             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 326             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 327
 328             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 329             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 330             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 331             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 332             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 333             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 334             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 335             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 336             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 337             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 338
 339             d                = _fjsp_sub_v2r8(r10,rswitch);
 340             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 341             d2               = _fjsp_mul_v2r8(d,d);
 342             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 343
 344             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 345
 346             /* Evaluate switch function */
 347             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 348             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 349             velec            = _fjsp_mul_v2r8(velec,sw);
 350             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 354             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 355
 356             fscal            = felec;
 357
 358             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 359
 360             /* Update vectorial force */
 361             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 362             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 363             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 364
 365             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 366             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 367             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 368
 369             }
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 376             {
 377
 378             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 379
 380             /* Compute parameters for interactions between i and j atoms */
 381             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 382
 383             /* EWALD ELECTROSTATICS */
 384
 385             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 386             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 387             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 388             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 389             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 390
 391             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 392             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 393             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 394             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 395             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 396             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 397             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 398             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 399             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 400             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 401
 402             d                = _fjsp_sub_v2r8(r20,rswitch);
 403             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 404             d2               = _fjsp_mul_v2r8(d,d);
 405             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 406
 407             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 408
 409             /* Evaluate switch function */
 410             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 411             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 412             velec            = _fjsp_mul_v2r8(velec,sw);
 413             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 414
 415             /* Update potential sum for this i atom from the interaction with this j atom. */
 416             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 417             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 418
 419             fscal            = felec;
 420
 421             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 422
 423             /* Update vectorial force */
 424             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 425             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 426             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 427
 428             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 429             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 430             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 431
 432             }
 433
 434             /**************************
 435              * CALCULATE INTERACTIONS *
 436              **************************/
 437
 438             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 439             {
 440
 441             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 442
 443             /* Compute parameters for interactions between i and j atoms */
 444             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 445
 446             /* EWALD ELECTROSTATICS */
 447
 448             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 449             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 450             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 451             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 452             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 453
 454             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 455             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 456             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 457             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 458             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 459             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 460             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 461             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 462             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 463             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 464
 465             d                = _fjsp_sub_v2r8(r30,rswitch);
 466             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 467             d2               = _fjsp_mul_v2r8(d,d);
 468             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 469
 470             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 471
 472             /* Evaluate switch function */
 473             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 474             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
 475             velec            = _fjsp_mul_v2r8(velec,sw);
 476             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 477
 478             /* Update potential sum for this i atom from the interaction with this j atom. */
 479             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 480             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 481
 482             fscal            = felec;
 483
 484             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 485
 486             /* Update vectorial force */
 487             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 488             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 489             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 490
 491             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 492             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 493             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 494
 495             }
 496
 497             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 498
 499             /* Inner loop uses 269 flops */
 500         }
 501
 502         if(jidx<j_index_end)
 503         {
 504
 505             jnrA             = jjnr[jidx];
 506             j_coord_offsetA  = DIM*jnrA;
 507
 508             /* load j atom coordinates */
 509             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 510                                               &jx0,&jy0,&jz0);
 511
 512             /* Calculate displacement vector */
 513             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 514             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 515             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 516             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 517             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 518             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 519             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 520             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 521             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 522             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 523             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 524             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 525
 526             /* Calculate squared distance and things based on it */
