src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRF_VdwLJ_GeomP1P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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_ElecRF_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRF_VdwLJ_GeomP1P1_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              vdwjidx0A,vdwjidx0B;
  82     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  83     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  84     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  85     real             *charge;
  86     int              nvdwtype;
  87     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  88     int              *vdwtype;
  89     real             *vdwparam;
  90     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  91     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  92     _fjsp_v2r8       itab_tmp;
  93     _fjsp_v2r8       dummy_mask,cutoff_mask;
  94     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  95     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  96     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  97
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 110     charge           = mdatoms->chargeA;
 111     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 112     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 113     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 114     nvdwtype         = fr->ntype;
 115     vdwparam         = fr->nbfp;
 116     vdwtype          = mdatoms->typeA;
 117
 118     /* Avoid stupid compiler warnings */
 119     jnrA = jnrB = 0;
 120     j_coord_offsetA = 0;
 121     j_coord_offsetB = 0;
 122
 123     outeriter        = 0;
 124     inneriter        = 0;
 125
 126     /* Start outer loop over neighborlists */
 127     for(iidx=0; iidx<nri; iidx++)
 128     {
 129         /* Load shift vector for this list */
 130         i_shift_offset   = DIM*shiftidx[iidx];
 131
 132         /* Load limits for loop over neighbors */
 133         j_index_start    = jindex[iidx];
 134         j_index_end      = jindex[iidx+1];
 135
 136         /* Get outer coordinate index */
 137         inr              = iinr[iidx];
 138         i_coord_offset   = DIM*inr;
 139
 140         /* Load i particle coords and add shift vector */
 141         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 142
 143         fix0             = _fjsp_setzero_v2r8();
 144         fiy0             = _fjsp_setzero_v2r8();
 145         fiz0             = _fjsp_setzero_v2r8();
 146
 147         /* Load parameters for i particles */
 148         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 149         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 150
 151         /* Reset potential sums */
 152         velecsum         = _fjsp_setzero_v2r8();
 153         vvdwsum          = _fjsp_setzero_v2r8();
 154
 155         /* Start inner kernel loop */
 156         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 157         {
 158
 159             /* Get j neighbor index, and coordinate index */
 160             jnrA             = jjnr[jidx];
 161             jnrB             = jjnr[jidx+1];
 162             j_coord_offsetA  = DIM*jnrA;
 163             j_coord_offsetB  = DIM*jnrB;
 164
 165             /* load j atom coordinates */
 166             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 167                                               &jx0,&jy0,&jz0);
 168
 169             /* Calculate displacement vector */
 170             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 171             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 172             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 173
 174             /* Calculate squared distance and things based on it */
 175             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 176
 177             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 178
 179             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 180
 181             /* Load parameters for j particles */
 182             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 183             vdwjidx0A        = 2*vdwtype[jnrA+0];
 184             vdwjidx0B        = 2*vdwtype[jnrB+0];
 185
 186             /**************************
 187              * CALCULATE INTERACTIONS *
 188              **************************/
 189
 190             /* Compute parameters for interactions between i and j atoms */
 191             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 192             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 193                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 194
 195             /* REACTION-FIELD ELECTROSTATICS */
 196             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 197             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 198
 199             /* LENNARD-JONES DISPERSION/REPULSION */
 200
 201             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 202             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 203             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 204             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 205             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 206
 207             /* Update potential sum for this i atom from the interaction with this j atom. */
 208             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 209             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 210
 211             fscal            = _fjsp_add_v2r8(felec,fvdw);
 212
 213             /* Update vectorial force */
 214             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 215             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 216             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 217
 218             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 219
 220             /* Inner loop uses 47 flops */
 221         }
 222
 223         if(jidx<j_index_end)
 224         {
 225
 226             jnrA             = jjnr[jidx];
 227             j_coord_offsetA  = DIM*jnrA;
 228
 229             /* load j atom coordinates */
 230             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 231                                               &jx0,&jy0,&jz0);
 232
 233             /* Calculate displacement vector */
 234             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 235             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 236             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 237
 238             /* Calculate squared distance and things based on it */
