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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "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                      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->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->rcoulomb;
 139     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 140     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 141
 142     rswitch_scalar   = fr->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_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 503
 504             /* EWALD ELECTROSTATICS */
 505
 506             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 507             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 508             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 509             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 510             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 511
 512             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 513             ewtabD           = _fjsp_setzero_v2r8();
 514             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 515             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 516             ewtabFn          = _fjsp_setzero_v2r8();
 517             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 518             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 519             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 520             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 521             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 522
 523             /* LENNARD-JONES DISPERSION/REPULSION */
 524
 525             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 526             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 527             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 528             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 529             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 530
 531             d                = _fjsp_sub_v2r8(r00,rswitch);
 532             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 533             d2               = _fjsp_mul_v2r8(d,d);
 534             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 535
 536             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 537
 538             /* Evaluate switch function */
 539             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 540             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 541             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 542             velec            = _fjsp_mul_v2r8(velec,sw);
 543             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 544             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 548             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 549             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 550             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 551             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 552             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 553
 554             fscal            = _fjsp_add_v2r8(felec,fvdw);
 555
 556             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 557
 558             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 559
 560             /* Update vectorial force */
 561             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 562             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 563             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 564
 565             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 566             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 567             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 568
 569             }
 570
 571             /**************************
 572              * CALCULATE INTERACTIONS *
 573              **************************/
 574
 575             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 576             {
 577
 578             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 579
 580             /* Compute parameters for interactions between i and j atoms */
 581             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 582
 583             /* EWALD ELECTROSTATICS */
 584
 585             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 586             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 587             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 588             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 589             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 590
 591             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 592             ewtabD           = _fjsp_setzero_v2r8();
 593             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 594             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 595             ewtabFn          = _fjsp_setzero_v2r8();
 596             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 597             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 598             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 599             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 600             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 601
 602             d                = _fjsp_sub_v2r8(r10,rswitch);
 603             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 604             d2               = _fjsp_mul_v2r8(d,d);
 605             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 606
 607             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 608
 609             /* Evaluate switch function */
 610             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 611             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 612             velec            = _fjsp_mul_v2r8(velec,sw);
 613             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 614
 615             /* Update potential sum for this i atom from the interaction with this j atom. */
 616             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 617             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 618             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 619
 620             fscal            = felec;
 621
 622             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 623
 624             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 625
 626             /* Update vectorial force */
 627             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 628             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 629             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 630
 631             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 632             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 633             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 634
 635             }
 636
 637             /**************************
 638              * CALCULATE INTERACTIONS *
 639              **************************/
 640
 641             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 642             {
 643
 644             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 645
 646             /* Compute parameters for interactions between i and j atoms */
 647             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 648
 649             /* EWALD ELECTROSTATICS */
 650
 651             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 652             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 653             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 654             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 655             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 656
 657             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 658             ewtabD           = _fjsp_setzero_v2r8();
 659             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 660             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 661             ewtabFn          = _fjsp_setzero_v2r8();
 662             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 663             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 664             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 665             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 666             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 667
 668             d                = _fjsp_sub_v2r8(r20,rswitch);
 669             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 670             d2               = _fjsp_mul_v2r8(d,d);
 671             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 672
 673             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 674
 675             /* Evaluate switch function */
 676             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 677             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 678             velec            = _fjsp_mul_v2r8(velec,sw);
 679             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 680
 681             /* Update potential sum for this i atom from the interaction with this j atom. */
 682             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 683             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 684             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 685
 686             fscal            = felec;
 687
 688             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 689
 690             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 691
 692             /* Update vectorial force */
 693             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 694             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 695             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 696
 697             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 698             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 699             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 700
 701             }
 702
 703             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 704
 705             /* Inner loop uses 225 flops */
 706         }
 707
 708         /* End of innermost loop */
 709
 710         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 711                                               f+i_coord_offset,fshift+i_shift_offset);
 712
 713         ggid                        = gid[iidx];
 714         /* Update potential energies */
 715         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 716         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 717
 718         /* Increment number of inner iterations */
 719         inneriter                  += j_index_end - j_index_start;
 720
 721         /* Outer loop uses 20 flops */
 722     }
 723
 724     /* Increment number of outer iterations */
 725     outeriter        += nri;
 726
 727     /* Update outer/inner flops */
 728
 729     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*225);
 730 }
 731 /*
 732  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 733  * Electrostatics interaction: Ewald
 734  * VdW interaction:            LennardJones
 735  * Geometry:                   Water3-Particle
 736  * Calculate force/pot:        Force
 737  */
 738 void
 739 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 740                     (t_nblist                    * gmx_restrict       nlist,
 741                      rvec                        * gmx_restrict          xx,
 742                      rvec                        * gmx_restrict          ff,
 743                      t_forcerec                  * gmx_restrict          fr,
 744                      t_mdatoms                   * gmx_restrict     mdatoms,
 745                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 746                      t_nrnb                      * gmx_restrict        nrnb)
 747 {
 748     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 749      * just 0 for non-waters.