 527             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 528             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 529             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 530             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 531
 532             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 533             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 534             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 535             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 536
 537             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 538             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 539             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 540             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 541
 542             /* Load parameters for j particles */
 543             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 544             vdwjidx0A        = 2*vdwtype[jnrA+0];
 545
 546             fjx0             = _fjsp_setzero_v2r8();
 547             fjy0             = _fjsp_setzero_v2r8();
 548             fjz0             = _fjsp_setzero_v2r8();
 549
 550             /**************************
 551              * CALCULATE INTERACTIONS *
 552              **************************/
 553
 554             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 555             {
 556
 557             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 558
 559             /* Compute parameters for interactions between i and j atoms */
 560             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 561                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 562
 563             /* LENNARD-JONES DISPERSION/REPULSION */
 564
 565             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 566             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 567             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 568             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 569             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 570
 571             d                = _fjsp_sub_v2r8(r00,rswitch);
 572             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 573             d2               = _fjsp_mul_v2r8(d,d);
 574             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 575
 576             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 577
 578             /* Evaluate switch function */
 579             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 580             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 581             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 582             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 583
 584             /* Update potential sum for this i atom from the interaction with this j atom. */
 585             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 586             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 587             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 588
 589             fscal            = fvdw;
 590
 591             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 592
 593             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 594
 595             /* Update vectorial force */
 596             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 597             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 598             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 599
 600             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 601             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 602             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 603
 604             }
 605
 606             /**************************
 607              * CALCULATE INTERACTIONS *
 608              **************************/
 609
 610             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 611             {
 612
 613             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 614
 615             /* Compute parameters for interactions between i and j atoms */
 616             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 617
 618             /* EWALD ELECTROSTATICS */
 619
 620             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 621             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 622             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 623             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 624             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 625
 626             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 627             ewtabD           = _fjsp_setzero_v2r8();
 628             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 629             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 630             ewtabFn          = _fjsp_setzero_v2r8();
 631             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 632             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 633             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 634             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 635             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 636
 637             d                = _fjsp_sub_v2r8(r10,rswitch);
 638             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 639             d2               = _fjsp_mul_v2r8(d,d);
 640             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 641
 642             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 643
 644             /* Evaluate switch function */
 645             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 646             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 647             velec            = _fjsp_mul_v2r8(velec,sw);
 648             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 649
 650             /* Update potential sum for this i atom from the interaction with this j atom. */
 651             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 652             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 653             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 654
 655             fscal            = felec;
 656
 657             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 658
 659             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 660
 661             /* Update vectorial force */
 662             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 663             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 664             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 665
 666             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 667             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 668             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 669
 670             }
 671
 672             /**************************
 673              * CALCULATE INTERACTIONS *
 674              **************************/
 675
 676             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 677             {
 678
 679             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 680
 681             /* Compute parameters for interactions between i and j atoms */
 682             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 683
 684             /* EWALD ELECTROSTATICS */
 685
 686             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 687             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 688             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 689             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 690             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 691
 692             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 693             ewtabD           = _fjsp_setzero_v2r8();
 694             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 695             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 696             ewtabFn          = _fjsp_setzero_v2r8();
 697             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 698             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 699             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 700             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 701             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 702
 703             d                = _fjsp_sub_v2r8(r20,rswitch);
 704             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 705             d2               = _fjsp_mul_v2r8(d,d);
 706             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 707
 708             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 709
 710             /* Evaluate switch function */
 711             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 712             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 713             velec            = _fjsp_mul_v2r8(velec,sw);
 714             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 715
 716             /* Update potential sum for this i atom from the interaction with this j atom. */
 717             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 718             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 719             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 720
 721             fscal            = felec;
 722
 723             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 724
 725             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 726
 727             /* Update vectorial force */
