src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRF_VdwLJ_GeomP1P1_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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwjidx0A,vdwjidx0B;
  84     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  85     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  87     real             *charge;
  88     int              nvdwtype;
  89     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  90     int              *vdwtype;
  91     real             *vdwparam;
  92     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  93     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  94     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 114     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 115     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 116     nvdwtype         = fr->ntype;
 117     vdwparam         = fr->nbfp;
 118     vdwtype          = mdatoms->typeA;
 119
 120     /* Avoid stupid compiler warnings */
 121     jnrA = jnrB = 0;
 122     j_coord_offsetA = 0;
 123     j_coord_offsetB = 0;
 124
 125     outeriter        = 0;
 126     inneriter        = 0;
 127
 128     /* Start outer loop over neighborlists */
 129     for(iidx=0; iidx<nri; iidx++)
 130     {
 131         /* Load shift vector for this list */
 132         i_shift_offset   = DIM*shiftidx[iidx];
 133
 134         /* Load limits for loop over neighbors */
 135         j_index_start    = jindex[iidx];
 136         j_index_end      = jindex[iidx+1];
 137
 138         /* Get outer coordinate index */
 139         inr              = iinr[iidx];
 140         i_coord_offset   = DIM*inr;
 141
 142         /* Load i particle coords and add shift vector */
 143         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 144
 145         fix0             = _fjsp_setzero_v2r8();
 146         fiy0             = _fjsp_setzero_v2r8();
 147         fiz0             = _fjsp_setzero_v2r8();
 148
 149         /* Load parameters for i particles */
 150         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 151         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 152
 153         /* Reset potential sums */
 154         velecsum         = _fjsp_setzero_v2r8();
 155         vvdwsum          = _fjsp_setzero_v2r8();
 156
 157         /* Start inner kernel loop */
 158         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 159         {
 160
 161             /* Get j neighbor index, and coordinate index */
 162             jnrA             = jjnr[jidx];
 163             jnrB             = jjnr[jidx+1];
 164             j_coord_offsetA  = DIM*jnrA;
 165             j_coord_offsetB  = DIM*jnrB;
 166
 167             /* load j atom coordinates */
 168             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 169                                               &jx0,&jy0,&jz0);
 170
 171             /* Calculate displacement vector */
 172             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 173             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 174             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 175
 176             /* Calculate squared distance and things based on it */
 177             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 178
 179             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 180
 181             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 182
 183             /* Load parameters for j particles */
 184             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 185             vdwjidx0A        = 2*vdwtype[jnrA+0];
 186             vdwjidx0B        = 2*vdwtype[jnrB+0];
 187
 188             /**************************
 189              * CALCULATE INTERACTIONS *
 190              **************************/
 191
 192             /* Compute parameters for interactions between i and j atoms */
 193             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 194             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 195                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 196
 197             /* REACTION-FIELD ELECTROSTATICS */
 198             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 199             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 200
 201             /* LENNARD-JONES DISPERSION/REPULSION */
 202
 203             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 204             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 205             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 206             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 207             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 208
 209             /* Update potential sum for this i atom from the interaction with this j atom. */
 210             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 211             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 212
 213             fscal            = _fjsp_add_v2r8(felec,fvdw);
 214
 215             /* Update vectorial force */
 216             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 217             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 218             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 219
 220             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 221
 222             /* Inner loop uses 47 flops */
 223         }
 224
 225         if(jidx<j_index_end)
 226         {
 227
 228             jnrA             = jjnr[jidx];
 229             j_coord_offsetA  = DIM*jnrA;
 230
 231             /* load j atom coordinates */
 232             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 233                                               &jx0,&jy0,&jz0);
 234
 235             /* Calculate displacement vector */
 236             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 237             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 238             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 239
 240             /* Calculate squared distance and things based on it */
 241             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 242
 243             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 244
 245             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 246
 247             /* Load parameters for j particles */
 248             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 249             vdwjidx0A        = 2*vdwtype[jnrA+0];
 250
 251             /**************************
 252              * CALCULATE INTERACTIONS *
 253              **************************/
 254
 255             /* Compute parameters for interactions between i and j atoms */
 256             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 257             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 258
 259             /* REACTION-FIELD ELECTROSTATICS */
 260             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 261             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 262
 263             /* LENNARD-JONES DISPERSION/REPULSION */
 264
 265             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 266             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 267             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 268             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 269             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 270
 271             /* Update potential sum for this i atom from the interaction with this j atom. */
 272             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 273             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 274             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 275             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 276
 277             fscal            = _fjsp_add_v2r8(felec,fvdw);
 278
 279             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 280
 281             /* Update vectorial force */
 282             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 283             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 284             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 285
 286             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 287
 288             /* Inner loop uses 47 flops */
 289         }
 290
 291         /* End of innermost loop */
 292
 293         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 294                                               f+i_coord_offset,fshift+i_shift_offset);
 295
 296         ggid                        = gid[iidx];
 297         /* Update potential energies */
 298         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 299         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 300
 301         /* Increment number of inner iterations */
 302         inneriter                  += j_index_end - j_index_start;
 303
 304         /* Outer loop uses 9 flops */
 305     }
 306
 307     /* Increment number of outer iterations */
 308     outeriter        += nri;
 309
 310     /* Update outer/inner flops */
 311
 312     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*47);
 313 }
 314 /*
 315  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 316  * Electrostatics interaction: ReactionField
 317  * VdW interaction:            LennardJones
 318  * Geometry:                   Particle-Particle
 319  * Calculate force/pot:        Force
 320  */
 321 void
 322 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 323                     (t_nblist                    * gmx_restrict       nlist,
 324                      rvec                        * gmx_restrict          xx,
 325                      rvec                        * gmx_restrict          ff,
 326                      t_forcerec                  * gmx_restrict          fr,
 327                      t_mdatoms                   * gmx_restrict     mdatoms,
 328                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 329                      t_nrnb                      * gmx_restrict        nrnb)
 330 {
 331     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 332      * just 0 for non-waters.
