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) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  99     real             *ewtab;
 100     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 101     real             rswitch_scalar,d_scalar;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 126     ewtab            = fr->ic->tabq_coul_FDV0;
 127     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 128     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 129
 130     /* Setup water-specific parameters */
 131     inr              = nlist->iinr[0];
 132     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 136
 137     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 138     rcutoff_scalar   = fr->ic->rcoulomb;
 139     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 140     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 141
 142     rswitch_scalar   = fr->ic->rcoulomb_switch;
 143     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 144     /* Setup switch parameters */
 145     d_scalar         = rcutoff_scalar-rswitch_scalar;
 146     d                = gmx_fjsp_set1_v2r8(d_scalar);
 147     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 148     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 149     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 150     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 151     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 152     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158
 159     outeriter        = 0;
 160     inneriter        = 0;
 161
 162     /* Start outer loop over neighborlists */
 163     for(iidx=0; iidx<nri; iidx++)
 164     {
 165         /* Load shift vector for this list */
 166         i_shift_offset   = DIM*shiftidx[iidx];
 167
 168         /* Load limits for loop over neighbors */
 169         j_index_start    = jindex[iidx];
 170         j_index_end      = jindex[iidx+1];
 171
 172         /* Get outer coordinate index */
 173         inr              = iinr[iidx];
 174         i_coord_offset   = DIM*inr;
 175
 176         /* Load i particle coords and add shift vector */
 177         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 178                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 179
 180         fix0             = _fjsp_setzero_v2r8();
 181         fiy0             = _fjsp_setzero_v2r8();
 182         fiz0             = _fjsp_setzero_v2r8();
 183         fix1             = _fjsp_setzero_v2r8();
 184         fiy1             = _fjsp_setzero_v2r8();
 185         fiz1             = _fjsp_setzero_v2r8();
 186         fix2             = _fjsp_setzero_v2r8();
 187         fiy2             = _fjsp_setzero_v2r8();
 188         fiz2             = _fjsp_setzero_v2r8();
 189
 190         /* Reset potential sums */
 191         velecsum         = _fjsp_setzero_v2r8();
 192         vvdwsum          = _fjsp_setzero_v2r8();
 193
 194         /* Start inner kernel loop */
 195         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 196         {
 197
 198             /* Get j neighbor index, and coordinate index */
 199             jnrA             = jjnr[jidx];
 200             jnrB             = jjnr[jidx+1];
 201             j_coord_offsetA  = DIM*jnrA;
 202             j_coord_offsetB  = DIM*jnrB;
 203
 204             /* load j atom coordinates */
 205             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 206                                               &jx0,&jy0,&jz0);
 207
 208             /* Calculate displacement vector */
 209             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 210             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 211             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 212             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 213             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 214             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 215             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 216             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 217             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 218
 219             /* Calculate squared distance and things based on it */
 220             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 221             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 222             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 223
 224             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 225             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 226             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 227
 228             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 229             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 230             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 231
 232             /* Load parameters for j particles */
 233             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 234             vdwjidx0A        = 2*vdwtype[jnrA+0];
 235             vdwjidx0B        = 2*vdwtype[jnrB+0];
 236
 237             fjx0             = _fjsp_setzero_v2r8();
 238             fjy0             = _fjsp_setzero_v2r8();
 239             fjz0             = _fjsp_setzero_v2r8();
 240
 241             /**************************
 242              * CALCULATE INTERACTIONS *
 243              **************************/
 244
 245             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 246             {
 247
 248             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 249
 250             /* Compute parameters for interactions between i and j atoms */
 251             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 252             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 253                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 254
 255             /* EWALD ELECTROSTATICS */
 256
 257             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 258             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 259             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 260             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 261             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 262
 263             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 264             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 265             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 266             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 267             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 268             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 269             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 270             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 271             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 272             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 273
 274             /* LENNARD-JONES DISPERSION/REPULSION */
 275
 276             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 277             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 278             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 279             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 280             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 281
 282             d                = _fjsp_sub_v2r8(r00,rswitch);
 283             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 284             d2               = _fjsp_mul_v2r8(d,d);
 285             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 286
 287             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 288
 289             /* Evaluate switch function */
 290             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 291             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 292             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 293             velec            = _fjsp_mul_v2r8(velec,sw);
 294             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 295             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 296
 297             /* Update potential sum for this i atom from the interaction with this j atom. */
 298             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 299             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 300             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 301             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 302
 303             fscal            = _fjsp_add_v2r8(felec,fvdw);
 304
 305             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 306
 307             /* Update vectorial force */
 308             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 309             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 310             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 311
 312             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 313             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 314             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 315
 316             }
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 323             {
 324
 325             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 329
 330             /* EWALD ELECTROSTATICS */
 331
 332             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 333             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 334             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 335             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 336             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 337
