src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_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_ElecCoul_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_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     nvdwtype         = fr->ntype;
 112     vdwparam         = fr->nbfp;
 113     vdwtype          = mdatoms->typeA;
 114
 115     /* Avoid stupid compiler warnings */
 116     jnrA = jnrB = 0;
 117     j_coord_offsetA = 0;
 118     j_coord_offsetB = 0;
 119
 120     outeriter        = 0;
 121     inneriter        = 0;
 122
 123     /* Start outer loop over neighborlists */
 124     for(iidx=0; iidx<nri; iidx++)
 125     {
 126         /* Load shift vector for this list */
 127         i_shift_offset   = DIM*shiftidx[iidx];
 128
 129         /* Load limits for loop over neighbors */
 130         j_index_start    = jindex[iidx];
 131         j_index_end      = jindex[iidx+1];
 132
 133         /* Get outer coordinate index */
 134         inr              = iinr[iidx];
 135         i_coord_offset   = DIM*inr;
 136
 137         /* Load i particle coords and add shift vector */
 138         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 139
 140         fix0             = _fjsp_setzero_v2r8();
 141         fiy0             = _fjsp_setzero_v2r8();
 142         fiz0             = _fjsp_setzero_v2r8();
 143
 144         /* Load parameters for i particles */
 145         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 146         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 147
 148         /* Reset potential sums */
 149         velecsum         = _fjsp_setzero_v2r8();
 150         vvdwsum          = _fjsp_setzero_v2r8();
 151
 152         /* Start inner kernel loop */
 153         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 154         {
 155
 156             /* Get j neighbor index, and coordinate index */
 157             jnrA             = jjnr[jidx];
 158             jnrB             = jjnr[jidx+1];
 159             j_coord_offsetA  = DIM*jnrA;
 160             j_coord_offsetB  = DIM*jnrB;
 161
 162             /* load j atom coordinates */
 163             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 164                                               &jx0,&jy0,&jz0);
 165
 166             /* Calculate displacement vector */
 167             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 168             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 169             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 170
 171             /* Calculate squared distance and things based on it */
 172             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 173
 174             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 175
 176             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 177
 178             /* Load parameters for j particles */
 179             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 180             vdwjidx0A        = 2*vdwtype[jnrA+0];
 181             vdwjidx0B        = 2*vdwtype[jnrB+0];
 182
 183             /**************************
 184              * CALCULATE INTERACTIONS *
 185              **************************/
 186
 187             /* Compute parameters for interactions between i and j atoms */
 188             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 189             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 190                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 191
 192             /* COULOMB ELECTROSTATICS */
 193             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 194             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 195
 196             /* LENNARD-JONES DISPERSION/REPULSION */
 197
 198             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 199             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 200             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 201             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 202             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 203
 204             /* Update potential sum for this i atom from the interaction with this j atom. */
 205             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 206             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 207
 208             fscal            = _fjsp_add_v2r8(felec,fvdw);
 209
 210             /* Update vectorial force */
 211             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 212             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 213             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 214
 215             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 216
 217             /* Inner loop uses 43 flops */
 218         }
 219
 220         if(jidx<j_index_end)
 221         {
 222
 223             jnrA             = jjnr[jidx];
 224             j_coord_offsetA  = DIM*jnrA;
 225
 226             /* load j atom coordinates */
 227             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 228                                               &jx0,&jy0,&jz0);
 229
 230             /* Calculate displacement vector */
 231             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 232             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 233             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 234
 235             /* Calculate squared distance and things based on it */
 236             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 237
 238             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 239
 240             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 241
 242             /* Load parameters for j particles */
 243             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 244             vdwjidx0A        = 2*vdwtype[jnrA+0];
 245
 246             /**************************
 247              * CALCULATE INTERACTIONS *
 248              **************************/
 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             /* COULOMB ELECTROSTATICS */
 256             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 257             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 258
 259             /* LENNARD-JONES DISPERSION/REPULSION */
 260
 261             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 262             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 263             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 264             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 265             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 266
 267             /* Update potential sum for this i atom from the interaction with this j atom. */
 268             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 269             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 270             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 271             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 272
 273             fscal            = _fjsp_add_v2r8(felec,fvdw);
 274
 275             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 276
 277             /* Update vectorial force */
 278             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 279             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 280             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 281
 282             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 283
 284             /* Inner loop uses 43 flops */
 285         }
 286
 287         /* End of innermost loop */
 288
 289         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 290                                               f+i_coord_offset,fshift+i_shift_offset);
 291
 292         ggid                        = gid[iidx];
 293         /* Update potential energies */
 294         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 295         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 296
 297         /* Increment number of inner iterations */
 298         inneriter                  += j_index_end - j_index_start;
 299
 300         /* Outer loop uses 9 flops */
 301     }
 302
 303     /* Increment number of outer iterations */
 304     outeriter        += nri;
 305
 306     /* Update outer/inner flops */
 307
 308     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*43);
 309 }
 310 /*
 311  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 312  * Electrostatics interaction: Coulomb
 313  * VdW interaction:            LennardJones
 314  * Geometry:                   Particle-Particle
 315  * Calculate force/pot:        Force
 316  */
 317 void
 318 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 319                     (t_nblist                    * gmx_restrict       nlist,
 320                      rvec                        * gmx_restrict          xx,
 321                      rvec                        * gmx_restrict          ff,
 322                      struct t_forcerec           * gmx_restrict          fr,
 323                      t_mdatoms                   * gmx_restrict     mdatoms,
 324                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 325                      t_nrnb                      * gmx_restrict        nrnb)
 326 {
 327     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 328      * just 0 for non-waters.
