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 "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_ElecNone_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: None
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecNone_VdwLJSh_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     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     nvdwtype         = fr->ntype;
 110     vdwparam         = fr->nbfp;
 111     vdwtype          = mdatoms->typeA;
 112
 113     rcutoff_scalar   = fr->rvdw;
 114     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 115     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 116
 117     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 118     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 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         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152         /* Reset potential sums */
 153         vvdwsum          = _fjsp_setzero_v2r8();
 154
 155         /* Start inner kernel loop */
 156         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 157         {
 158
 159             /* Get j neighbor index, and coordinate index */
 160             jnrA             = jjnr[jidx];
 161             jnrB             = jjnr[jidx+1];
 162             j_coord_offsetA  = DIM*jnrA;
 163             j_coord_offsetB  = DIM*jnrB;
 164
 165             /* load j atom coordinates */
 166             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 167                                               &jx0,&jy0,&jz0);
 168
 169             /* Calculate displacement vector */
 170             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 171             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 172             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 173
 174             /* Calculate squared distance and things based on it */
 175             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 176
 177             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 178
 179             /* Load parameters for j particles */
 180             vdwjidx0A        = 2*vdwtype[jnrA+0];
 181             vdwjidx0B        = 2*vdwtype[jnrB+0];
 182
 183             /**************************
 184              * CALCULATE INTERACTIONS *
 185              **************************/
 186
 187             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 188             {
 189
 190             /* Compute parameters for interactions between i and j atoms */
 191             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 192                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 193
 194             /* LENNARD-JONES DISPERSION/REPULSION */
 195
 196             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 197             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 198             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 199             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 200                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 201             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 202
 203             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 204
 205             /* Update potential sum for this i atom from the interaction with this j atom. */
 206             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 207             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 208
 209             fscal            = fvdw;
 210
 211             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 212
 213             /* Update vectorial force */
 214             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 215             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 216             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 217
 218             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 219
 220             }
 221
 222             /* Inner loop uses 44 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             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 244
 245             /* Load parameters for j particles */
 246             vdwjidx0A        = 2*vdwtype[jnrA+0];
 247
 248             /**************************
 249              * CALCULATE INTERACTIONS *
 250              **************************/
 251
 252             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 253             {
 254
 255             /* Compute parameters for interactions between i and j atoms */
 256             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 257                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 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(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 265                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 266             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 267
 268             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 269
 270             /* Update potential sum for this i atom from the interaction with this j atom. */
 271             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 272             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 273             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 274
 275             fscal            = fvdw;
 276
 277             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 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             }
 289
 290             /* Inner loop uses 44 flops */
 291         }
 292
 293         /* End of innermost loop */
 294
 295         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 296                                               f+i_coord_offset,fshift+i_shift_offset);
 297
 298         ggid                        = gid[iidx];
 299         /* Update potential energies */
 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 7 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_VDW_VF,outeriter*7 + inneriter*44);
 314 }
 315 /*
 316  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 317  * Electrostatics interaction: None
 318  * VdW interaction:            LennardJones
 319  * Geometry:                   Particle-Particle
 320  * Calculate force/pot:        Force
 321  */
 322 void
 323 nb_kernel_ElecNone_VdwLJSh_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     int              nvdwtype;
 351     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 352     int              *vdwtype;
 353     real             *vdwparam;
 354     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 355     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 356     _fjsp_v2r8       itab_tmp;
 357     _fjsp_v2r8       dummy_mask,cutoff_mask;
 358     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 359     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 360     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 361
 362     x                = xx[0];
 363     f                = ff[0];
 364
 365     nri              = nlist->nri;
 366     iinr             = nlist->iinr;
 367     jindex           = nlist->jindex;
 368     jjnr             = nlist->jjnr;
 369     shiftidx         = nlist->shift;
 370     gid              = nlist->gid;
 371     shiftvec         = fr->shift_vec[0];
 372     fshift           = fr->fshift[0];
 373     nvdwtype         = fr->ntype;
 374     vdwparam         = fr->nbfp;
 375     vdwtype          = mdatoms->typeA;
 376
 377     rcutoff_scalar   = fr->rvdw;
 378     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 379     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 380
 381     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 382     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 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         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             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 439
 440             /* Load parameters for j particles */
 441             vdwjidx0A        = 2*vdwtype[jnrA+0];
 442             vdwjidx0B        = 2*vdwtype[jnrB+0];
 443
 444             /**************************
 445              * CALCULATE INTERACTIONS *
 446              **************************/
 447
 448             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 449             {
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 453                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 454
 455             /* LENNARD-JONES DISPERSION/REPULSION */
 456
 457             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 458             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 459
 460             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 461
 462             fscal            = fvdw;
 463
 464             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 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             }
 474
 475             /* Inner loop uses 33 flops */
 476         }
 477
 478         if(jidx<j_index_end)
 479         {
 480
 481             jnrA             = jjnr[jidx];
 482             j_coord_offsetA  = DIM*jnrA;
 483
 484             /* load j atom coordinates */
 485             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 486                                               &jx0,&jy0,&jz0);
 487
 488             /* Calculate displacement vector */
 489             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 490             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 491             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 492
 493             /* Calculate squared distance and things based on it */
 494             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 495
 496             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 497
 498             /* Load parameters for j particles */
 499             vdwjidx0A        = 2*vdwtype[jnrA+0];
 500
 501             /**************************
 502              * CALCULATE INTERACTIONS *
 503              **************************/
 504
 505             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 506             {
 507
 508             /* Compute parameters for interactions between i and j atoms */
 509             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 510                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 511
 512             /* LENNARD-JONES DISPERSION/REPULSION */
 513
 514             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 515             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 516
 517             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 518
 519             fscal            = fvdw;
 520
 521             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 522
 523             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 524
 525             /* Update vectorial force */
 526             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 527             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 528             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 529
 530             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 531
 532             }
 533
 534             /* Inner loop uses 33 flops */
 535         }
 536
 537         /* End of innermost loop */
 538
 539         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 540                                               f+i_coord_offset,fshift+i_shift_offset);
 541
 542         /* Increment number of inner iterations */
 543         inneriter                  += j_index_end - j_index_start;
 544
 545         /* Outer loop uses 6 flops */
 546     }
 547
 548     /* Increment number of outer iterations */
 549     outeriter        += nri;
 550
 551     /* Update outer/inner flops */
 552
 553     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*33);
 554 }