 728             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 729             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 730             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 731
 732             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 733             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 734             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 735
 736             }
 737
 738             /**************************
 739              * CALCULATE INTERACTIONS *
 740              **************************/
 741
 742             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 743             {
 744
 745             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 746
 747             /* Compute parameters for interactions between i and j atoms */
 748             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 749
 750             /* EWALD ELECTROSTATICS */
 751
 752             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 753             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 754             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 755             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 756             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 757
 758             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 759             ewtabD           = _fjsp_setzero_v2r8();
 760             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 761             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 762             ewtabFn          = _fjsp_setzero_v2r8();
 763             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 764             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 765             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 766             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 767             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 768
 769             d                = _fjsp_sub_v2r8(r30,rswitch);
 770             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 771             d2               = _fjsp_mul_v2r8(d,d);
 772             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 773
 774             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 775
 776             /* Evaluate switch function */
 777             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 778             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
 779             velec            = _fjsp_mul_v2r8(velec,sw);
 780             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 781
 782             /* Update potential sum for this i atom from the interaction with this j atom. */
 783             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 784             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 785             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 786
 787             fscal            = felec;
 788
 789             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 790
 791             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 792
 793             /* Update vectorial force */
 794             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 795             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 796             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 797
 798             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 799             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 800             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 801
 802             }
 803
 804             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 805
 806             /* Inner loop uses 269 flops */
 807         }
 808
 809         /* End of innermost loop */
 810
 811         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 812                                               f+i_coord_offset,fshift+i_shift_offset);
 813
 814         ggid                        = gid[iidx];
 815         /* Update potential energies */
 816         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 817         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 818
 819         /* Increment number of inner iterations */
 820         inneriter                  += j_index_end - j_index_start;
 821
 822         /* Outer loop uses 26 flops */
 823     }
 824
 825     /* Increment number of outer iterations */
 826     outeriter        += nri;
 827
 828     /* Update outer/inner flops */
 829
 830     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*269);
 831 }
 832 /*
 833  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 834  * Electrostatics interaction: Ewald
 835  * VdW interaction:            LennardJones
 836  * Geometry:                   Water4-Particle
 837  * Calculate force/pot:        Force
 838  */
 839 void
 840 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 841                     (t_nblist                    * gmx_restrict       nlist,
 842                      rvec                        * gmx_restrict          xx,
 843                      rvec                        * gmx_restrict          ff,
 844                      t_forcerec                  * gmx_restrict          fr,
 845                      t_mdatoms                   * gmx_restrict     mdatoms,
 846                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 847                      t_nrnb                      * gmx_restrict        nrnb)
 848 {
 849     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 850      * just 0 for non-waters.
 851      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 852      * jnr indices corresponding to data put in the four positions in the SIMD register.
 853      */
 854     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 855     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 856     int              jnrA,jnrB;
 857     int              j_coord_offsetA,j_coord_offsetB;
 858     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 859     real             rcutoff_scalar;
 860     real             *shiftvec,*fshift,*x,*f;
 861     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 862     int              vdwioffset0;
 863     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 864     int              vdwioffset1;
 865     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 866     int              vdwioffset2;
 867     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 868     int              vdwioffset3;
 869     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 870     int              vdwjidx0A,vdwjidx0B;
 871     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 872     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 873     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 874     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 875     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 876     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 877     real             *charge;
 878     int              nvdwtype;
 879     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 880     int              *vdwtype;
 881     real             *vdwparam;
 882     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 883     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 884     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 885     real             *ewtab;
 886     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 887     real             rswitch_scalar,d_scalar;
 888     _fjsp_v2r8       itab_tmp;
 889     _fjsp_v2r8       dummy_mask,cutoff_mask;
 890     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 891     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 892     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 893
 894     x                = xx[0];
 895     f                = ff[0];
 896
 897     nri              = nlist->nri;
 898     iinr             = nlist->iinr;
 899     jindex           = nlist->jindex;
 900     jjnr             = nlist->jjnr;
 901     shiftidx         = nlist->shift;
 902     gid              = nlist->gid;
 903     shiftvec         = fr->shift_vec[0];
 904     fshift           = fr->fshift[0];
 905     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 906     charge           = mdatoms->chargeA;
 907     nvdwtype         = fr->ntype;
 908     vdwparam         = fr->nbfp;
 909     vdwtype          = mdatoms->typeA;
 910
 911     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 912     ewtab            = fr->ic->tabq_coul_FDV0;
 913     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 914     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 915
 916     /* Setup water-specific parameters */
 917     inr              = nlist->iinr[0];
 918     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 919     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 920     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 921     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 922
 923     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 924     rcutoff_scalar   = fr->rcoulomb;
 925     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 926     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 927
 928     rswitch_scalar   = fr->rcoulomb_switch;
 929     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 930     /* Setup switch parameters */
 931     d_scalar         = rcutoff_scalar-rswitch_scalar;
 932     d                = gmx_fjsp_set1_v2r8(d_scalar);