 750      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 751      * jnr indices corresponding to data put in the four positions in the SIMD register.
 752      */
 753     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 754     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 755     int              jnrA,jnrB;
 756     int              j_coord_offsetA,j_coord_offsetB;
 757     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 758     real             rcutoff_scalar;
 759     real             *shiftvec,*fshift,*x,*f;
 760     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 761     int              vdwioffset0;
 762     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 763     int              vdwioffset1;
 764     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 765     int              vdwioffset2;
 766     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 767     int              vdwjidx0A,vdwjidx0B;
 768     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 769     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 770     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 771     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 772     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 773     real             *charge;
 774     int              nvdwtype;
 775     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 776     int              *vdwtype;
 777     real             *vdwparam;
 778     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 779     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 780     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 781     real             *ewtab;
 782     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 783     real             rswitch_scalar,d_scalar;
 784     _fjsp_v2r8       itab_tmp;
 785     _fjsp_v2r8       dummy_mask,cutoff_mask;
 786     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 787     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 788     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 789
 790     x                = xx[0];
 791     f                = ff[0];
 792
 793     nri              = nlist->nri;
 794     iinr             = nlist->iinr;
 795     jindex           = nlist->jindex;
 796     jjnr             = nlist->jjnr;
 797     shiftidx         = nlist->shift;
 798     gid              = nlist->gid;
 799     shiftvec         = fr->shift_vec[0];
 800     fshift           = fr->fshift[0];
 801     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 802     charge           = mdatoms->chargeA;
 803     nvdwtype         = fr->ntype;
 804     vdwparam         = fr->nbfp;
 805     vdwtype          = mdatoms->typeA;
 806
 807     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 808     ewtab            = fr->ic->tabq_coul_FDV0;
 809     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 810     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 811
 812     /* Setup water-specific parameters */
 813     inr              = nlist->iinr[0];
 814     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 815     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 816     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 817     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 818
 819     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 820     rcutoff_scalar   = fr->rcoulomb;
 821     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 822     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 823
 824     rswitch_scalar   = fr->rcoulomb_switch;
 825     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 826     /* Setup switch parameters */
 827     d_scalar         = rcutoff_scalar-rswitch_scalar;
 828     d                = gmx_fjsp_set1_v2r8(d_scalar);
 829     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 830     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 831     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 832     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 833     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 834     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 835
 836     /* Avoid stupid compiler warnings */
 837     jnrA = jnrB = 0;
 838     j_coord_offsetA = 0;
 839     j_coord_offsetB = 0;
 840
 841     outeriter        = 0;
 842     inneriter        = 0;
 843
 844     /* Start outer loop over neighborlists */
 845     for(iidx=0; iidx<nri; iidx++)
 846     {
 847         /* Load shift vector for this list */
 848         i_shift_offset   = DIM*shiftidx[iidx];
 849
 850         /* Load limits for loop over neighbors */
 851         j_index_start    = jindex[iidx];
 852         j_index_end      = jindex[iidx+1];
 853
 854         /* Get outer coordinate index */
 855         inr              = iinr[iidx];
 856         i_coord_offset   = DIM*inr;
 857
 858         /* Load i particle coords and add shift vector */
 859         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 860                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 861
 862         fix0             = _fjsp_setzero_v2r8();
 863         fiy0             = _fjsp_setzero_v2r8();
 864         fiz0             = _fjsp_setzero_v2r8();
 865         fix1             = _fjsp_setzero_v2r8();
 866         fiy1             = _fjsp_setzero_v2r8();
 867         fiz1             = _fjsp_setzero_v2r8();
 868         fix2             = _fjsp_setzero_v2r8();
 869         fiy2             = _fjsp_setzero_v2r8();
 870         fiz2             = _fjsp_setzero_v2r8();
 871
 872         /* Start inner kernel loop */
 873         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 874         {
 875
 876             /* Get j neighbor index, and coordinate index */
 877             jnrA             = jjnr[jidx];
 878             jnrB             = jjnr[jidx+1];
 879             j_coord_offsetA  = DIM*jnrA;
 880             j_coord_offsetB  = DIM*jnrB;
 881
 882             /* load j atom coordinates */
 883             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 884                                               &jx0,&jy0,&jz0);
 885
 886             /* Calculate displacement vector */
 887             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 888             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 889             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 890             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 891             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 892             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 893             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 894             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 895             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 896
 897             /* Calculate squared distance and things based on it */
 898             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 899             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 900             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 901
 902             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 903             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 904             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 905
 906             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 907             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 908             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 909
 910             /* Load parameters for j particles */
 911             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 912             vdwjidx0A        = 2*vdwtype[jnrA+0];
 913             vdwjidx0B        = 2*vdwtype[jnrB+0];
 914
 915             fjx0             = _fjsp_setzero_v2r8();
 916             fjy0             = _fjsp_setzero_v2r8();
 917             fjz0             = _fjsp_setzero_v2r8();
 918
 919             /**************************
 920              * CALCULATE INTERACTIONS *
 921              **************************/
 922
 923             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 924             {
 925
 926             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 927
 928             /* Compute parameters for interactions between i and j atoms */
 929             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 930             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 931                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 932
 933             /* EWALD ELECTROSTATICS */
 934
 935             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 936             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 937             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 938             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 939             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 940
 941             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 942             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 943             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 944             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 945             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 946             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 947             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 948             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 949             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 950             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 951
