src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecNone_VdwLJSh_GeomP1P1_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 sse4_1_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sse4_1_double
  54  * Electrostatics interaction: None
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Particle-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sse4_1_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwjidx0A,vdwjidx0B;
  85     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  86     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  87     int              nvdwtype;
  88     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  89     int              *vdwtype;
  90     real             *vdwparam;
  91     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  92     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  93     __m128d          dummy_mask,cutoff_mask;
  94     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  95     __m128d          one     = _mm_set1_pd(1.0);
  96     __m128d          two     = _mm_set1_pd(2.0);
  97     x                = xx[0];
  98     f                = ff[0];
  99
 100     nri              = nlist->nri;
 101     iinr             = nlist->iinr;
 102     jindex           = nlist->jindex;
 103     jjnr             = nlist->jjnr;
 104     shiftidx         = nlist->shift;
 105     gid              = nlist->gid;
 106     shiftvec         = fr->shift_vec[0];
 107     fshift           = fr->fshift[0];
 108     nvdwtype         = fr->ntype;
 109     vdwparam         = fr->nbfp;
 110     vdwtype          = mdatoms->typeA;
 111
 112     rcutoff_scalar   = fr->rvdw;
 113     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 114     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 115
 116     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 117     rvdw             = _mm_set1_pd(fr->rvdw);
 118
 119     /* Avoid stupid compiler warnings */
 120     jnrA = jnrB = 0;
 121     j_coord_offsetA = 0;
 122     j_coord_offsetB = 0;
 123
 124     outeriter        = 0;
 125     inneriter        = 0;
 126
 127     /* Start outer loop over neighborlists */
 128     for(iidx=0; iidx<nri; iidx++)
 129     {
 130         /* Load shift vector for this list */
 131         i_shift_offset   = DIM*shiftidx[iidx];
 132
 133         /* Load limits for loop over neighbors */
 134         j_index_start    = jindex[iidx];
 135         j_index_end      = jindex[iidx+1];
 136
 137         /* Get outer coordinate index */
 138         inr              = iinr[iidx];
 139         i_coord_offset   = DIM*inr;
 140
 141         /* Load i particle coords and add shift vector */
 142         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 143
 144         fix0             = _mm_setzero_pd();
 145         fiy0             = _mm_setzero_pd();
 146         fiz0             = _mm_setzero_pd();
 147
 148         /* Load parameters for i particles */
 149         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 150
 151         /* Reset potential sums */
 152         vvdwsum          = _mm_setzero_pd();
 153
 154         /* Start inner kernel loop */
 155         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 156         {
 157
 158             /* Get j neighbor index, and coordinate index */
 159             jnrA             = jjnr[jidx];
 160             jnrB             = jjnr[jidx+1];
 161             j_coord_offsetA  = DIM*jnrA;
 162             j_coord_offsetB  = DIM*jnrB;
 163
 164             /* load j atom coordinates */
 165             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 166                                               &jx0,&jy0,&jz0);
 167
 168             /* Calculate displacement vector */
 169             dx00             = _mm_sub_pd(ix0,jx0);
 170             dy00             = _mm_sub_pd(iy0,jy0);
 171             dz00             = _mm_sub_pd(iz0,jz0);
 172
 173             /* Calculate squared distance and things based on it */
 174             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 175
 176             rinvsq00         = gmx_mm_inv_pd(rsq00);
 177
 178             /* Load parameters for j particles */
 179             vdwjidx0A        = 2*vdwtype[jnrA+0];
 180             vdwjidx0B        = 2*vdwtype[jnrB+0];
 181
 182             /**************************
 183              * CALCULATE INTERACTIONS *
 184              **************************/
 185
 186             if (gmx_mm_any_lt(rsq00,rcutoff2))
 187             {
 188
 189             /* Compute parameters for interactions between i and j atoms */
 190             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 191                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 192
 193             /* LENNARD-JONES DISPERSION/REPULSION */
 194
 195             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 196             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 197             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 198             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 199                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 200             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 201
 202             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 203
 204             /* Update potential sum for this i atom from the interaction with this j atom. */
 205             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 206             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 207
 208             fscal            = fvdw;
 209
 210             fscal            = _mm_and_pd(fscal,cutoff_mask);
 211
 212             /* Calculate temporary vectorial force */
 213             tx               = _mm_mul_pd(fscal,dx00);
 214             ty               = _mm_mul_pd(fscal,dy00);
 215             tz               = _mm_mul_pd(fscal,dz00);
 216
 217             /* Update vectorial force */
 218             fix0             = _mm_add_pd(fix0,tx);
 219             fiy0             = _mm_add_pd(fiy0,ty);
 220             fiz0             = _mm_add_pd(fiz0,tz);
 221
 222             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 223
 224             }
 225
 226             /* Inner loop uses 41 flops */
 227         }
 228
 229         if(jidx<j_index_end)
 230         {
 231
 232             jnrA             = jjnr[jidx];
 233             j_coord_offsetA  = DIM*jnrA;
 234
 235             /* load j atom coordinates */
 236             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 237                                               &jx0,&jy0,&jz0);
