src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSw_VdwLJSw_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_ElecEwSw_VdwLJSw_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSw_VdwLJSw_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       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 101     real             *ewtab;
 102     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 103     real             rswitch_scalar,d_scalar;
 104     _fjsp_v2r8       itab_tmp;
 105     _fjsp_v2r8       dummy_mask,cutoff_mask;
 106     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 107     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 108     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 109
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 128     ewtab            = fr->ic->tabq_coul_FDV0;
 129     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 130     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 131
 132     /* Setup water-specific parameters */
 133     inr              = nlist->iinr[0];
 134     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 135     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 136     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 137     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 138
 139     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 140     rcutoff_scalar   = fr->rcoulomb;
 141     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 142     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 143
 144     rswitch_scalar   = fr->rcoulomb_switch;
 145     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 146     /* Setup switch parameters */
 147     d_scalar         = rcutoff_scalar-rswitch_scalar;
 148     d                = gmx_fjsp_set1_v2r8(d_scalar);
 149     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 150     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 151     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 152     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 153     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 154     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 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             /* EWALD ELECTROSTATICS */
 258
 259             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 260             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 261             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 262             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 263             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 264
 265             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 266             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 267             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 268             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 269             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 270             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 271             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 272             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 273             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 274             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 275
 276             /* LENNARD-JONES DISPERSION/REPULSION */
 277
 278             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 279             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 280             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 281             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 282             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 283
 284             d                = _fjsp_sub_v2r8(r00,rswitch);
 285             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 286             d2               = _fjsp_mul_v2r8(d,d);
 287             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 288
 289             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 290
 291             /* Evaluate switch function */
 292             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 293             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 294             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 295             velec            = _fjsp_mul_v2r8(velec,sw);
 296             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 297             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 298
 299             /* Update potential sum for this i atom from the interaction with this j atom. */
 300             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 301             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 302             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 303             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 304
 305             fscal            = _fjsp_add_v2r8(felec,fvdw);
 306
 307             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 308
 309             /* Update vectorial force */
 310             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 311             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 312             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 313
 314             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 315             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 316             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 317
 318             }
 319
 320             /**************************
 321              * CALCULATE INTERACTIONS *
 322              **************************/
 323
 324             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 325             {
 326
 327             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 328
 329             /* Compute parameters for interactions between i and j atoms */
 330             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 331
 332             /* EWALD ELECTROSTATICS */
 333
 334             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 335             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 336             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 337             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 338             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 339
 340             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 341             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 342             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 343             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 344             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 345             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 346             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 347             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 348             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 349             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 350
 351             d                = _fjsp_sub_v2r8(r10,rswitch);
 352             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 353             d2               = _fjsp_mul_v2r8(d,d);
 354             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 355
 356             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 357
 358             /* Evaluate switch function */
 359             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 360             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 361             velec            = _fjsp_mul_v2r8(velec,sw);
 362             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 363
 364             /* Update potential sum for this i atom from the interaction with this j atom. */
 365             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 366             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 371
 372             /* Update vectorial force */
 373             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 374             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 375             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 376
 377             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 378             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 379             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 380
 381             }
 382
 383             /**************************
 384              * CALCULATE INTERACTIONS *
 385              **************************/
 386
 387             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 388             {
 389
 390             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 391
 392             /* Compute parameters for interactions between i and j atoms */
 393             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 394
 395             /* EWALD ELECTROSTATICS */
 396
 397             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 398             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 399             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 400             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 401             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 402
 403             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 404             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 405             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 406             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 407             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 408             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 409             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 410             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 411             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 412             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 413
 414             d                = _fjsp_sub_v2r8(r20,rswitch);
 415             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 416             d2               = _fjsp_mul_v2r8(d,d);
 417             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 418
 419             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 420
 421             /* Evaluate switch function */
 422             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 423             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 424             velec            = _fjsp_mul_v2r8(velec,sw);
 425             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 426
 427             /* Update potential sum for this i atom from the interaction with this j atom. */
 428             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 429             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 434
 435             /* Update vectorial force */
 436             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 437             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 438             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 439
 440             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 441             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 442             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 443
 444             }
 445
 446             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 447
 448             /* Inner loop uses 225 flops */
 449         }
 450
 451         if(jidx<j_index_end)
 452         {
 453
 454             jnrA             = jjnr[jidx];
 455             j_coord_offsetA  = DIM*jnrA;
 456
 457             /* load j atom coordinates */
 458             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 459                                               &jx0,&jy0,&jz0);
 460
 461             /* Calculate displacement vector */
 462             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 463             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 464             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 465             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 466             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 467             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 468             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 469             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 470             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 471
 472             /* Calculate squared distance and things based on it */
 473             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 474             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 475             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 476
 477             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 478             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 479             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 480
