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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 505                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 506
 507             /* EWALD ELECTROSTATICS */
 508
 509             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 510             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 511             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 512             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 513             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 514
 515             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 516             ewtabD           = _fjsp_setzero_v2r8();
 517             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 518             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 519             ewtabFn          = _fjsp_setzero_v2r8();
 520             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 521             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 522             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 523             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 524             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 525
 526             /* LENNARD-JONES DISPERSION/REPULSION */
 527
 528             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 529             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 530             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 531             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 532             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 533
 534             d                = _fjsp_sub_v2r8(r00,rswitch);
 535             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 536             d2               = _fjsp_mul_v2r8(d,d);
 537             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 538
 539             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 540
 541             /* Evaluate switch function */
 542             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 543             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 544             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 545             velec            = _fjsp_mul_v2r8(velec,sw);
 546             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 547             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 548
 549             /* Update potential sum for this i atom from the interaction with this j atom. */
 550             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 551             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 552             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 553             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 554             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 555             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 556
 557             fscal            = _fjsp_add_v2r8(felec,fvdw);
 558
 559             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 560
 561             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 562
 563             /* Update vectorial force */
 564             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 565             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 566             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 567
 568             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 569             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 570             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 571
 572             }
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 576              **************************/
 577
 578             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 579             {
 580
 581             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 582
 583             /* Compute parameters for interactions between i and j atoms */
 584             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 585
 586             /* EWALD ELECTROSTATICS */
 587
 588             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 589             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 590             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 591             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 592             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 593
 594             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 595             ewtabD           = _fjsp_setzero_v2r8();
 596             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 597             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 598             ewtabFn          = _fjsp_setzero_v2r8();
 599             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 600             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 601             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 602             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 603             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 604
 605             d                = _fjsp_sub_v2r8(r10,rswitch);
 606             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 607             d2               = _fjsp_mul_v2r8(d,d);
 608             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 609
 610             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 611
 612             /* Evaluate switch function */
 613             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 614             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 615             velec            = _fjsp_mul_v2r8(velec,sw);
 616             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 617
 618             /* Update potential sum for this i atom from the interaction with this j atom. */
 619             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 620             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 621             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 622
 623             fscal            = felec;
 624
 625             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 626
 627             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 628
 629             /* Update vectorial force */
 630             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 631             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 632             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 633
 634             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 635             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 636             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 637
 638             }
 639
 640             /**************************
 641              * CALCULATE INTERACTIONS *
 642              **************************/
 643
 644             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 645             {
 646
 647             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 648
 649             /* Compute parameters for interactions between i and j atoms */
 650             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 651
 652             /* EWALD ELECTROSTATICS */
 653
 654             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 655             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 656             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 657             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 658             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 659
 660             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 661             ewtabD           = _fjsp_setzero_v2r8();
 662             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 663             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 664             ewtabFn          = _fjsp_setzero_v2r8();
 665             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 666             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 667             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 668             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 669             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 670
 671             d                = _fjsp_sub_v2r8(r20,rswitch);
 672             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 673             d2               = _fjsp_mul_v2r8(d,d);
 674             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 675
 676             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 677
 678             /* Evaluate switch function */
 679             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 680             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 681             velec            = _fjsp_mul_v2r8(velec,sw);
 682             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 683
 684             /* Update potential sum for this i atom from the interaction with this j atom. */
 685             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 686             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 687             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 688
 689             fscal            = felec;
 690
 691             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 692
 693             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 694
 695             /* Update vectorial force */
 696             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 697             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 698             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 699
 700             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 701             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 702             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 703
 704             }
 705
 706             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 707
 708             /* Inner loop uses 225 flops */
 709         }
 710
 711         /* End of innermost loop */
 712
 713         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 714                                               f+i_coord_offset,fshift+i_shift_offset);
 715
 716         ggid                        = gid[iidx];
 717         /* Update potential energies */
 718         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 719         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 720
 721         /* Increment number of inner iterations */
 722         inneriter                  += j_index_end - j_index_start;
 723
 724         /* Outer loop uses 20 flops */
 725     }
 726
 727     /* Increment number of outer iterations */
 728     outeriter        += nri;
 729
 730     /* Update outer/inner flops */
 731
 732     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*225);
 733 }
 734 /*
 735  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 736  * Electrostatics interaction: Ewald
 737  * VdW interaction:            LennardJones
 738  * Geometry:                   Water3-Particle
 739  * Calculate force/pot:        Force
 740  */
 741 void
 742 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 743                     (t_nblist                    * gmx_restrict       nlist,
 744                      rvec                        * gmx_restrict          xx,
 745                      rvec                        * gmx_restrict          ff,
 746                      t_forcerec                  * gmx_restrict          fr,
 747                      t_mdatoms                   * gmx_restrict     mdatoms,
 748                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 749                      t_nrnb                      * gmx_restrict        nrnb)
 750 {
 751     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 752      * just 0 for non-waters.