 338             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 339             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 340             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 341             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 342             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 343             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 344             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 345             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 346             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 347             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 348
 349             d                = _fjsp_sub_v2r8(r10,rswitch);
 350             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 351             d2               = _fjsp_mul_v2r8(d,d);
 352             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 353
 354             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 355
 356             /* Evaluate switch function */
 357             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 358             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 359             velec            = _fjsp_mul_v2r8(velec,sw);
 360             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 361
 362             /* Update potential sum for this i atom from the interaction with this j atom. */
 363             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 364             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 369
 370             /* Update vectorial force */
 371             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 372             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 373             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 374
 375             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 376             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 377             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 378
 379             }
 380
 381             /**************************
 382              * CALCULATE INTERACTIONS *
 383              **************************/
 384
 385             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 386             {
 387
 388             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 389
 390             /* Compute parameters for interactions between i and j atoms */
 391             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 392
 393             /* EWALD ELECTROSTATICS */
 394
 395             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 396             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 397             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 398             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 399             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 400
 401             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 402             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 403             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 404             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 405             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 406             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 407             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 408             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 409             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 410             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 411
 412             d                = _fjsp_sub_v2r8(r20,rswitch);
 413             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 414             d2               = _fjsp_mul_v2r8(d,d);
 415             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 416
 417             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 418
 419             /* Evaluate switch function */
 420             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 421             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 422             velec            = _fjsp_mul_v2r8(velec,sw);
 423             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 424
 425             /* Update potential sum for this i atom from the interaction with this j atom. */
 426             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 427             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 428
 429             fscal            = felec;
 430
 431             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 432
 433             /* Update vectorial force */
 434             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 435             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 436             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 437
 438             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 439             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 440             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 441
 442             }
 443
 444             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 445
 446             /* Inner loop uses 225 flops */
 447         }
 448
 449         if(jidx<j_index_end)
 450         {
 451
 452             jnrA             = jjnr[jidx];
 453             j_coord_offsetA  = DIM*jnrA;
 454
 455             /* load j atom coordinates */
 456             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 457                                               &jx0,&jy0,&jz0);
 458
 459             /* Calculate displacement vector */
 460             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 461             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 462             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 463             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 464             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 465             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 466             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 467             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 468             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 469
 470             /* Calculate squared distance and things based on it */
 471             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 472             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 473             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 474
 475             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 476             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 477             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 478
 479             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 480             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 481             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 482
 483             /* Load parameters for j particles */
 484             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 485             vdwjidx0A        = 2*vdwtype[jnrA+0];
 486
 487             fjx0             = _fjsp_setzero_v2r8();
 488             fjy0             = _fjsp_setzero_v2r8();
 489             fjz0             = _fjsp_setzero_v2r8();
 490
 491             /**************************
 492              * CALCULATE INTERACTIONS *
 493              **************************/
 494
 495             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 496             {
 497
 498             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 499
 500             /* Compute parameters for interactions between i and j atoms */
 501             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 502             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 503                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 504
 505             /* EWALD ELECTROSTATICS */
 506
 507             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 508             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 509             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 510             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 511             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 512
 513             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 514             ewtabD           = _fjsp_setzero_v2r8();
 515             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 516             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 517             ewtabFn          = _fjsp_setzero_v2r8();
 518             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 519             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 520             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 521             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 522             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 523
 524             /* LENNARD-JONES DISPERSION/REPULSION */
 525
 526             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 527             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 528             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 529             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 530             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 531
 532             d                = _fjsp_sub_v2r8(r00,rswitch);
 533             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 534             d2               = _fjsp_mul_v2r8(d,d);
 535             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 536
 537             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 538
 539             /* Evaluate switch function */