 933     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 934     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 935     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 936     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 937     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 938     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 939
 940     /* Avoid stupid compiler warnings */
 941     jnrA = jnrB = 0;
 942     j_coord_offsetA = 0;
 943     j_coord_offsetB = 0;
 944
 945     outeriter        = 0;
 946     inneriter        = 0;
 947
 948     /* Start outer loop over neighborlists */
 949     for(iidx=0; iidx<nri; iidx++)
 950     {
 951         /* Load shift vector for this list */
 952         i_shift_offset   = DIM*shiftidx[iidx];
 953
 954         /* Load limits for loop over neighbors */
 955         j_index_start    = jindex[iidx];
 956         j_index_end      = jindex[iidx+1];
 957
 958         /* Get outer coordinate index */
 959         inr              = iinr[iidx];
 960         i_coord_offset   = DIM*inr;
 961
 962         /* Load i particle coords and add shift vector */
 963         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 964                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 965
 966         fix0             = _fjsp_setzero_v2r8();
 967         fiy0             = _fjsp_setzero_v2r8();
 968         fiz0             = _fjsp_setzero_v2r8();
 969         fix1             = _fjsp_setzero_v2r8();
 970         fiy1             = _fjsp_setzero_v2r8();
 971         fiz1             = _fjsp_setzero_v2r8();
 972         fix2             = _fjsp_setzero_v2r8();
 973         fiy2             = _fjsp_setzero_v2r8();
 974         fiz2             = _fjsp_setzero_v2r8();
 975         fix3             = _fjsp_setzero_v2r8();
 976         fiy3             = _fjsp_setzero_v2r8();
 977         fiz3             = _fjsp_setzero_v2r8();
 978
 979         /* Start inner kernel loop */
 980         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 981         {
 982
 983             /* Get j neighbor index, and coordinate index */
 984             jnrA             = jjnr[jidx];
 985             jnrB             = jjnr[jidx+1];
 986             j_coord_offsetA  = DIM*jnrA;
 987             j_coord_offsetB  = DIM*jnrB;
 988
 989             /* load j atom coordinates */
 990             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 991                                               &jx0,&jy0,&jz0);
 992
 993             /* Calculate displacement vector */
 994             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 995             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 996             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 997             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 998             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 999             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1000             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1001             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1002             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1003             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1004             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1005             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1006
1007             /* Calculate squared distance and things based on it */
1008             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1009             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1010             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1011             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1012
1013             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1014             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1015             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1016             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1017
1018             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1019             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1020             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1021             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1022
1023             /* Load parameters for j particles */
1024             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
1025             vdwjidx0A        = 2*vdwtype[jnrA+0];
1026             vdwjidx0B        = 2*vdwtype[jnrB+0];
1027
1028             fjx0             = _fjsp_setzero_v2r8();
1029             fjy0             = _fjsp_setzero_v2r8();
1030             fjz0             = _fjsp_setzero_v2r8();
1031
1032             /**************************
1033              * CALCULATE INTERACTIONS *
1034              **************************/
1035
1036             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1037             {
1038
1039             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1040
1041             /* Compute parameters for interactions between i and j atoms */
1042             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1043                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1044
1045             /* LENNARD-JONES DISPERSION/REPULSION */
1046
1047             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1048             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1049             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1050             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1051             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1052
1053             d                = _fjsp_sub_v2r8(r00,rswitch);
1054             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1055             d2               = _fjsp_mul_v2r8(d,d);
1056             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1057
1058             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1059
1060             /* Evaluate switch function */
1061             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1062             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1063             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1064
1065             fscal            = fvdw;
1066
1067             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1068
1069             /* Update vectorial force */
1070             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1071             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1072             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1073
1074             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1075             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1076             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1077
1078             }
1079
1080             /**************************
1081              * CALCULATE INTERACTIONS *
1082              **************************/
1083
1084             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1085             {
1086
1087             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1088
1089             /* Compute parameters for interactions between i and j atoms */
1090             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1091
1092             /* EWALD ELECTROSTATICS */
1093
1094             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1095             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1096             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1097             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1098             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1099
1100             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1101             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1102             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1103             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1104             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1105             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1106             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1107             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1108             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1109             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1110
1111             d                = _fjsp_sub_v2r8(r10,rswitch);
1112             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1113             d2               = _fjsp_mul_v2r8(d,d);
1114             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1115
1116             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1117
1118             /* Evaluate switch function */
1119             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1120             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1121             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1122
1123             fscal            = felec;
1124
1125             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1126
1127             /* Update vectorial force */
1128             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1129             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1130             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1131
1132             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1133             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1134             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1135
1136             }
1137
1138             /**************************
1139              * CALCULATE INTERACTIONS *
1140              **************************/
1141
1142             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1143             {
1144