 481             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 482             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 483             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 484
 485             /* Load parameters for j particles */
 486             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 487             vdwjidx0A        = 2*vdwtype[jnrA+0];
 488
 489             fjx0             = _fjsp_setzero_v2r8();
 490             fjy0             = _fjsp_setzero_v2r8();
 491             fjz0             = _fjsp_setzero_v2r8();
 492
 493             /**************************
 494              * CALCULATE INTERACTIONS *
 495              **************************/
 496
 497             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 498             {
 499
 500             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 501
 502             /* Compute parameters for interactions between i and j atoms */
 503             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 504             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 505
 506             /* EWALD ELECTROSTATICS */
 507
 508             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 509             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 510             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 511             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 512             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 513
 514             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 515             ewtabD           = _fjsp_setzero_v2r8();
 516             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 517             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 518             ewtabFn          = _fjsp_setzero_v2r8();
 519             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 520             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 521             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 522             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 523             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 524
 525             /* LENNARD-JONES DISPERSION/REPULSION */
 526
 527             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 528             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 529             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 530             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 531             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 532
 533             d                = _fjsp_sub_v2r8(r00,rswitch);
 534             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 535             d2               = _fjsp_mul_v2r8(d,d);
 536             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 537
 538             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 539
 540             /* Evaluate switch function */
 541             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 542             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 543             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 544             velec            = _fjsp_mul_v2r8(velec,sw);
 545             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 546             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 547
 548             /* Update potential sum for this i atom from the interaction with this j atom. */
 549             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 550             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 551             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 552             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 553             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 554             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 555
 556             fscal            = _fjsp_add_v2r8(felec,fvdw);
 557
 558             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 559
 560             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 561
 562             /* Update vectorial force */
 563             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 564             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 565             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 566
 567             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 568             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 569             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 570
 571             }
 572
 573             /**************************
 574              * CALCULATE INTERACTIONS *
 575              **************************/
 576
 577             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 578             {
 579
 580             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 581
 582             /* Compute parameters for interactions between i and j atoms */
 583             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 584
 585             /* EWALD ELECTROSTATICS */
 586
 587             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 588             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 589             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 590             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 591             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 592
 593             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 594             ewtabD           = _fjsp_setzero_v2r8();
 595             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 596             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 597             ewtabFn          = _fjsp_setzero_v2r8();
 598             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 599             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 600             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 601             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 602             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 603
 604             d                = _fjsp_sub_v2r8(r10,rswitch);
 605             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 606             d2               = _fjsp_mul_v2r8(d,d);
 607             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 608
 609             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 610
 611             /* Evaluate switch function */
 612             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 613             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 614             velec            = _fjsp_mul_v2r8(velec,sw);
 615             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 616
 617             /* Update potential sum for this i atom from the interaction with this j atom. */
 618             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 619             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 620             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 621
 622             fscal            = felec;
 623
 624             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 625
 626             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 627
 628             /* Update vectorial force */
 629             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 630             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 631             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 632
 633             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 634             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 635             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 636
 637             }
 638
 639             /**************************
 640              * CALCULATE INTERACTIONS *
 641              **************************/
 642
 643             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 644             {
 645
 646             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 647
 648             /* Compute parameters for interactions between i and j atoms */
 649             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 650
 651             /* EWALD ELECTROSTATICS */
 652
 653             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 654             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 655             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 656             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 657             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 658
 659             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 660             ewtabD           = _fjsp_setzero_v2r8();
 661             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 662             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 663             ewtabFn          = _fjsp_setzero_v2r8();
 664             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 665             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 666             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 667             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 668             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 669
 670             d                = _fjsp_sub_v2r8(r20,rswitch);
 671             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 672             d2               = _fjsp_mul_v2r8(d,d);
 673             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 674
 675             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 676
 677             /* Evaluate switch function */
 678             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 679             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 680             velec            = _fjsp_mul_v2r8(velec,sw);
 681             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 682
 683             /* Update potential sum for this i atom from the interaction with this j atom. */
 684             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 685             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 686             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 687
 688             fscal            = felec;
 689
 690             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 691
 692             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 693
 694             /* Update vectorial force */
 695             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 696             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 697             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 698
 699             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 700             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 701             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 702
 703             }
 704
 705             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 706
 707             /* Inner loop uses 225 flops */
 708         }
 709
 710         /* End of innermost loop */
 711
 712         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 713                                               f+i_coord_offset,fshift+i_shift_offset);
 714
 715         ggid                        = gid[iidx];
 716         /* Update potential energies */
 717         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 718         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 719
 720         /* Increment number of inner iterations */
 721         inneriter                  += j_index_end - j_index_start;
 722
 723         /* Outer loop uses 20 flops */
 724     }
 725
 726     /* Increment number of outer iterations */
 727     outeriter        += nri;
 728
 729     /* Update outer/inner flops */
 730
 731     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*225);
 732 }
 733 /*
 734  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 735  * Electrostatics interaction: Ewald
 736  * VdW interaction:            LennardJones
 737  * Geometry:                   Water3-Particle
 738  * Calculate force/pot:        Force
 739  */
 740 void
 741 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 742                     (t_nblist                    * gmx_restrict       nlist,
 743                      rvec                        * gmx_restrict          xx,
 744                      rvec                        * gmx_restrict          ff,
 745                      t_forcerec                  * gmx_restrict          fr,
 746                      t_mdatoms                   * gmx_restrict     mdatoms,
 747                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 748                      t_nrnb                      * gmx_restrict        nrnb)
 749 {
 750     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 751      * just 0 for non-waters.