 540             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 541             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 542             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 543             velec            = _fjsp_mul_v2r8(velec,sw);
 544             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 545             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 549             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 550             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 551             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 552             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 553             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 554
 555             fscal            = _fjsp_add_v2r8(felec,fvdw);
 556
 557             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 558
 559             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 560
 561             /* Update vectorial force */
 562             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 563             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 564             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 565
 566             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 567             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 568             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 569
 570             }
 571
 572             /**************************
 573              * CALCULATE INTERACTIONS *
 574              **************************/
 575
 576             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 577             {
 578
 579             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 580
 581             /* Compute parameters for interactions between i and j atoms */
 582             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 583
 584             /* EWALD ELECTROSTATICS */
 585
 586             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 587             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 588             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 589             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 590             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 591
 592             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 593             ewtabD           = _fjsp_setzero_v2r8();
 594             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 595             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 596             ewtabFn          = _fjsp_setzero_v2r8();
 597             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 598             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 599             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 600             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 601             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 602
 603             d                = _fjsp_sub_v2r8(r10,rswitch);
 604             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 605             d2               = _fjsp_mul_v2r8(d,d);
 606             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 607
 608             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 609
 610             /* Evaluate switch function */
 611             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 612             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 613             velec            = _fjsp_mul_v2r8(velec,sw);
 614             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 615
 616             /* Update potential sum for this i atom from the interaction with this j atom. */
 617             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 618             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 619             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 620
 621             fscal            = felec;
 622
 623             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 624
 625             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 626
 627             /* Update vectorial force */
 628             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 629             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 630             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 631
 632             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 633             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 634             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 635
 636             }
 637
 638             /**************************
 639              * CALCULATE INTERACTIONS *
 640              **************************/
 641
 642             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 643             {
 644
 645             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 646
 647             /* Compute parameters for interactions between i and j atoms */
 648             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 649
 650             /* EWALD ELECTROSTATICS */
 651
 652             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 653             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 654             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 655             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 656             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 657
 658             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 659             ewtabD           = _fjsp_setzero_v2r8();
 660             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 661             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 662             ewtabFn          = _fjsp_setzero_v2r8();
 663             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 664             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 665             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 666             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 667             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 668
 669             d                = _fjsp_sub_v2r8(r20,rswitch);
 670             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 671             d2               = _fjsp_mul_v2r8(d,d);
 672             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 673
 674             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 675
 676             /* Evaluate switch function */
 677             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 678             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 679             velec            = _fjsp_mul_v2r8(velec,sw);
 680             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 681
 682             /* Update potential sum for this i atom from the interaction with this j atom. */
 683             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 684             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 685             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 686
 687             fscal            = felec;
 688
 689             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 690
 691             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 692
 693             /* Update vectorial force */
 694             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 695             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 696             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 697
 698             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 699             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 700             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 701
 702             }
 703
 704             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 705
 706             /* Inner loop uses 225 flops */
 707         }
 708
 709         /* End of innermost loop */
 710
 711         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 712                                               f+i_coord_offset,fshift+i_shift_offset);
 713
 714         ggid                        = gid[iidx];
 715         /* Update potential energies */
 716         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 717         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 718
 719         /* Increment number of inner iterations */
 720         inneriter                  += j_index_end - j_index_start;
 721
 722         /* Outer loop uses 20 flops */
 723     }
 724
 725     /* Increment number of outer iterations */
 726     outeriter        += nri;
 727
 728     /* Update outer/inner flops */
 729
 730     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*225);
 731 }
 732 /*
 733  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 734  * Electrostatics interaction: Ewald
 735  * VdW interaction:            LennardJones
 736  * Geometry:                   Water3-Particle
 737  * Calculate force/pot:        Force
 738  */
 739 void
 740 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 741                     (t_nblist                    * gmx_restrict       nlist,
 742                      rvec                        * gmx_restrict          xx,
 743                      rvec                        * gmx_restrict          ff,
 744                      struct t_forcerec           * gmx_restrict          fr,
 745                      t_mdatoms                   * gmx_restrict     mdatoms,
 746                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 747                      t_nrnb                      * gmx_restrict        nrnb)
 748 {
 749     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 750      * just 0 for non-waters.
 751      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 752      * jnr indices corresponding to data put in the four positions in the SIMD register.