 329      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 330      * jnr indices corresponding to data put in the four positions in the SIMD register.
 331      */
 332     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 333     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 334     int              jnrA,jnrB;
 335     int              j_coord_offsetA,j_coord_offsetB;
 336     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 337     real             rcutoff_scalar;
 338     real             *shiftvec,*fshift,*x,*f;
 339     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 340     int              vdwioffset0;
 341     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 342     int              vdwjidx0A,vdwjidx0B;
 343     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 344     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 345     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 346     real             *charge;
 347     int              nvdwtype;
 348     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 349     int              *vdwtype;
 350     real             *vdwparam;
 351     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 352     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 353     _fjsp_v2r8       itab_tmp;
 354     _fjsp_v2r8       dummy_mask,cutoff_mask;
 355     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 356     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 357     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 358
 359     x                = xx[0];
 360     f                = ff[0];
 361
 362     nri              = nlist->nri;
 363     iinr             = nlist->iinr;
 364     jindex           = nlist->jindex;
 365     jjnr             = nlist->jjnr;
 366     shiftidx         = nlist->shift;
 367     gid              = nlist->gid;
 368     shiftvec         = fr->shift_vec[0];
 369     fshift           = fr->fshift[0];
 370     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 371     charge           = mdatoms->chargeA;
 372     nvdwtype         = fr->ntype;
 373     vdwparam         = fr->nbfp;
 374     vdwtype          = mdatoms->typeA;
 375
 376     /* Avoid stupid compiler warnings */
 377     jnrA = jnrB = 0;
 378     j_coord_offsetA = 0;
 379     j_coord_offsetB = 0;
 380
 381     outeriter        = 0;
 382     inneriter        = 0;
 383
 384     /* Start outer loop over neighborlists */
 385     for(iidx=0; iidx<nri; iidx++)
 386     {
 387         /* Load shift vector for this list */
 388         i_shift_offset   = DIM*shiftidx[iidx];
 389
 390         /* Load limits for loop over neighbors */
 391         j_index_start    = jindex[iidx];
 392         j_index_end      = jindex[iidx+1];
 393
 394         /* Get outer coordinate index */
 395         inr              = iinr[iidx];
 396         i_coord_offset   = DIM*inr;
 397
 398         /* Load i particle coords and add shift vector */
 399         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 400
 401         fix0             = _fjsp_setzero_v2r8();
 402         fiy0             = _fjsp_setzero_v2r8();
 403         fiz0             = _fjsp_setzero_v2r8();
 404
 405         /* Load parameters for i particles */
 406         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 407         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 408
 409         /* Start inner kernel loop */
 410         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 411         {
 412
 413             /* Get j neighbor index, and coordinate index */
 414             jnrA             = jjnr[jidx];
 415             jnrB             = jjnr[jidx+1];
 416             j_coord_offsetA  = DIM*jnrA;
 417             j_coord_offsetB  = DIM*jnrB;
 418
 419             /* load j atom coordinates */
 420             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 421                                               &jx0,&jy0,&jz0);
 422
 423             /* Calculate displacement vector */
 424             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 425             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 426             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 427
 428             /* Calculate squared distance and things based on it */
 429             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 430
 431             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 432
 433             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 434
 435             /* Load parameters for j particles */
 436             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 437             vdwjidx0A        = 2*vdwtype[jnrA+0];
 438             vdwjidx0B        = 2*vdwtype[jnrB+0];
 439
 440             /**************************
 441              * CALCULATE INTERACTIONS *
 442              **************************/
 443
 444             /* Compute parameters for interactions between i and j atoms */
 445             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 446             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 447                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 448
 449             /* COULOMB ELECTROSTATICS */
 450             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 451             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 452
 453             /* LENNARD-JONES DISPERSION/REPULSION */
 454
 455             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 456             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 457
 458             fscal            = _fjsp_add_v2r8(felec,fvdw);
 459
 460             /* Update vectorial force */
 461             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 462             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 463             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 464
 465             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 466
 467             /* Inner loop uses 37 flops */
 468         }
 469
 470         if(jidx<j_index_end)
 471         {
 472
 473             jnrA             = jjnr[jidx];
 474             j_coord_offsetA  = DIM*jnrA;
 475
 476             /* load j atom coordinates */
 477             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 478                                               &jx0,&jy0,&jz0);
 479
 480             /* Calculate displacement vector */
 481             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 482             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 483             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 484
 485             /* Calculate squared distance and things based on it */
 486             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 487
 488             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 489
 490             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 491
 492             /* Load parameters for j particles */
 493             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 494             vdwjidx0A        = 2*vdwtype[jnrA+0];
 495
 496             /**************************
 497              * CALCULATE INTERACTIONS *
 498              **************************/
 499
 500             /* Compute parameters for interactions between i and j atoms */
 501             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 502             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 503                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 504
 505             /* COULOMB ELECTROSTATICS */
 506             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 507             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 508
 509             /* LENNARD-JONES DISPERSION/REPULSION */
 510
 511             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 512             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 513
 514             fscal            = _fjsp_add_v2r8(felec,fvdw);
 515
 516             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 517
 518             /* Update vectorial force */
 519             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 520             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 521             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 522
 523             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 524
 525             /* Inner loop uses 37 flops */
 526         }
 527
 528         /* End of innermost loop */
 529
 530         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 531                                               f+i_coord_offset,fshift+i_shift_offset);
 532
 533         /* Increment number of inner iterations */
 534         inneriter                  += j_index_end - j_index_start;
 535
 536         /* Outer loop uses 7 flops */
 537     }
 538
 539     /* Increment number of outer iterations */
 540     outeriter        += nri;
 541
 542     /* Update outer/inner flops */
 543
 544     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 545 }