 239             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 240
 241             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 242
 243             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 244
 245             /* Load parameters for j particles */
 246             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 247             vdwjidx0A        = 2*vdwtype[jnrA+0];
 248
 249             /**************************
 250              * CALCULATE INTERACTIONS *
 251              **************************/
 252
 253             /* Compute parameters for interactions between i and j atoms */
 254             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 255             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 256                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 257
 258             /* REACTION-FIELD ELECTROSTATICS */
 259             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 260             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 261
 262             /* LENNARD-JONES DISPERSION/REPULSION */
 263
 264             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 265             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 266             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 267             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 268             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 269
 270             /* Update potential sum for this i atom from the interaction with this j atom. */
 271             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 272             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 273             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 274             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 275
 276             fscal            = _fjsp_add_v2r8(felec,fvdw);
 277
 278             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 279
 280             /* Update vectorial force */
 281             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 282             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 283             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 284
 285             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 286
 287             /* Inner loop uses 47 flops */
 288         }
 289
 290         /* End of innermost loop */
 291
 292         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 293                                               f+i_coord_offset,fshift+i_shift_offset);
 294
 295         ggid                        = gid[iidx];
 296         /* Update potential energies */
 297         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 298         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 299
 300         /* Increment number of inner iterations */
 301         inneriter                  += j_index_end - j_index_start;
 302
 303         /* Outer loop uses 9 flops */
 304     }
 305
 306     /* Increment number of outer iterations */
 307     outeriter        += nri;
 308
 309     /* Update outer/inner flops */
 310
 311     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*47);
 312 }
 313 /*
 314  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 315  * Electrostatics interaction: ReactionField
 316  * VdW interaction:            LennardJones
 317  * Geometry:                   Particle-Particle
 318  * Calculate force/pot:        Force
 319  */
 320 void
 321 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 322                     (t_nblist                    * gmx_restrict       nlist,
 323                      rvec                        * gmx_restrict          xx,
 324                      rvec                        * gmx_restrict          ff,
 325                      struct t_forcerec           * gmx_restrict          fr,
 326                      t_mdatoms                   * gmx_restrict     mdatoms,
 327                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 328                      t_nrnb                      * gmx_restrict        nrnb)
 329 {
 330     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 331      * just 0 for non-waters.
 332      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 333      * jnr indices corresponding to data put in the four positions in the SIMD register.
 334      */
 335     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 336     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 337     int              jnrA,jnrB;
 338     int              j_coord_offsetA,j_coord_offsetB;
 339     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 340     real             rcutoff_scalar;
 341     real             *shiftvec,*fshift,*x,*f;
 342     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 343     int              vdwioffset0;
 344     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 345     int              vdwjidx0A,vdwjidx0B;
 346     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 347     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 348     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 349     real             *charge;
 350     int              nvdwtype;
 351     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 352     int              *vdwtype;
 353     real             *vdwparam;
 354     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 355     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 356     _fjsp_v2r8       itab_tmp;
 357     _fjsp_v2r8       dummy_mask,cutoff_mask;
 358     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 359     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 360     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 361
 362     x                = xx[0];
 363     f                = ff[0];
 364
 365     nri              = nlist->nri;
 366     iinr             = nlist->iinr;
 367     jindex           = nlist->jindex;
 368     jjnr             = nlist->jjnr;
 369     shiftidx         = nlist->shift;
 370     gid              = nlist->gid;
 371     shiftvec         = fr->shift_vec[0];
 372     fshift           = fr->fshift[0];
 373     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 374     charge           = mdatoms->chargeA;
 375     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 376     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 377     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 378     nvdwtype         = fr->ntype;
 379     vdwparam         = fr->nbfp;
 380     vdwtype          = mdatoms->typeA;
 381
 382     /* Avoid stupid compiler warnings */
 383     jnrA = jnrB = 0;
 384     j_coord_offsetA = 0;
