src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecNone_VdwLJSh_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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: None
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              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     int              nvdwtype;
  85     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  86     int              *vdwtype;
  87     real             *vdwparam;
  88     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  89     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  90     _fjsp_v2r8       itab_tmp;
  91     _fjsp_v2r8       dummy_mask,cutoff_mask;
  92     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  93     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  94     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  95
  96     x                = xx[0];
  97     f                = ff[0];
  98
  99     nri              = nlist->nri;
 100     iinr             = nlist->iinr;
 101     jindex           = nlist->jindex;
 102     jjnr             = nlist->jjnr;
 103     shiftidx         = nlist->shift;
 104     gid              = nlist->gid;
 105     shiftvec         = fr->shift_vec[0];
 106     fshift           = fr->fshift[0];
 107     nvdwtype         = fr->ntype;
 108     vdwparam         = fr->nbfp;
 109     vdwtype          = mdatoms->typeA;
 110
 111     rcutoff_scalar   = fr->rvdw;
 112     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 113     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 114
 115     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 116     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 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         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 149
 150         /* Reset potential sums */
 151         vvdwsum          = _fjsp_setzero_v2r8();
 152
 153         /* Start inner kernel loop */
 154         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 155         {
 156
 157             /* Get j neighbor index, and coordinate index */
 158             jnrA             = jjnr[jidx];
 159             jnrB             = jjnr[jidx+1];
 160             j_coord_offsetA  = DIM*jnrA;
 161             j_coord_offsetB  = DIM*jnrB;
 162
 163             /* load j atom coordinates */
 164             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 165                                               &jx0,&jy0,&jz0);
 166
 167             /* Calculate displacement vector */
 168             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 169             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 170             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 171
 172             /* Calculate squared distance and things based on it */
 173             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 174
 175             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 176
 177             /* Load parameters for j particles */
 178             vdwjidx0A        = 2*vdwtype[jnrA+0];
 179             vdwjidx0B        = 2*vdwtype[jnrB+0];
 180
 181             /**************************
 182              * CALCULATE INTERACTIONS *
 183              **************************/
 184
 185             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 186             {
 187
 188             /* Compute parameters for interactions between i and j atoms */
 189             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 190                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 191
 192             /* LENNARD-JONES DISPERSION/REPULSION */
 193
 194             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 195             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 196             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 197             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 198                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 199             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 200
 201             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 202
 203             /* Update potential sum for this i atom from the interaction with this j atom. */
 204             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 205             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 206
 207             fscal            = fvdw;
 208
 209             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 210
 211             /* Update vectorial force */
 212             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 213             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 214             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 215
 216             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 217
 218             }
 219
 220             /* Inner loop uses 44 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             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 242
 243             /* Load parameters for j particles */
 244             vdwjidx0A        = 2*vdwtype[jnrA+0];
 245
 246             /**************************
 247              * CALCULATE INTERACTIONS *
 248              **************************/
 249
 250             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 251             {
 252
 253             /* Compute parameters for interactions between i and j atoms */
 254             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 255
 256             /* LENNARD-JONES DISPERSION/REPULSION */
 257
 258             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 259             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 260             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 261             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 262                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 263             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 264
 265             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 266
 267             /* Update potential sum for this i atom from the interaction with this j atom. */
 268             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 269             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 270             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 271
 272             fscal            = fvdw;
 273
 274             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 275
 276             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 277
 278             /* Update vectorial force */
 279             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 280             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 281             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 282
 283             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 284
 285             }
 286
 287             /* Inner loop uses 44 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(vvdwsum,kernel_data->energygrp_vdw+ggid);
 298
 299         /* Increment number of inner iterations */
 300         inneriter                  += j_index_end - j_index_start;
 301
 302         /* Outer loop uses 7 flops */
 303     }
 304
 305     /* Increment number of outer iterations */
 306     outeriter        += nri;
 307
 308     /* Update outer/inner flops */
 309
 310     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*44);
 311 }
 312 /*
 313  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 314  * Electrostatics interaction: None
 315  * VdW interaction:            LennardJones
 316  * Geometry:                   Particle-Particle
 317  * Calculate force/pot:        Force
 318  */
 319 void
 320 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 321                     (t_nblist                    * gmx_restrict       nlist,
 322                      rvec                        * gmx_restrict          xx,
 323                      rvec                        * gmx_restrict          ff,
 324                      t_forcerec                  * gmx_restrict          fr,
 325                      t_mdatoms                   * gmx_restrict     mdatoms,
 326                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 327                      t_nrnb                      * gmx_restrict        nrnb)
 328 {
 329     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 330      * just 0 for non-waters.