 333      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 334      * jnr indices corresponding to data put in the four positions in the SIMD register.
 335      */
 336     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 337     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 338     int              jnrA,jnrB;
 339     int              j_coord_offsetA,j_coord_offsetB;
 340     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 341     real             rcutoff_scalar;
 342     real             *shiftvec,*fshift,*x,*f;
 343     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 344     int              vdwioffset0;
 345     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 346     int              vdwjidx0A,vdwjidx0B;
 347     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 348     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 349     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 350     real             *charge;
 351     int              nvdwtype;
 352     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 353     int              *vdwtype;
 354     real             *vdwparam;
 355     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 356     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 357     _fjsp_v2r8       itab_tmp;
 358     _fjsp_v2r8       dummy_mask,cutoff_mask;
 359     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 360     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 361     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 362
 363     x                = xx[0];
 364     f                = ff[0];
 365
 366     nri              = nlist->nri;
 367     iinr             = nlist->iinr;
 368     jindex           = nlist->jindex;
 369     jjnr             = nlist->jjnr;
 370     shiftidx         = nlist->shift;
 371     gid              = nlist->gid;
 372     shiftvec         = fr->shift_vec[0];
 373     fshift           = fr->fshift[0];
 374     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 375     charge           = mdatoms->chargeA;
 376     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 377     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 378     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 379     nvdwtype         = fr->ntype;
 380     vdwparam         = fr->nbfp;
 381     vdwtype          = mdatoms->typeA;
 382
 383     /* Avoid stupid compiler warnings */
 384     jnrA = jnrB = 0;
 385     j_coord_offsetA = 0;
 386     j_coord_offsetB = 0;
 387
 388     outeriter        = 0;
 389     inneriter        = 0;
 390
 391     /* Start outer loop over neighborlists */
 392     for(iidx=0; iidx<nri; iidx++)
 393     {
 394         /* Load shift vector for this list */
 395         i_shift_offset   = DIM*shiftidx[iidx];
 396
 397         /* Load limits for loop over neighbors */
 398         j_index_start    = jindex[iidx];
 399         j_index_end      = jindex[iidx+1];
 400
 401         /* Get outer coordinate index */
 402         inr              = iinr[iidx];
 403         i_coord_offset   = DIM*inr;
 404
 405         /* Load i particle coords and add shift vector */
 406         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 407
 408         fix0             = _fjsp_setzero_v2r8();
 409         fiy0             = _fjsp_setzero_v2r8();
 410         fiz0             = _fjsp_setzero_v2r8();
 411
 412         /* Load parameters for i particles */
 413         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 414         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 415
 416         /* Start inner kernel loop */
 417         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 418         {
 419
 420             /* Get j neighbor index, and coordinate index */
 421             jnrA             = jjnr[jidx];
 422             jnrB             = jjnr[jidx+1];
 423             j_coord_offsetA  = DIM*jnrA;
 424             j_coord_offsetB  = DIM*jnrB;
 425
 426             /* load j atom coordinates */
 427             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 428                                               &jx0,&jy0,&jz0);
 429
 430             /* Calculate displacement vector */
 431             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 432             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 433             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 434
 435             /* Calculate squared distance and things based on it */
 436             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 437
 438             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 439
 440             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 441
 442             /* Load parameters for j particles */
 443             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 444             vdwjidx0A        = 2*vdwtype[jnrA+0];
 445             vdwjidx0B        = 2*vdwtype[jnrB+0];
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 453             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 454                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 455
 456             /* REACTION-FIELD ELECTROSTATICS */
 457             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 458
 459             /* LENNARD-JONES DISPERSION/REPULSION */
 460
 461             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 462             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 463
 464             fscal            = _fjsp_add_v2r8(felec,fvdw);
 465
 466             /* Update vectorial force */
 467             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 468             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 469             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 470
 471             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 472
 473             /* Inner loop uses 37 flops */
 474         }
 475
 476         if(jidx<j_index_end)
 477         {
 478
 479             jnrA             = jjnr[jidx];
 480             j_coord_offsetA  = DIM*jnrA;
 481
 482             /* load j atom coordinates */
 483             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 484                                               &jx0,&jy0,&jz0);
 485
 486             /* Calculate displacement vector */
 487             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 488             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 489             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 490
 491             /* Calculate squared distance and things based on it */
 492             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 493
 494             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 495
 496             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 497
 498             /* Load parameters for j particles */
 499             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 500             vdwjidx0A        = 2*vdwtype[jnrA+0];
 501
 502             /**************************
 503              * CALCULATE INTERACTIONS *
 504              **************************/
 505
 506             /* Compute parameters for interactions between i and j atoms */
 507             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 508             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&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 }