 385     j_coord_offsetB = 0;
 386
 387     outeriter        = 0;
 388     inneriter        = 0;
 389
 390     /* Start outer loop over neighborlists */
 391     for(iidx=0; iidx<nri; iidx++)
 392     {
 393         /* Load shift vector for this list */
 394         i_shift_offset   = DIM*shiftidx[iidx];
 395
 396         /* Load limits for loop over neighbors */
 397         j_index_start    = jindex[iidx];
 398         j_index_end      = jindex[iidx+1];
 399
 400         /* Get outer coordinate index */
 401         inr              = iinr[iidx];
 402         i_coord_offset   = DIM*inr;
 403
 404         /* Load i particle coords and add shift vector */
 405         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 406
 407         fix0             = _fjsp_setzero_v2r8();
 408         fiy0             = _fjsp_setzero_v2r8();
 409         fiz0             = _fjsp_setzero_v2r8();
 410
 411         /* Load parameters for i particles */
 412         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 413         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 414
 415         /* Start inner kernel loop */
 416         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 417         {
 418
 419             /* Get j neighbor index, and coordinate index */
 420             jnrA             = jjnr[jidx];
 421             jnrB             = jjnr[jidx+1];
 422             j_coord_offsetA  = DIM*jnrA;
 423             j_coord_offsetB  = DIM*jnrB;
 424
 425             /* load j atom coordinates */
 426             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 427                                               &jx0,&jy0,&jz0);
 428
 429             /* Calculate displacement vector */
 430             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 431             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 432             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 433
 434             /* Calculate squared distance and things based on it */
 435             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 436
 437             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 438
 439             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 440
 441             /* Load parameters for j particles */
 442             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 443             vdwjidx0A        = 2*vdwtype[jnrA+0];
 444             vdwjidx0B        = 2*vdwtype[jnrB+0];
 445
 446             /**************************
 447              * CALCULATE INTERACTIONS *
 448              **************************/
 449
 450             /* Compute parameters for interactions between i and j atoms */
 451             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 452             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 453                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 454
 455             /* REACTION-FIELD ELECTROSTATICS */
 456             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 457
 458             /* LENNARD-JONES DISPERSION/REPULSION */
 459
 460             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 461             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 462
 463             fscal            = _fjsp_add_v2r8(felec,fvdw);
 464
 465             /* Update vectorial force */
 466             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 467             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 468             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 469
 470             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 471
 472             /* Inner loop uses 37 flops */
 473         }
 474
 475         if(jidx<j_index_end)
 476         {
 477
 478             jnrA             = jjnr[jidx];
 479             j_coord_offsetA  = DIM*jnrA;
 480
 481             /* load j atom coordinates */
 482             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 483                                               &jx0,&jy0,&jz0);
 484
 485             /* Calculate displacement vector */
 486             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 487             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 488             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 489
 490             /* Calculate squared distance and things based on it */
 491             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 492
 493             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 494
 495             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 496
 497             /* Load parameters for j particles */
 498             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 499             vdwjidx0A        = 2*vdwtype[jnrA+0];
 500
 501             /**************************
 502              * CALCULATE INTERACTIONS *
 503              **************************/
 504
 505             /* Compute parameters for interactions between i and j atoms */
 506             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 507             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 508                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 509
 510             /* REACTION-FIELD ELECTROSTATICS */
 511             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 512
 513             /* LENNARD-JONES DISPERSION/REPULSION */
 514
 515             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 516             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 517
 518             fscal            = _fjsp_add_v2r8(felec,fvdw);
 519
 520             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 521
 522             /* Update vectorial force */
 523             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 524             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 525             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 526
 527             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 528
 529             /* Inner loop uses 37 flops */
 530         }
 531
 532         /* End of innermost loop */
 533
 534         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 535                                               f+i_coord_offset,fshift+i_shift_offset);
 536
 537         /* Increment number of inner iterations */
 538         inneriter                  += j_index_end - j_index_start;
 539
 540         /* Outer loop uses 7 flops */
 541     }
 542
 543     /* Increment number of outer iterations */
 544     outeriter        += nri;
 545
 546     /* Update outer/inner flops */
 547
 548     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 549 }