 752      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 753      * jnr indices corresponding to data put in the four positions in the SIMD register.
 754      */
 755     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 756     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 757     int              jnrA,jnrB;
 758     int              j_coord_offsetA,j_coord_offsetB;
 759     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 760     real             rcutoff_scalar;
 761     real             *shiftvec,*fshift,*x,*f;
 762     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 763     int              vdwioffset0;
 764     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 765     int              vdwioffset1;
 766     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 767     int              vdwioffset2;
 768     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 769     int              vdwjidx0A,vdwjidx0B;
 770     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 771     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 772     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 773     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 774     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 775     real             *charge;
 776     int              nvdwtype;
 777     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 778     int              *vdwtype;
 779     real             *vdwparam;
 780     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 781     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 782     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 783     real             *ewtab;
 784     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 785     real             rswitch_scalar,d_scalar;
 786     _fjsp_v2r8       itab_tmp;
 787     _fjsp_v2r8       dummy_mask,cutoff_mask;
 788     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 789     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 790     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 791
 792     x                = xx[0];
 793     f                = ff[0];
 794
 795     nri              = nlist->nri;
 796     iinr             = nlist->iinr;
 797     jindex           = nlist->jindex;
 798     jjnr             = nlist->jjnr;
 799     shiftidx         = nlist->shift;
 800     gid              = nlist->gid;
 801     shiftvec         = fr->shift_vec[0];
 802     fshift           = fr->fshift[0];
 803     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 804     charge           = mdatoms->chargeA;
 805     nvdwtype         = fr->ntype;
 806     vdwparam         = fr->nbfp;
 807     vdwtype          = mdatoms->typeA;
 808
 809     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 810     ewtab            = fr->ic->tabq_coul_FDV0;
 811     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 812     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 813
 814     /* Setup water-specific parameters */
 815     inr              = nlist->iinr[0];
 816     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 817     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 818     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 819     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 820
 821     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 822     rcutoff_scalar   = fr->rcoulomb;
 823     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 824     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 825
 826     rswitch_scalar   = fr->rcoulomb_switch;
 827     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 828     /* Setup switch parameters */
 829     d_scalar         = rcutoff_scalar-rswitch_scalar;
 830     d                = gmx_fjsp_set1_v2r8(d_scalar);
 831     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 832     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 833     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 834     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 835     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 836     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 837
 838     /* Avoid stupid compiler warnings */
 839     jnrA = jnrB = 0;
 840     j_coord_offsetA = 0;
 841     j_coord_offsetB = 0;
 842
 843     outeriter        = 0;
 844     inneriter        = 0;
 845
 846     /* Start outer loop over neighborlists */
 847     for(iidx=0; iidx<nri; iidx++)
 848     {
 849         /* Load shift vector for this list */
 850         i_shift_offset   = DIM*shiftidx[iidx];
 851
 852         /* Load limits for loop over neighbors */
 853         j_index_start    = jindex[iidx];
 854         j_index_end      = jindex[iidx+1];
 855
 856         /* Get outer coordinate index */
 857         inr              = iinr[iidx];
 858         i_coord_offset   = DIM*inr;
 859
 860         /* Load i particle coords and add shift vector */
 861         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 862                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 863
 864         fix0             = _fjsp_setzero_v2r8();
 865         fiy0             = _fjsp_setzero_v2r8();
 866         fiz0             = _fjsp_setzero_v2r8();
 867         fix1             = _fjsp_setzero_v2r8();
 868         fiy1             = _fjsp_setzero_v2r8();
 869         fiz1             = _fjsp_setzero_v2r8();
 870         fix2             = _fjsp_setzero_v2r8();
 871         fiy2             = _fjsp_setzero_v2r8();
 872         fiz2             = _fjsp_setzero_v2r8();
 873
 874         /* Start inner kernel loop */
 875         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 876         {
 877
 878             /* Get j neighbor index, and coordinate index */
 879             jnrA             = jjnr[jidx];
 880             jnrB             = jjnr[jidx+1];
 881             j_coord_offsetA  = DIM*jnrA;
 882             j_coord_offsetB  = DIM*jnrB;
 883
 884             /* load j atom coordinates */
 885             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 886                                               &jx0,&jy0,&jz0);
 887
 888             /* Calculate displacement vector */
 889             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 890             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 891             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 892             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 893             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 894             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 895             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 896             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 897             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 898
 899             /* Calculate squared distance and things based on it */
 900             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 901             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 902             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 903
 904             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 905             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 906             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 907
 908             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 909             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 910             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 911
 912             /* Load parameters for j particles */
 913             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 914             vdwjidx0A        = 2*vdwtype[jnrA+0];
 915             vdwjidx0B        = 2*vdwtype[jnrB+0];
 916
 917             fjx0             = _fjsp_setzero_v2r8();
 918             fjy0             = _fjsp_setzero_v2r8();
 919             fjz0             = _fjsp_setzero_v2r8();
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 926             {
 927
 928             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 929
 930             /* Compute parameters for interactions between i and j atoms */
 931             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 932             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 933                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 934
 935             /* EWALD ELECTROSTATICS */
 936
 937             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 938             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 939             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 940             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 941             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 942
 943             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 944             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 945             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 946             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 947             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 948             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 949             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 950             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 951             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 952             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 953
 954             /* LENNARD-JONES DISPERSION/REPULSION */
 955