 952             /* LENNARD-JONES DISPERSION/REPULSION */
 953
 954             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 955             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 956             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 957             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 958             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 959
 960             d                = _fjsp_sub_v2r8(r00,rswitch);
 961             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 962             d2               = _fjsp_mul_v2r8(d,d);
 963             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 964
 965             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 966
 967             /* Evaluate switch function */
 968             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 969             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 970             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 971             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 972
 973             fscal            = _fjsp_add_v2r8(felec,fvdw);
 974
 975             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 976
 977             /* Update vectorial force */
 978             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 979             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 980             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 981
 982             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 983             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 984             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 985
 986             }
 987
 988             /**************************
 989              * CALCULATE INTERACTIONS *
 990              **************************/
 991
 992             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 993             {
 994
 995             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 996
 997             /* Compute parameters for interactions between i and j atoms */
 998             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 999
1000             /* EWALD ELECTROSTATICS */
1001
1002             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1003             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1004             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1005             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1006             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1007
1008             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1009             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1010             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1011             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1012             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1013             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1014             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1015             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1016             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1017             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1018
1019             d                = _fjsp_sub_v2r8(r10,rswitch);
1020             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1021             d2               = _fjsp_mul_v2r8(d,d);
1022             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1023
1024             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1025
1026             /* Evaluate switch function */
1027             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1028             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1029             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1030
1031             fscal            = felec;
1032
1033             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1034
1035             /* Update vectorial force */
1036             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1037             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1038             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1039
1040             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1041             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1042             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1043
1044             }
1045
1046             /**************************
1047              * CALCULATE INTERACTIONS *
1048              **************************/
1049
1050             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1051             {
1052
1053             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1054
1055             /* Compute parameters for interactions between i and j atoms */
1056             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1057
1058             /* EWALD ELECTROSTATICS */
1059
1060             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1061             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1062             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1063             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1064             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1065
1066             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1067             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1068             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1069             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1070             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1071             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1072             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1073             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1074             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1075             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1076
1077             d                = _fjsp_sub_v2r8(r20,rswitch);
1078             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1079             d2               = _fjsp_mul_v2r8(d,d);
1080             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1081
1082             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1083
1084             /* Evaluate switch function */
1085             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1086             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1087             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1088
1089             fscal            = felec;
1090
1091             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1092
1093             /* Update vectorial force */
1094             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1095             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1096             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1097
1098             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1099             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1100             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1101
1102             }
1103
1104             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1105
1106             /* Inner loop uses 213 flops */
1107         }
1108
1109         if(jidx<j_index_end)
1110         {
1111
1112             jnrA             = jjnr[jidx];
1113             j_coord_offsetA  = DIM*jnrA;
1114
1115             /* load j atom coordinates */
1116             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1117                                               &jx0,&jy0,&jz0);
1118
1119             /* Calculate displacement vector */
1120             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1121             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1122             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1123             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1124             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1125             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1126             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1127             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1128             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1129
1130             /* Calculate squared distance and things based on it */
1131             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1132             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1133             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1134
1135             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1136             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1137             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1138
1139             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1140             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1141             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1142
1143             /* Load parameters for j particles */
1144             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1145             vdwjidx0A        = 2*vdwtype[jnrA+0];
1146
1147             fjx0             = _fjsp_setzero_v2r8();
1148             fjy0             = _fjsp_setzero_v2r8();
1149             fjz0             = _fjsp_setzero_v2r8();
1150
1151             /**************************
1152              * CALCULATE INTERACTIONS *
1153              **************************/
1154
