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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 258                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 259
 260             /* REACTION-FIELD ELECTROSTATICS */
 261             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 262             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 263
 264             /* LENNARD-JONES DISPERSION/REPULSION */
 265
 266             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 267             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 268             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 269             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 270             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 271
 272             /* Update potential sum for this i atom from the interaction with this j atom. */
 273             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 274             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 275             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 276             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 277
 278             fscal            = _fjsp_add_v2r8(felec,fvdw);
 279
 280             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 281
 282             /* Update vectorial force */
 283             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 284             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 285             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 286
 287             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 288
 289             /* Inner loop uses 47 flops */
 290         }
 291
 292         /* End of innermost loop */
 293
 294         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 295                                               f+i_coord_offset,fshift+i_shift_offset);
 296
 297         ggid                        = gid[iidx];
 298         /* Update potential energies */
 299         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 300         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 301
 302         /* Increment number of inner iterations */
 303         inneriter                  += j_index_end - j_index_start;
 304
 305         /* Outer loop uses 9 flops */
 306     }
 307
 308     /* Increment number of outer iterations */
 309     outeriter        += nri;
 310
 311     /* Update outer/inner flops */
 312
 313     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*47);
 314 }
 315 /*
 316  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 317  * Electrostatics interaction: ReactionField
 318  * VdW interaction:            LennardJones
 319  * Geometry:                   Particle-Particle
 320  * Calculate force/pot:        Force
 321  */
 322 void
 323 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 324                     (t_nblist                    * gmx_restrict       nlist,
 325                      rvec                        * gmx_restrict          xx,
 326                      rvec                        * gmx_restrict          ff,
 327                      t_forcerec                  * gmx_restrict          fr,
 328                      t_mdatoms                   * gmx_restrict     mdatoms,
 329                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 330                      t_nrnb                      * gmx_restrict        nrnb)
 331 {
 332     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 333      * just 0 for non-waters.
 334      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 335      * jnr indices corresponding to data put in the four positions in the SIMD register.
 336      */
 337     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 338     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 339     int              jnrA,jnrB;
 340     int              j_coord_offsetA,j_coord_offsetB;
 341     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 342     real             rcutoff_scalar;
 343     real             *shiftvec,*fshift,*x,*f;
 344     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 345     int              vdwioffset0;
 346     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 347     int              vdwjidx0A,vdwjidx0B;
 348     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 349     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 350     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 351     real             *charge;
 352     int              nvdwtype;
 353     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 354     int              *vdwtype;
 355     real             *vdwparam;
 356     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 357     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 358     _fjsp_v2r8       itab_tmp;
 359     _fjsp_v2r8       dummy_mask,cutoff_mask;
 360     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 361     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 362     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 363
 364     x                = xx[0];
 365     f                = ff[0];
 366
 367     nri              = nlist->nri;
 368     iinr             = nlist->iinr;
 369     jindex           = nlist->jindex;
 370     jjnr             = nlist->jjnr;
 371     shiftidx         = nlist->shift;
 372     gid              = nlist->gid;
 373     shiftvec         = fr->shift_vec[0];
 374     fshift           = fr->fshift[0];
 375     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 376     charge           = mdatoms->chargeA;
 377     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 378     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 379     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 380     nvdwtype         = fr->ntype;
 381     vdwparam         = fr->nbfp;
 382     vdwtype          = mdatoms->typeA;
 383
 384     /* Avoid stupid compiler warnings */
 385     jnrA = jnrB = 0;
 386     j_coord_offsetA = 0;
 387     j_coord_offsetB = 0;
 388
 389     outeriter        = 0;
 390     inneriter        = 0;
 391
 392     /* Start outer loop over neighborlists */
 393     for(iidx=0; iidx<nri; iidx++)
 394     {
 395         /* Load shift vector for this list */
 396         i_shift_offset   = DIM*shiftidx[iidx];
 397
 398         /* Load limits for loop over neighbors */
 399         j_index_start    = jindex[iidx];
 400         j_index_end      = jindex[iidx+1];
 401
 402         /* Get outer coordinate index */
 403         inr              = iinr[iidx];
 404         i_coord_offset   = DIM*inr;
 405
 406         /* Load i particle coords and add shift vector */
 407         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 408
 409         fix0             = _fjsp_setzero_v2r8();
 410         fiy0             = _fjsp_setzero_v2r8();
 411         fiz0             = _fjsp_setzero_v2r8();
 412
 413         /* Load parameters for i particles */
 414         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 415         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 416
 417         /* Start inner kernel loop */
 418         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 419         {
 420
 421             /* Get j neighbor index, and coordinate index */
 422             jnrA             = jjnr[jidx];
 423             jnrB             = jjnr[jidx+1];
 424             j_coord_offsetA  = DIM*jnrA;
 425             j_coord_offsetB  = DIM*jnrB;
 426
 427             /* load j atom coordinates */
 428             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 429                                               &jx0,&jy0,&jz0);
 430
 431             /* Calculate displacement vector */
 432             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 433             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 434             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 435
 436             /* Calculate squared distance and things based on it */
 437             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 438
 439             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 440
 441             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 442
 443             /* Load parameters for j particles */
 444             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 445             vdwjidx0A        = 2*vdwtype[jnrA+0];
 446             vdwjidx0B        = 2*vdwtype[jnrB+0];
 447
 448             /**************************
 449              * CALCULATE INTERACTIONS *
 450              **************************/
 451
 452             /* Compute parameters for interactions between i and j atoms */
 453             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 454             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 455                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 456
 457             /* REACTION-FIELD ELECTROSTATICS */
 458             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 459
 460             /* LENNARD-JONES DISPERSION/REPULSION */
 461
 462             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 463             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 464
 465             fscal            = _fjsp_add_v2r8(felec,fvdw);
 466
 467             /* Update vectorial force */
 468             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 469             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 470             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 471
 472             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 473
 474             /* Inner loop uses 37 flops */
 475         }
 476
 477         if(jidx<j_index_end)
 478         {
 479
 480             jnrA             = jjnr[jidx];
 481             j_coord_offsetA  = DIM*jnrA;
 482
 483             /* load j atom coordinates */
 484             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 485                                               &jx0,&jy0,&jz0);
 486
 487             /* Calculate displacement vector */
 488             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 489             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 490             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 491
 492             /* Calculate squared distance and things based on it */
 493             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 494
 495             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 496
 497             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 498
 499             /* Load parameters for j particles */
 500             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 501             vdwjidx0A        = 2*vdwtype[jnrA+0];
 502
 503             /**************************
 504              * CALCULATE INTERACTIONS *
 505              **************************/
 506
 507             /* Compute parameters for interactions between i and j atoms */
 508             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 509             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 510                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 511
 512             /* REACTION-FIELD ELECTROSTATICS */
 513             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 514
 515             /* LENNARD-JONES DISPERSION/REPULSION */
 516
 517             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 518             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 519
 520             fscal            = _fjsp_add_v2r8(felec,fvdw);
 521
 522             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 523
 524             /* Update vectorial force */
 525             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 526             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 527             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 528
 529             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 530
 531             /* Inner loop uses 37 flops */
 532         }
 533
 534         /* End of innermost loop */
 535
 536         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 537                                               f+i_coord_offset,fshift+i_shift_offset);
 538
 539         /* Increment number of inner iterations */
 540         inneriter                  += j_index_end - j_index_start;
 541
 542         /* Outer loop uses 7 flops */
 543     }
 544
 545     /* Increment number of outer iterations */
 546     outeriter        += nri;
 547
 548     /* Update outer/inner flops */
 549
 550     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 551 }