 238
 239             /* Calculate displacement vector */
 240             dx00             = _mm_sub_pd(ix0,jx0);
 241             dy00             = _mm_sub_pd(iy0,jy0);
 242             dz00             = _mm_sub_pd(iz0,jz0);
 243
 244             /* Calculate squared distance and things based on it */
 245             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 246
 247             rinvsq00         = gmx_mm_inv_pd(rsq00);
 248
 249             /* Load parameters for j particles */
 250             vdwjidx0A        = 2*vdwtype[jnrA+0];
 251
 252             /**************************
 253              * CALCULATE INTERACTIONS *
 254              **************************/
 255
 256             if (gmx_mm_any_lt(rsq00,rcutoff2))
 257             {
 258
 259             /* Compute parameters for interactions between i and j atoms */
 260             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 261
 262             /* LENNARD-JONES DISPERSION/REPULSION */
 263
 264             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 265             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 266             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 267             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 268                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 269             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 270
 271             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 272
 273             /* Update potential sum for this i atom from the interaction with this j atom. */
 274             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 275             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 276             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 277
 278             fscal            = fvdw;
 279
 280             fscal            = _mm_and_pd(fscal,cutoff_mask);
 281
 282             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 283
 284             /* Calculate temporary vectorial force */
 285             tx               = _mm_mul_pd(fscal,dx00);
 286             ty               = _mm_mul_pd(fscal,dy00);
 287             tz               = _mm_mul_pd(fscal,dz00);
 288
 289             /* Update vectorial force */
 290             fix0             = _mm_add_pd(fix0,tx);
 291             fiy0             = _mm_add_pd(fiy0,ty);
 292             fiz0             = _mm_add_pd(fiz0,tz);
 293
 294             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 295
 296             }
 297
 298             /* Inner loop uses 41 flops */
 299         }
 300
 301         /* End of innermost loop */
 302
 303         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 304                                               f+i_coord_offset,fshift+i_shift_offset);
 305
 306         ggid                        = gid[iidx];
 307         /* Update potential energies */
 308         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 309
 310         /* Increment number of inner iterations */
 311         inneriter                  += j_index_end - j_index_start;
 312
 313         /* Outer loop uses 7 flops */
 314     }
 315
 316     /* Increment number of outer iterations */
 317     outeriter        += nri;
 318
 319     /* Update outer/inner flops */
 320
 321     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*41);
 322 }
 323 /*
 324  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 325  * Electrostatics interaction: None
 326  * VdW interaction:            LennardJones
 327  * Geometry:                   Particle-Particle
 328  * Calculate force/pot:        Force
 329  */
 330 void
 331 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 332                     (t_nblist                    * gmx_restrict       nlist,
 333                      rvec                        * gmx_restrict          xx,
 334                      rvec                        * gmx_restrict          ff,
 335                      t_forcerec                  * gmx_restrict          fr,
 336                      t_mdatoms                   * gmx_restrict     mdatoms,
 337                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 338                      t_nrnb                      * gmx_restrict        nrnb)
 339 {
 340     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 341      * just 0 for non-waters.
 342      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 343      * jnr indices corresponding to data put in the four positions in the SIMD register.
 344      */
 345     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 346     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 347     int              jnrA,jnrB;
 348     int              j_coord_offsetA,j_coord_offsetB;
 349     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 350     real             rcutoff_scalar;
 351     real             *shiftvec,*fshift,*x,*f;
 352     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 353     int              vdwioffset0;
 354     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 355     int              vdwjidx0A,vdwjidx0B;
 356     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 357     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 358     int              nvdwtype;
 359     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 360     int              *vdwtype;
 361     real             *vdwparam;
 362     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 363     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 364     __m128d          dummy_mask,cutoff_mask;
 365     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 366     __m128d          one     = _mm_set1_pd(1.0);
 367     __m128d          two     = _mm_set1_pd(2.0);
 368     x                = xx[0];
 369     f                = ff[0];
 370
 371     nri              = nlist->nri;
 372     iinr             = nlist->iinr;
 373     jindex           = nlist->jindex;
 374     jjnr             = nlist->jjnr;
 375     shiftidx         = nlist->shift;
 376     gid              = nlist->gid;
 377     shiftvec         = fr->shift_vec[0];
 378     fshift           = fr->fshift[0];
 379     nvdwtype         = fr->ntype;
 380     vdwparam         = fr->nbfp;
 381     vdwtype          = mdatoms->typeA;
 382
 383     rcutoff_scalar   = fr->rvdw;
 384     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 385     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 386
 387     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 388     rvdw             = _mm_set1_pd(fr->rvdw);