 753      */
 754     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 755     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 756     int              jnrA,jnrB;
 757     int              j_coord_offsetA,j_coord_offsetB;
 758     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 759     real             rcutoff_scalar;
 760     real             *shiftvec,*fshift,*x,*f;
 761     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 762     int              vdwioffset0;
 763     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 764     int              vdwioffset1;
 765     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 766     int              vdwioffset2;
 767     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 768     int              vdwjidx0A,vdwjidx0B;
 769     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 770     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 771     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 772     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 773     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 774     real             *charge;
 775     int              nvdwtype;
 776     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 777     int              *vdwtype;
 778     real             *vdwparam;
 779     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 780     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 781     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 782     real             *ewtab;
 783     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 784     real             rswitch_scalar,d_scalar;
 785     _fjsp_v2r8       itab_tmp;
 786     _fjsp_v2r8       dummy_mask,cutoff_mask;
 787     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 788     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 789     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 790
 791     x                = xx[0];
 792     f                = ff[0];
 793
 794     nri              = nlist->nri;
 795     iinr             = nlist->iinr;
 796     jindex           = nlist->jindex;
 797     jjnr             = nlist->jjnr;
 798     shiftidx         = nlist->shift;
 799     gid              = nlist->gid;
 800     shiftvec         = fr->shift_vec[0];
 801     fshift           = fr->fshift[0];
 802     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 803     charge           = mdatoms->chargeA;
 804     nvdwtype         = fr->ntype;
 805     vdwparam         = fr->nbfp;
 806     vdwtype          = mdatoms->typeA;
 807
 808     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 809     ewtab            = fr->ic->tabq_coul_FDV0;
 810     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 811     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 812
 813     /* Setup water-specific parameters */
 814     inr              = nlist->iinr[0];
 815     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 816     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 817     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 818     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 819
 820     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 821     rcutoff_scalar   = fr->ic->rcoulomb;
 822     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 823     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 824
 825     rswitch_scalar   = fr->ic->rcoulomb_switch;
 826     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 827     /* Setup switch parameters */
 828     d_scalar         = rcutoff_scalar-rswitch_scalar;
 829     d                = gmx_fjsp_set1_v2r8(d_scalar);
 830     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 831     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 832     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 833     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 834     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 835     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 836
 837     /* Avoid stupid compiler warnings */
 838     jnrA = jnrB = 0;
 839     j_coord_offsetA = 0;
 840     j_coord_offsetB = 0;
 841
 842     outeriter        = 0;
 843     inneriter        = 0;
 844
 845     /* Start outer loop over neighborlists */
 846     for(iidx=0; iidx<nri; iidx++)
 847     {
 848         /* Load shift vector for this list */
 849         i_shift_offset   = DIM*shiftidx[iidx];
 850
 851         /* Load limits for loop over neighbors */
 852         j_index_start    = jindex[iidx];
 853         j_index_end      = jindex[iidx+1];
 854
 855         /* Get outer coordinate index */
 856         inr              = iinr[iidx];
 857         i_coord_offset   = DIM*inr;
 858
 859         /* Load i particle coords and add shift vector */
 860         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 861                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 862
 863         fix0             = _fjsp_setzero_v2r8();
 864         fiy0             = _fjsp_setzero_v2r8();
 865         fiz0             = _fjsp_setzero_v2r8();
 866         fix1             = _fjsp_setzero_v2r8();
 867         fiy1             = _fjsp_setzero_v2r8();
 868         fiz1             = _fjsp_setzero_v2r8();
 869         fix2             = _fjsp_setzero_v2r8();
 870         fiy2             = _fjsp_setzero_v2r8();
 871         fiz2             = _fjsp_setzero_v2r8();
 872
 873         /* Start inner kernel loop */
 874         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 875         {
 876
 877             /* Get j neighbor index, and coordinate index */
 878             jnrA             = jjnr[jidx];
 879             jnrB             = jjnr[jidx+1];
 880             j_coord_offsetA  = DIM*jnrA;
 881             j_coord_offsetB  = DIM*jnrB;
 882
 883             /* load j atom coordinates */
 884             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 885                                               &jx0,&jy0,&jz0);
 886
 887             /* Calculate displacement vector */
 888             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 889             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 890             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 891             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 892             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 893             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 894             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 895             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 896             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 897
 898             /* Calculate squared distance and things based on it */
 899             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 900             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 901             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 902
 903             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 904             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 905             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 906
 907             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 908             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 909             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 910
 911             /* Load parameters for j particles */
 912             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 913             vdwjidx0A        = 2*vdwtype[jnrA+0];
 914             vdwjidx0B        = 2*vdwtype[jnrB+0];
 915
 916             fjx0             = _fjsp_setzero_v2r8();
 917             fjy0             = _fjsp_setzero_v2r8();
 918             fjz0             = _fjsp_setzero_v2r8();
 919
 920             /**************************
 921              * CALCULATE INTERACTIONS *
 922              **************************/
 923
 924             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 925             {
 926
 927             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 928
 929             /* Compute parameters for interactions between i and j atoms */
 930             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 931             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 932                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 933
 934             /* EWALD ELECTROSTATICS */
 935
 936             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 937             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 938             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 939             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 940             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 941
 942             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 943             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 944             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 945             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 946             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 947             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 948             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 949             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 950             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 951             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 952
 953             /* LENNARD-JONES DISPERSION/REPULSION */
 954
 955             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 956             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 957             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 958             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 959             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 960