 753      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 754      * jnr indices corresponding to data put in the four positions in the SIMD register.
 755      */
 756     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 757     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 758     int              jnrA,jnrB;
 759     int              j_coord_offsetA,j_coord_offsetB;
 760     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 761     real             rcutoff_scalar;
 762     real             *shiftvec,*fshift,*x,*f;
 763     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 764     int              vdwioffset0;
 765     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 766     int              vdwioffset1;
 767     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 768     int              vdwioffset2;
 769     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 770     int              vdwjidx0A,vdwjidx0B;
 771     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 772     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 773     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 774     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 775     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 776     real             *charge;
 777     int              nvdwtype;
 778     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 779     int              *vdwtype;
 780     real             *vdwparam;
 781     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 782     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 783     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 784     real             *ewtab;
 785     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 786     real             rswitch_scalar,d_scalar;
 787     _fjsp_v2r8       itab_tmp;
 788     _fjsp_v2r8       dummy_mask,cutoff_mask;
 789     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 790     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 791     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 792
 793     x                = xx[0];
 794     f                = ff[0];
 795
 796     nri              = nlist->nri;
 797     iinr             = nlist->iinr;
 798     jindex           = nlist->jindex;
 799     jjnr             = nlist->jjnr;
 800     shiftidx         = nlist->shift;
 801     gid              = nlist->gid;
 802     shiftvec         = fr->shift_vec[0];
 803     fshift           = fr->fshift[0];
 804     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 805     charge           = mdatoms->chargeA;
 806     nvdwtype         = fr->ntype;
 807     vdwparam         = fr->nbfp;
 808     vdwtype          = mdatoms->typeA;
 809
 810     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 811     ewtab            = fr->ic->tabq_coul_FDV0;
 812     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 813     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 814
 815     /* Setup water-specific parameters */
 816     inr              = nlist->iinr[0];
 817     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 818     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 819     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 820     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 821
 822     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 823     rcutoff_scalar   = fr->rcoulomb;
 824     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 825     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 826
 827     rswitch_scalar   = fr->rcoulomb_switch;
 828     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 829     /* Setup switch parameters */
 830     d_scalar         = rcutoff_scalar-rswitch_scalar;
 831     d                = gmx_fjsp_set1_v2r8(d_scalar);
 832     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 833     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 834     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 835     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 836     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 837     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 838
 839     /* Avoid stupid compiler warnings */
 840     jnrA = jnrB = 0;
 841     j_coord_offsetA = 0;
 842     j_coord_offsetB = 0;
 843
 844     outeriter        = 0;
 845     inneriter        = 0;
 846
 847     /* Start outer loop over neighborlists */
 848     for(iidx=0; iidx<nri; iidx++)
 849     {
 850         /* Load shift vector for this list */
 851         i_shift_offset   = DIM*shiftidx[iidx];
 852
 853         /* Load limits for loop over neighbors */
 854         j_index_start    = jindex[iidx];
 855         j_index_end      = jindex[iidx+1];
 856
 857         /* Get outer coordinate index */
 858         inr              = iinr[iidx];
 859         i_coord_offset   = DIM*inr;
 860
 861         /* Load i particle coords and add shift vector */
 862         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 863                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 864
 865         fix0             = _fjsp_setzero_v2r8();
 866         fiy0             = _fjsp_setzero_v2r8();
 867         fiz0             = _fjsp_setzero_v2r8();
 868         fix1             = _fjsp_setzero_v2r8();
 869         fiy1             = _fjsp_setzero_v2r8();
 870         fiz1             = _fjsp_setzero_v2r8();
 871         fix2             = _fjsp_setzero_v2r8();
 872         fiy2             = _fjsp_setzero_v2r8();
 873         fiz2             = _fjsp_setzero_v2r8();
 874
 875         /* Start inner kernel loop */
 876         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 877         {
 878
 879             /* Get j neighbor index, and coordinate index */
 880             jnrA             = jjnr[jidx];
 881             jnrB             = jjnr[jidx+1];
 882             j_coord_offsetA  = DIM*jnrA;
 883             j_coord_offsetB  = DIM*jnrB;
 884
 885             /* load j atom coordinates */
 886             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 887                                               &jx0,&jy0,&jz0);
 888
 889             /* Calculate displacement vector */
 890             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 891             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 892             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 893             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 894             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 895             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 896             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 897             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 898             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 899
 900             /* Calculate squared distance and things based on it */
 901             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 902             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 903             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 904
 905             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 906             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 907             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 908
 909             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 910             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 911             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 912
 913             /* Load parameters for j particles */
 914             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 915             vdwjidx0A        = 2*vdwtype[jnrA+0];
 916             vdwjidx0B        = 2*vdwtype[jnrB+0];
 917
 918             fjx0             = _fjsp_setzero_v2r8();
 919             fjy0             = _fjsp_setzero_v2r8();
 920             fjz0             = _fjsp_setzero_v2r8();
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 927             {
 928
 929             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 930
 931             /* Compute parameters for interactions between i and j atoms */
 932             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 933             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 934                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 935
 936             /* EWALD ELECTROSTATICS */
 937
 938             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 939             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 940             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 941             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 942             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 943
 944             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 945             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 946             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 947             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 948             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 949             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 950             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 951             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 952             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 953             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 954
 955             /* LENNARD-JONES DISPERSION/REPULSION */
 956