1145             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1146
1147             /* Compute parameters for interactions between i and j atoms */
1148             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1149
1150             /* EWALD ELECTROSTATICS */
1151
1152             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1153             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1154             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1155             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1156             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1157
1158             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1159             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1160             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1161             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1162             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1163             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1164             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1165             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1166             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1167             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1168
1169             d                = _fjsp_sub_v2r8(r20,rswitch);
1170             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1171             d2               = _fjsp_mul_v2r8(d,d);
1172             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1173
1174             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1175
1176             /* Evaluate switch function */
1177             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1178             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1179             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1180
1181             fscal            = felec;
1182
1183             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1184
1185             /* Update vectorial force */
1186             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1187             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1188             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1189
1190             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1191             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1192             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1193
1194             }
1195
1196             /**************************
1197              * CALCULATE INTERACTIONS *
1198              **************************/
1199
1200             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1201             {
1202
1203             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1204
1205             /* Compute parameters for interactions between i and j atoms */
1206             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1207
1208             /* EWALD ELECTROSTATICS */
1209
1210             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1211             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1212             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1213             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1214             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1215
1216             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1217             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1218             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1219             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1220             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1221             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1222             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1223             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1224             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1225             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1226
1227             d                = _fjsp_sub_v2r8(r30,rswitch);
1228             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1229             d2               = _fjsp_mul_v2r8(d,d);
1230             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1231
1232             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1233
1234             /* Evaluate switch function */
1235             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1236             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1237             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1238
1239             fscal            = felec;
1240
1241             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1242
1243             /* Update vectorial force */
1244             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1245             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1246             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1247
1248             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1249             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1250             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1251
1252             }
1253
1254             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1255
1256             /* Inner loop uses 257 flops */
1257         }
1258
1259         if(jidx<j_index_end)
1260         {
1261
1262             jnrA             = jjnr[jidx];
1263             j_coord_offsetA  = DIM*jnrA;
1264
1265             /* load j atom coordinates */
1266             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1267                                               &jx0,&jy0,&jz0);
1268
1269             /* Calculate displacement vector */
1270             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1271             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1272             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1273             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1274             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1275             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1276             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1277             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1278             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1279             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1280             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1281             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1282
1283             /* Calculate squared distance and things based on it */
1284             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1285             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1286             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1287             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1288
1289             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1290             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1291             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1292             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1293
1294             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1295             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1296             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1297             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1298
1299             /* Load parameters for j particles */
1300             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1301             vdwjidx0A        = 2*vdwtype[jnrA+0];
1302
1303             fjx0             = _fjsp_setzero_v2r8();
1304             fjy0             = _fjsp_setzero_v2r8();
1305             fjz0             = _fjsp_setzero_v2r8();
1306
1307             /**************************
1308              * CALCULATE INTERACTIONS *
1309              **************************/
1310
1311             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1312             {
1313
1314             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1315
1316             /* Compute parameters for interactions between i and j atoms */
1317             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1318                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1319
1320             /* LENNARD-JONES DISPERSION/REPULSION */
1321
1322             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1323             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1324             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1325             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1326             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1327
1328             d                = _fjsp_sub_v2r8(r00,rswitch);
1329             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1330             d2               = _fjsp_mul_v2r8(d,d);
1331             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1332
1333             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1334
1335             /* Evaluate switch function */
1336             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1337             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1338             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1339
1340             fscal            = fvdw;
1341
1342             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1343
1344             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1345