 331      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 332      * jnr indices corresponding to data put in the four positions in the SIMD register.
 333      */
 334     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 335     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 336     int              jnrA,jnrB;
 337     int              j_coord_offsetA,j_coord_offsetB;
 338     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 339     real             rcutoff_scalar;
 340     real             *shiftvec,*fshift,*x,*f;
 341     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 342     int              vdwioffset0;
 343     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 344     int              vdwjidx0A,vdwjidx0B;
 345     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 346     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 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     nvdwtype         = fr->ntype;
 371     vdwparam         = fr->nbfp;
 372     vdwtype          = mdatoms->typeA;
 373
 374     rcutoff_scalar   = fr->rvdw;
 375     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 376     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 377
 378     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 379     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 380
 381     /* Avoid stupid compiler warnings */
 382     jnrA = jnrB = 0;
 383     j_coord_offsetA = 0;
 384     j_coord_offsetB = 0;
 385
 386     outeriter        = 0;
 387     inneriter        = 0;
 388
 389     /* Start outer loop over neighborlists */
 390     for(iidx=0; iidx<nri; iidx++)
 391     {
 392         /* Load shift vector for this list */
 393         i_shift_offset   = DIM*shiftidx[iidx];
 394
 395         /* Load limits for loop over neighbors */
 396         j_index_start    = jindex[iidx];
 397         j_index_end      = jindex[iidx+1];
 398
 399         /* Get outer coordinate index */
 400         inr              = iinr[iidx];
 401         i_coord_offset   = DIM*inr;
 402
 403         /* Load i particle coords and add shift vector */
 404         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 405
 406         fix0             = _fjsp_setzero_v2r8();
 407         fiy0             = _fjsp_setzero_v2r8();
 408         fiz0             = _fjsp_setzero_v2r8();
 409
 410         /* Load parameters for i particles */
 411         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 412
 413         /* Start inner kernel loop */
 414         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 415         {
 416
 417             /* Get j neighbor index, and coordinate index */
 418             jnrA             = jjnr[jidx];
 419             jnrB             = jjnr[jidx+1];
 420             j_coord_offsetA  = DIM*jnrA;
 421             j_coord_offsetB  = DIM*jnrB;
 422
 423             /* load j atom coordinates */
 424             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 425                                               &jx0,&jy0,&jz0);
 426
 427             /* Calculate displacement vector */
 428             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 429             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 430             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 431
 432             /* Calculate squared distance and things based on it */
 433             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 434
 435             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 436
 437             /* Load parameters for j particles */
 438             vdwjidx0A        = 2*vdwtype[jnrA+0];
 439             vdwjidx0B        = 2*vdwtype[jnrB+0];
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 446             {
 447
 448             /* Compute parameters for interactions between i and j atoms */
 449             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 450                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 451
 452             /* LENNARD-JONES DISPERSION/REPULSION */
 453
 454             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 455             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 456
 457             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 458
 459             fscal            = fvdw;
 460
 461             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 462
 463             /* Update vectorial force */
 464             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 465             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 466             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 467
 468             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 469
 470             }
 471
 472             /* Inner loop uses 33 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             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 494
 495             /* Load parameters for j particles */
 496             vdwjidx0A        = 2*vdwtype[jnrA+0];
 497
 498             /**************************
 499              * CALCULATE INTERACTIONS *
 500              **************************/
 501
 502             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 503             {
 504
 505             /* Compute parameters for interactions between i and j atoms */
 506             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 507
 508             /* LENNARD-JONES DISPERSION/REPULSION */
 509
 510             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 511             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 512
 513             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 514
 515             fscal            = fvdw;
 516
 517             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 518
 519             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 520
 521             /* Update vectorial force */
 522             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 523             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 524             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 525
 526             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 527
 528             }
 529
 530             /* Inner loop uses 33 flops */
 531         }
 532
 533         /* End of innermost loop */
 534
 535         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 536                                               f+i_coord_offset,fshift+i_shift_offset);
 537
 538         /* Increment number of inner iterations */
 539         inneriter                  += j_index_end - j_index_start;
 540
 541         /* Outer loop uses 6 flops */
 542     }
 543
 544     /* Increment number of outer iterations */
 545     outeriter        += nri;
 546
 547     /* Update outer/inner flops */
 548
 549     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*33);
 550 }