 957             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 958             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 959             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 960             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 961             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 962
 963             d                = _fjsp_sub_v2r8(r00,rswitch);
 964             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 965             d2               = _fjsp_mul_v2r8(d,d);
 966             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 967
 968             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 969
 970             /* Evaluate switch function */
 971             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 972             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 973             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 974             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 975
 976             fscal            = _fjsp_add_v2r8(felec,fvdw);
 977
 978             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 979
 980             /* Update vectorial force */
 981             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 982             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 983             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 984
 985             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 986             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 987             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 988
 989             }
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 996             {
 997
 998             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 999
1000             /* Compute parameters for interactions between i and j atoms */
1001             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1002
1003             /* EWALD ELECTROSTATICS */
1004
1005             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1006             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1007             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1008             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1009             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1010
1011             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1012             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1013             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1014             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1015             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1016             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1017             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1018             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1019             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1020             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1021
1022             d                = _fjsp_sub_v2r8(r10,rswitch);
1023             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1024             d2               = _fjsp_mul_v2r8(d,d);
1025             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1026
1027             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1028
1029             /* Evaluate switch function */
1030             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1031             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1032             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1033
1034             fscal            = felec;
1035
1036             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1037
1038             /* Update vectorial force */
1039             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1040             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1041             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1042
1043             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1044             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1045             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1046
1047             }
1048
1049             /**************************
1050              * CALCULATE INTERACTIONS *
1051              **************************/
1052
1053             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1054             {
1055
1056             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1057
1058             /* Compute parameters for interactions between i and j atoms */
1059             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1060
1061             /* EWALD ELECTROSTATICS */
1062
1063             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1064             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1065             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1066             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1067             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1068
1069             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1070             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1071             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1072             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1073             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1074             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1075             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1076             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1077             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1078             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1079
1080             d                = _fjsp_sub_v2r8(r20,rswitch);
1081             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1082             d2               = _fjsp_mul_v2r8(d,d);
1083             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1084
1085             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1086
1087             /* Evaluate switch function */
1088             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1089             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1090             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1091
1092             fscal            = felec;
1093
1094             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1095
1096             /* Update vectorial force */
1097             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1098             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1099             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1100
1101             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1102             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1103             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1104
1105             }
1106
1107             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1108
1109             /* Inner loop uses 213 flops */
1110         }
1111
1112         if(jidx<j_index_end)
1113         {
1114
1115             jnrA             = jjnr[jidx];
1116             j_coord_offsetA  = DIM*jnrA;
1117
1118             /* load j atom coordinates */
1119             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1120                                               &jx0,&jy0,&jz0);
1121
1122             /* Calculate displacement vector */
1123             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1124             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1125             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1126             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1127             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1128             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1129             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1130             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1131             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1132
1133             /* Calculate squared distance and things based on it */
1134             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1135             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1136             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1137
1138             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1139             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1140             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1141
1142             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1143             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1144             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1145
1146             /* Load parameters for j particles */
1147             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1148             vdwjidx0A        = 2*vdwtype[jnrA+0];
1149
1150             fjx0             = _fjsp_setzero_v2r8();
1151             fjy0             = _fjsp_setzero_v2r8();
1152             fjz0             = _fjsp_setzero_v2r8();
1153
1154             /**************************
1155              * CALCULATE INTERACTIONS *
1156              **************************/
1157
1158             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1159             {
1160
1161             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1162