1346             /* Update vectorial force */
1347             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1348             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1349             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1350
1351             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1352             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1353             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1354
1355             }
1356
1357             /**************************
1358              * CALCULATE INTERACTIONS *
1359              **************************/
1360
1361             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1362             {
1363
1364             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1365
1366             /* Compute parameters for interactions between i and j atoms */
1367             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1368
1369             /* EWALD ELECTROSTATICS */
1370
1371             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1372             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1373             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1374             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1375             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1376
1377             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1378             ewtabD           = _fjsp_setzero_v2r8();
1379             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1380             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1381             ewtabFn          = _fjsp_setzero_v2r8();
1382             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1383             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1384             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1385             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1386             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1387
1388             d                = _fjsp_sub_v2r8(r10,rswitch);
1389             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1390             d2               = _fjsp_mul_v2r8(d,d);
1391             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1392
1393             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1394
1395             /* Evaluate switch function */
1396             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1397             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1398             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1399
1400             fscal            = felec;
1401
1402             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1403
1404             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1405
1406             /* Update vectorial force */
1407             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1408             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1409             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1410
1411             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1412             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1413             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1414
1415             }
1416
1417             /**************************
1418              * CALCULATE INTERACTIONS *
1419              **************************/
1420
1421             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1422             {
1423
1424             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1425
1426             /* Compute parameters for interactions between i and j atoms */
1427             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1428
1429             /* EWALD ELECTROSTATICS */
1430
1431             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1432             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1433             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1434             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1435             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1436
1437             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1438             ewtabD           = _fjsp_setzero_v2r8();
1439             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1440             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1441             ewtabFn          = _fjsp_setzero_v2r8();
1442             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1443             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1444             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1445             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1446             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1447
1448             d                = _fjsp_sub_v2r8(r20,rswitch);
1449             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1450             d2               = _fjsp_mul_v2r8(d,d);
1451             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1452
1453             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1454
1455             /* Evaluate switch function */
1456             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1457             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1458             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1459
1460             fscal            = felec;
1461
1462             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1463
1464             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1465
1466             /* Update vectorial force */
1467             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1468             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1469             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1470
1471             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1472             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1473             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1474
1475             }
1476
1477             /**************************
1478              * CALCULATE INTERACTIONS *
1479              **************************/
1480
1481             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1482             {
1483
1484             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1485
1486             /* Compute parameters for interactions between i and j atoms */
1487             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1488
1489             /* EWALD ELECTROSTATICS */
1490
1491             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1492             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1493             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1494             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1495             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1496
1497             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1498             ewtabD           = _fjsp_setzero_v2r8();
1499             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1500             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1501             ewtabFn          = _fjsp_setzero_v2r8();
1502             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1503             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1504             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1505             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1506             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1507
1508             d                = _fjsp_sub_v2r8(r30,rswitch);
1509             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1510             d2               = _fjsp_mul_v2r8(d,d);
1511             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1512
1513             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1514
1515             /* Evaluate switch function */
1516             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1517             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1518             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1519
1520             fscal            = felec;
1521
1522             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1523
1524             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1525
1526             /* Update vectorial force */
1527             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1528             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1529             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1530
1531             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1532             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1533             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1534
1535             }
1536
1537             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1538
1539             /* Inner loop uses 257 flops */
1540         }
1541
1542         /* End of innermost loop */
1543
1544         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1545                                               f+i_coord_offset,fshift+i_shift_offset);
1546
1547         /* Increment number of inner iterations */
1548         inneriter                  += j_index_end - j_index_start;
1549
1550         /* Outer loop uses 24 flops */
1551     }
1552
1553     /* Increment number of outer iterations */
1554     outeriter        += nri;
1555
1556     /* Update outer/inner flops */
1557
1558     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*257);
1559 }