1155             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1156             {
1157
1158             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1159
1160             /* Compute parameters for interactions between i and j atoms */
1161             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1162             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1163
1164             /* EWALD ELECTROSTATICS */
1165
1166             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1167             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1168             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1169             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1170             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1171
1172             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1173             ewtabD           = _fjsp_setzero_v2r8();
1174             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1175             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1176             ewtabFn          = _fjsp_setzero_v2r8();
1177             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1178             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1179             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1180             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
1181             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1182
1183             /* LENNARD-JONES DISPERSION/REPULSION */
1184
1185             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1186             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1187             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1188             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1189             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1190
1191             d                = _fjsp_sub_v2r8(r00,rswitch);
1192             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1193             d2               = _fjsp_mul_v2r8(d,d);
1194             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1195
1196             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1197
1198             /* Evaluate switch function */
1199             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1200             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
1201             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1202             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1203
1204             fscal            = _fjsp_add_v2r8(felec,fvdw);
1205
1206             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1207
1208             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1209
1210             /* Update vectorial force */
1211             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1212             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1213             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1214
1215             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1216             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1217             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1218
1219             }
1220
1221             /**************************
1222              * CALCULATE INTERACTIONS *
1223              **************************/
1224
1225             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1226             {
1227
1228             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1229
1230             /* Compute parameters for interactions between i and j atoms */
1231             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1232
1233             /* EWALD ELECTROSTATICS */
1234
1235             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1236             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1237             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1238             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1239             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1240
1241             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1242             ewtabD           = _fjsp_setzero_v2r8();
1243             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1244             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1245             ewtabFn          = _fjsp_setzero_v2r8();
1246             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1247             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1248             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1249             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1250             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1251
1252             d                = _fjsp_sub_v2r8(r10,rswitch);
1253             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1254             d2               = _fjsp_mul_v2r8(d,d);
1255             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1256
1257             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1258
1259             /* Evaluate switch function */
1260             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1261             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1262             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1263
1264             fscal            = felec;
1265
1266             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1267
1268             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1269
1270             /* Update vectorial force */
1271             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1272             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1273             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1274
1275             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1276             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1277             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1278
1279             }
1280
1281             /**************************
1282              * CALCULATE INTERACTIONS *
1283              **************************/
1284
1285             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1286             {
1287
1288             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1289
1290             /* Compute parameters for interactions between i and j atoms */
1291             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1292
1293             /* EWALD ELECTROSTATICS */
1294
1295             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1296             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1297             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1298             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1299             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1300
1301             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1302             ewtabD           = _fjsp_setzero_v2r8();
1303             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1304             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1305             ewtabFn          = _fjsp_setzero_v2r8();
1306             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1307             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1308             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1309             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1310             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1311
1312             d                = _fjsp_sub_v2r8(r20,rswitch);
1313             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1314             d2               = _fjsp_mul_v2r8(d,d);
1315             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1316
1317             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1318
1319             /* Evaluate switch function */
1320             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1321             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1322             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1323
1324             fscal            = felec;
1325
1326             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1327
1328             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1329
1330             /* Update vectorial force */
1331             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1332             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1333             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1334
1335             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1336             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1337             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1338
1339             }
1340
1341             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1342
1343             /* Inner loop uses 213 flops */
1344         }
1345
1346         /* End of innermost loop */
1347
1348         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1349                                               f+i_coord_offset,fshift+i_shift_offset);
1350
1351         /* Increment number of inner iterations */
1352         inneriter                  += j_index_end - j_index_start;
1353
1354         /* Outer loop uses 18 flops */
1355     }
1356
1357     /* Increment number of outer iterations */
1358     outeriter        += nri;
1359
1360     /* Update outer/inner flops */
1361
1362     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*213);
1363 }