 389
 390     /* Avoid stupid compiler warnings */
 391     jnrA = jnrB = 0;
 392     j_coord_offsetA = 0;
 393     j_coord_offsetB = 0;
 394
 395     outeriter        = 0;
 396     inneriter        = 0;
 397
 398     /* Start outer loop over neighborlists */
 399     for(iidx=0; iidx<nri; iidx++)
 400     {
 401         /* Load shift vector for this list */
 402         i_shift_offset   = DIM*shiftidx[iidx];
 403
 404         /* Load limits for loop over neighbors */
 405         j_index_start    = jindex[iidx];
 406         j_index_end      = jindex[iidx+1];
 407
 408         /* Get outer coordinate index */
 409         inr              = iinr[iidx];
 410         i_coord_offset   = DIM*inr;
 411
 412         /* Load i particle coords and add shift vector */
 413         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 414
 415         fix0             = _mm_setzero_pd();
 416         fiy0             = _mm_setzero_pd();
 417         fiz0             = _mm_setzero_pd();
 418
 419         /* Load parameters for i particles */
 420         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 421
 422         /* Start inner kernel loop */
 423         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 424         {
 425
 426             /* Get j neighbor index, and coordinate index */
 427             jnrA             = jjnr[jidx];
 428             jnrB             = jjnr[jidx+1];
 429             j_coord_offsetA  = DIM*jnrA;
 430             j_coord_offsetB  = DIM*jnrB;
 431
 432             /* load j atom coordinates */
 433             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 434                                               &jx0,&jy0,&jz0);
 435
 436             /* Calculate displacement vector */
 437             dx00             = _mm_sub_pd(ix0,jx0);
 438             dy00             = _mm_sub_pd(iy0,jy0);
 439             dz00             = _mm_sub_pd(iz0,jz0);
 440
 441             /* Calculate squared distance and things based on it */
 442             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 443
 444             rinvsq00         = gmx_mm_inv_pd(rsq00);
 445
 446             /* Load parameters for j particles */
 447             vdwjidx0A        = 2*vdwtype[jnrA+0];
 448             vdwjidx0B        = 2*vdwtype[jnrB+0];
 449
 450             /**************************
 451              * CALCULATE INTERACTIONS *
 452              **************************/
 453
 454             if (gmx_mm_any_lt(rsq00,rcutoff2))
 455             {
 456
 457             /* Compute parameters for interactions between i and j atoms */
 458             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 459                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 460
 461             /* LENNARD-JONES DISPERSION/REPULSION */
 462
 463             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 464             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 465
 466             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 467
 468             fscal            = fvdw;
 469
 470             fscal            = _mm_and_pd(fscal,cutoff_mask);
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm_mul_pd(fscal,dx00);
 474             ty               = _mm_mul_pd(fscal,dy00);
 475             tz               = _mm_mul_pd(fscal,dz00);
 476
 477             /* Update vectorial force */
 478             fix0             = _mm_add_pd(fix0,tx);
 479             fiy0             = _mm_add_pd(fiy0,ty);
 480             fiz0             = _mm_add_pd(fiz0,tz);
 481
 482             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 483
 484             }
 485
 486             /* Inner loop uses 30 flops */
 487         }
 488
 489         if(jidx<j_index_end)
 490         {
 491
 492             jnrA             = jjnr[jidx];
 493             j_coord_offsetA  = DIM*jnrA;
 494
 495             /* load j atom coordinates */
 496             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 497                                               &jx0,&jy0,&jz0);
 498
 499             /* Calculate displacement vector */
 500             dx00             = _mm_sub_pd(ix0,jx0);
 501             dy00             = _mm_sub_pd(iy0,jy0);
 502             dz00             = _mm_sub_pd(iz0,jz0);
 503
 504             /* Calculate squared distance and things based on it */
 505             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 506
 507             rinvsq00         = gmx_mm_inv_pd(rsq00);
 508
 509             /* Load parameters for j particles */
 510             vdwjidx0A        = 2*vdwtype[jnrA+0];
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             if (gmx_mm_any_lt(rsq00,rcutoff2))
 517             {
 518
 519             /* Compute parameters for interactions between i and j atoms */
 520             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 521
 522             /* LENNARD-JONES DISPERSION/REPULSION */
 523
 524             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 525             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 526
 527             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 528
 529             fscal            = fvdw;
 530
 531             fscal            = _mm_and_pd(fscal,cutoff_mask);
 532
 533             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 534
 535             /* Calculate temporary vectorial force */
 536             tx               = _mm_mul_pd(fscal,dx00);
 537             ty               = _mm_mul_pd(fscal,dy00);
 538             tz               = _mm_mul_pd(fscal,dz00);
 539
 540             /* Update vectorial force */
 541             fix0             = _mm_add_pd(fix0,tx);
 542             fiy0             = _mm_add_pd(fiy0,ty);
 543             fiz0             = _mm_add_pd(fiz0,tz);
 544
 545             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 546
 547             }
 548
 549             /* Inner loop uses 30 flops */
 550         }
 551
 552         /* End of innermost loop */
 553
 554         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 555                                               f+i_coord_offset,fshift+i_shift_offset);
 556
 557         /* Increment number of inner iterations */
 558         inneriter                  += j_index_end - j_index_start;
 559
 560         /* Outer loop uses 6 flops */
 561     }
 562
 563     /* Increment number of outer iterations */
 564     outeriter        += nri;
 565
 566     /* Update outer/inner flops */
 567
 568     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*30);
 569 }