 956             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 957             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 958             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 959             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 960             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 961
 962             d                = _fjsp_sub_v2r8(r00,rswitch);
 963             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 964             d2               = _fjsp_mul_v2r8(d,d);
 965             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 966
 967             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 968
 969             /* Evaluate switch function */
 970             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 971             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 972             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 973             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 974
 975             fscal            = _fjsp_add_v2r8(felec,fvdw);
 976
 977             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 978
 979             /* Update vectorial force */
 980             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 981             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 982             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 983
 984             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 985             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 986             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 987
 988             }
 989
 990             /**************************
 991              * CALCULATE INTERACTIONS *
 992              **************************/
 993
 994             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 995             {
 996
 997             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 998
 999             /* Compute parameters for interactions between i and j atoms */
1000             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1001
1002             /* EWALD ELECTROSTATICS */
1003
1004             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1005             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1006             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1007             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1008             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1009
1010             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1011             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1012             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1013             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1014             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1015             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1016             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1017             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1018             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1019             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1020
1021             d                = _fjsp_sub_v2r8(r10,rswitch);
1022             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1023             d2               = _fjsp_mul_v2r8(d,d);
1024             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1025
1026             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1027
1028             /* Evaluate switch function */
1029             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1030             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1031             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1032
1033             fscal            = felec;
1034
1035             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1036
1037             /* Update vectorial force */
1038             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1039             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1040             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1041
1042             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1043             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1044             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1045
1046             }
1047
1048             /**************************
1049              * CALCULATE INTERACTIONS *
1050              **************************/
1051
1052             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1053             {
1054
1055             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1056
1057             /* Compute parameters for interactions between i and j atoms */
1058             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1059
1060             /* EWALD ELECTROSTATICS */
1061
1062             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1063             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1064             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1065             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1066             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1067
1068             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1069             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1070             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1071             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1072             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1073             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1074             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1075             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1076             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1077             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1078
1079             d                = _fjsp_sub_v2r8(r20,rswitch);
1080             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1081             d2               = _fjsp_mul_v2r8(d,d);
1082             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1083
1084             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1085
1086             /* Evaluate switch function */
1087             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1088             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1089             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1090
1091             fscal            = felec;
1092
1093             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1094
1095             /* Update vectorial force */
1096             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1097             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1098             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1099
1100             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1101             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1102             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1103
1104             }
1105
1106             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1107
1108             /* Inner loop uses 213 flops */
1109         }
1110
1111         if(jidx<j_index_end)
1112         {
1113
1114             jnrA             = jjnr[jidx];
1115             j_coord_offsetA  = DIM*jnrA;
1116
1117             /* load j atom coordinates */
1118             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1119                                               &jx0,&jy0,&jz0);
1120
1121             /* Calculate displacement vector */
1122             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1123             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1124             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1125             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1126             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1127             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1128             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1129             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1130             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1131
1132             /* Calculate squared distance and things based on it */
1133             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1134             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1135             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1136
1137             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1138             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1139             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1140
1141             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1142             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1143             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1144
1145             /* Load parameters for j particles */
1146             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1147             vdwjidx0A        = 2*vdwtype[jnrA+0];
1148
1149             fjx0             = _fjsp_setzero_v2r8();
1150             fjy0             = _fjsp_setzero_v2r8();
1151             fjz0             = _fjsp_setzero_v2r8();
1152
1153             /**************************
1154              * CALCULATE INTERACTIONS *
1155              **************************/
1156
1157             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1158             {