1163             /* Compute parameters for interactions between i and j atoms */
1164             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1165             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1166                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1167
1168             /* EWALD ELECTROSTATICS */
1169
1170             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1171             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1172             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1173             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1174             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1175
1176             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1177             ewtabD           = _fjsp_setzero_v2r8();
1178             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1179             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1180             ewtabFn          = _fjsp_setzero_v2r8();
1181             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1182             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1183             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1184             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
1185             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1186
1187             /* LENNARD-JONES DISPERSION/REPULSION */
1188
1189             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1190             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1191             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1192             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1193             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1194
1195             d                = _fjsp_sub_v2r8(r00,rswitch);
1196             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1197             d2               = _fjsp_mul_v2r8(d,d);
1198             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1199
1200             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1201
1202             /* Evaluate switch function */
1203             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1204             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
1205             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1206             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1207
1208             fscal            = _fjsp_add_v2r8(felec,fvdw);
1209
1210             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1211
1212             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1213
1214             /* Update vectorial force */
1215             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1216             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1217             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1218
1219             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1220             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1221             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1222
1223             }
1224
1225             /**************************
1226              * CALCULATE INTERACTIONS *
1227              **************************/
1228
1229             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1230             {
1231
1232             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1233
1234             /* Compute parameters for interactions between i and j atoms */
1235             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1236
1237             /* EWALD ELECTROSTATICS */
1238
1239             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1240             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1241             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1242             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1243             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1244
1245             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1246             ewtabD           = _fjsp_setzero_v2r8();
1247             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1248             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1249             ewtabFn          = _fjsp_setzero_v2r8();
1250             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1251             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1252             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1253             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1254             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1255
1256             d                = _fjsp_sub_v2r8(r10,rswitch);
1257             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1258             d2               = _fjsp_mul_v2r8(d,d);
1259             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1260
1261             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1262
1263             /* Evaluate switch function */
1264             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1265             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1266             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1267
1268             fscal            = felec;
1269
1270             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1271
1272             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1273
1274             /* Update vectorial force */
1275             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1276             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1277             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1278
1279             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1280             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1281             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1282
1283             }
1284
1285             /**************************
1286              * CALCULATE INTERACTIONS *
1287              **************************/
1288
1289             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1290             {
1291
1292             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1293
1294             /* Compute parameters for interactions between i and j atoms */
1295             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1296
1297             /* EWALD ELECTROSTATICS */
1298
1299             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1300             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1301             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1302             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1303             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1304
1305             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1306             ewtabD           = _fjsp_setzero_v2r8();
1307             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1308             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1309             ewtabFn          = _fjsp_setzero_v2r8();
1310             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1311             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1312             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1313             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1314             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1315
1316             d                = _fjsp_sub_v2r8(r20,rswitch);
1317             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1318             d2               = _fjsp_mul_v2r8(d,d);
1319             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1320
1321             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1322
1323             /* Evaluate switch function */
1324             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1325             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1326             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1327
1328             fscal            = felec;
1329
1330             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1331
1332             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1333
1334             /* Update vectorial force */
1335             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1336             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1337             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1338
1339             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1340             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1341             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1342
1343             }
1344
1345             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1346
1347             /* Inner loop uses 213 flops */
1348         }
1349
1350         /* End of innermost loop */
1351
1352         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1353                                               f+i_coord_offset,fshift+i_shift_offset);
1354
1355         /* Increment number of inner iterations */
1356         inneriter                  += j_index_end - j_index_start;
1357
1358         /* Outer loop uses 18 flops */
1359     }
1360
1361     /* Increment number of outer iterations */
1362     outeriter        += nri;
1363
1364     /* Update outer/inner flops */
1365
1366     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*213);
1367 }