 961             d                = _fjsp_sub_v2r8(r00,rswitch);
 962             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 963             d2               = _fjsp_mul_v2r8(d,d);
 964             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 965
 966             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 967
 968             /* Evaluate switch function */
 969             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 970             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 971             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 972             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 973
 974             fscal            = _fjsp_add_v2r8(felec,fvdw);
 975
 976             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 977
 978             /* Update vectorial force */
 979             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 980             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 981             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 982
 983             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 984             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 985             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 986
 987             }
 988
 989             /**************************
 990              * CALCULATE INTERACTIONS *
 991              **************************/
 992
 993             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 994             {
 995
 996             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 997
 998             /* Compute parameters for interactions between i and j atoms */
 999             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1000
1001             /* EWALD ELECTROSTATICS */
1002
1003             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1004             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1005             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1006             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1007             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1008
1009             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1010             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1011             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1012             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1013             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1014             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1015             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1016             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1017             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1018             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1019
1020             d                = _fjsp_sub_v2r8(r10,rswitch);
1021             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1022             d2               = _fjsp_mul_v2r8(d,d);
1023             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1024
1025             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1026
1027             /* Evaluate switch function */
1028             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1029             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1030             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1031
1032             fscal            = felec;
1033
1034             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1035
1036             /* Update vectorial force */
1037             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1038             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1039             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1040
1041             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1042             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1043             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1044
1045             }
1046
1047             /**************************
1048              * CALCULATE INTERACTIONS *
1049              **************************/
1050
1051             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1052             {
1053
1054             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1055
1056             /* Compute parameters for interactions between i and j atoms */
1057             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1058
1059             /* EWALD ELECTROSTATICS */
1060
1061             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1062             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1063             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1064             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1065             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1066
1067             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1068             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1069             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1070             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1071             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1072             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1073             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1074             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1075             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1076             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1077
1078             d                = _fjsp_sub_v2r8(r20,rswitch);
1079             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1080             d2               = _fjsp_mul_v2r8(d,d);
1081             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1082
1083             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1084
1085             /* Evaluate switch function */
1086             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1087             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1088             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1089
1090             fscal            = felec;
1091
1092             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1093
1094             /* Update vectorial force */
1095             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1096             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1097             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1098
1099             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1100             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1101             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1102
1103             }
1104
1105             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1106
1107             /* Inner loop uses 213 flops */
1108         }
1109
1110         if(jidx<j_index_end)
1111         {
1112
1113             jnrA             = jjnr[jidx];
1114             j_coord_offsetA  = DIM*jnrA;
1115
1116             /* load j atom coordinates */
1117             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1118                                               &jx0,&jy0,&jz0);
1119
1120             /* Calculate displacement vector */
1121             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1122             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1123             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1124             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1125             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1126             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1127             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1128             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1129             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1130
1131             /* Calculate squared distance and things based on it */
1132             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1133             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1134             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1135
1136             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1137             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1138             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1139
1140             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1141             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1142             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1143
1144             /* Load parameters for j particles */
1145             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1146             vdwjidx0A        = 2*vdwtype[jnrA+0];
1147
1148             fjx0             = _fjsp_setzero_v2r8();
1149             fjy0             = _fjsp_setzero_v2r8();
1150             fjz0             = _fjsp_setzero_v2r8();
1151
1152             /**************************
1153              * CALCULATE INTERACTIONS *
1154              **************************/
1155
1156             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1157             {
1158
1159             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1160
1161             /* Compute parameters for interactions between i and j atoms */
1162             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1163             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1164                                          vdwparam+vdwioffset0+vdwjidx0B,&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 }