1159
1160             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1161
1162             /* Compute parameters for interactions between i and j atoms */
1163             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1164             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1165
1166             /* EWALD ELECTROSTATICS */
1167
1168             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1169             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1170             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1171             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1172             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1173
1174             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1175             ewtabD           = _fjsp_setzero_v2r8();
1176             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1177             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1178             ewtabFn          = _fjsp_setzero_v2r8();
1179             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1180             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1181             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1182             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
1183             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1184
1185             /* LENNARD-JONES DISPERSION/REPULSION */
1186
1187             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1188             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1189             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1190             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1191             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1192
1193             d                = _fjsp_sub_v2r8(r00,rswitch);
1194             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1195             d2               = _fjsp_mul_v2r8(d,d);
1196             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1197
1198             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1199
1200             /* Evaluate switch function */
1201             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1202             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
1203             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1204             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1205
1206             fscal            = _fjsp_add_v2r8(felec,fvdw);
1207
1208             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1209
1210             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1211
1212             /* Update vectorial force */
1213             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1214             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1215             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1216
1217             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1218             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1219             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1220
1221             }
1222
1223             /**************************
1224              * CALCULATE INTERACTIONS *
1225              **************************/
1226
1227             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1228             {
1229
1230             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1231
1232             /* Compute parameters for interactions between i and j atoms */
1233             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1234
1235             /* EWALD ELECTROSTATICS */
1236
1237             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1238             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1239             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1240             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1241             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1242
1243             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1244             ewtabD           = _fjsp_setzero_v2r8();
1245             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1246             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1247             ewtabFn          = _fjsp_setzero_v2r8();
1248             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1249             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1250             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1251             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1252             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1253
1254             d                = _fjsp_sub_v2r8(r10,rswitch);
1255             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1256             d2               = _fjsp_mul_v2r8(d,d);
1257             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1258
1259             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1260
1261             /* Evaluate switch function */
1262             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1263             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1264             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1265
1266             fscal            = felec;
1267
1268             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1269
1270             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1271
1272             /* Update vectorial force */
1273             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1274             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1275             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1276
1277             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1278             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1279             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1280
1281             }
1282
1283             /**************************
1284              * CALCULATE INTERACTIONS *
1285              **************************/
1286
1287             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1288             {
1289
1290             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1291
1292             /* Compute parameters for interactions between i and j atoms */
1293             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1294
1295             /* EWALD ELECTROSTATICS */
1296
1297             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1298             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1299             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1300             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1301             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1302
1303             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1304             ewtabD           = _fjsp_setzero_v2r8();
1305             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1306             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1307             ewtabFn          = _fjsp_setzero_v2r8();
1308             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1309             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1310             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1311             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1312             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1313
1314             d                = _fjsp_sub_v2r8(r20,rswitch);
1315             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1316             d2               = _fjsp_mul_v2r8(d,d);
1317             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1318
1319             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1320
1321             /* Evaluate switch function */
1322             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1323             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1324             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1325
1326             fscal            = felec;
1327
1328             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1329
1330             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1331
1332             /* Update vectorial force */
1333             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1334             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1335             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1336
1337             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1338             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1339             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1340
1341             }
1342
1343             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1344
1345             /* Inner loop uses 213 flops */
1346         }
1347
1348         /* End of innermost loop */
1349
1350         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1351                                               f+i_coord_offset,fshift+i_shift_offset);
1352
1353         /* Increment number of inner iterations */
1354         inneriter                  += j_index_end - j_index_start;
1355
1356         /* Outer loop uses 18 flops */
1357     }
1358
1359     /* Increment number of outer iterations */
1360     outeriter        += nri;
1361
1362     /* Update outer/inner flops */
1363
1364     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*213);
1365 }