src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecNone_VdwLJSh_GeomP1P1_sse4_1_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 sse4_1_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_x86_sse4_1_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sse4_1_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_sse4_1_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 SSE double precision, 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     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwjidx0A,vdwjidx0B;
  82     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  83     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  84     int              nvdwtype;
  85     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  86     int              *vdwtype;
  87     real             *vdwparam;
  88     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  89     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  90     __m128d          dummy_mask,cutoff_mask;
  91     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  92     __m128d          one     = _mm_set1_pd(1.0);
  93     __m128d          two     = _mm_set1_pd(2.0);
  94     x                = xx[0];
  95     f                = ff[0];
  96
  97     nri              = nlist->nri;
  98     iinr             = nlist->iinr;
  99     jindex           = nlist->jindex;
 100     jjnr             = nlist->jjnr;
 101     shiftidx         = nlist->shift;
 102     gid              = nlist->gid;
 103     shiftvec         = fr->shift_vec[0];
 104     fshift           = fr->fshift[0];
 105     nvdwtype         = fr->ntype;
 106     vdwparam         = fr->nbfp;
 107     vdwtype          = mdatoms->typeA;
 108
 109     rcutoff_scalar   = fr->ic->rvdw;
 110     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 111     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 112
 113     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 114     rvdw             = _mm_set1_pd(fr->ic->rvdw);
 115
 116     /* Avoid stupid compiler warnings */
 117     jnrA = jnrB = 0;
 118     j_coord_offsetA = 0;
 119     j_coord_offsetB = 0;
 120
 121     outeriter        = 0;
 122     inneriter        = 0;
 123
 124     /* Start outer loop over neighborlists */
 125     for(iidx=0; iidx<nri; iidx++)
 126     {
 127         /* Load shift vector for this list */
 128         i_shift_offset   = DIM*shiftidx[iidx];
 129
 130         /* Load limits for loop over neighbors */
 131         j_index_start    = jindex[iidx];
 132         j_index_end      = jindex[iidx+1];
 133
 134         /* Get outer coordinate index */
 135         inr              = iinr[iidx];
 136         i_coord_offset   = DIM*inr;
 137
 138         /* Load i particle coords and add shift vector */
 139         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 140
 141         fix0             = _mm_setzero_pd();
 142         fiy0             = _mm_setzero_pd();
 143         fiz0             = _mm_setzero_pd();
 144
 145         /* Load parameters for i particles */
 146         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 147
 148         /* Reset potential sums */
 149         vvdwsum          = _mm_setzero_pd();
 150
 151         /* Start inner kernel loop */
 152         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 153         {
 154
 155             /* Get j neighbor index, and coordinate index */
 156             jnrA             = jjnr[jidx];
 157             jnrB             = jjnr[jidx+1];
 158             j_coord_offsetA  = DIM*jnrA;
 159             j_coord_offsetB  = DIM*jnrB;
 160
 161             /* load j atom coordinates */
 162             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 163                                               &jx0,&jy0,&jz0);
 164
 165             /* Calculate displacement vector */
 166             dx00             = _mm_sub_pd(ix0,jx0);
 167             dy00             = _mm_sub_pd(iy0,jy0);
 168             dz00             = _mm_sub_pd(iz0,jz0);
 169
 170             /* Calculate squared distance and things based on it */
 171             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 172
 173             rinvsq00         = sse41_inv_d(rsq00);
 174
 175             /* Load parameters for j particles */
 176             vdwjidx0A        = 2*vdwtype[jnrA+0];
 177             vdwjidx0B        = 2*vdwtype[jnrB+0];
 178
 179             /**************************
 180              * CALCULATE INTERACTIONS *
 181              **************************/
 182
 183             if (gmx_mm_any_lt(rsq00,rcutoff2))
 184             {
 185
 186             /* Compute parameters for interactions between i and j atoms */
 187             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 188                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 189
 190             /* LENNARD-JONES DISPERSION/REPULSION */
 191
 192             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 193             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 194             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 195             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) ,
 196                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 197             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 198
 199             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 200
 201             /* Update potential sum for this i atom from the interaction with this j atom. */
 202             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 203             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 204
 205             fscal            = fvdw;
 206
 207             fscal            = _mm_and_pd(fscal,cutoff_mask);
 208
 209             /* Calculate temporary vectorial force */
 210             tx               = _mm_mul_pd(fscal,dx00);
 211             ty               = _mm_mul_pd(fscal,dy00);
 212             tz               = _mm_mul_pd(fscal,dz00);
 213
 214             /* Update vectorial force */
 215             fix0             = _mm_add_pd(fix0,tx);
 216             fiy0             = _mm_add_pd(fiy0,ty);
 217             fiz0             = _mm_add_pd(fiz0,tz);
 218
 219             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 220
 221             }
 222
 223             /* Inner loop uses 41 flops */
 224         }
 225
 226         if(jidx<j_index_end)
 227         {
 228
 229             jnrA             = jjnr[jidx];
 230             j_coord_offsetA  = DIM*jnrA;
 231
 232             /* load j atom coordinates */
 233             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 234                                               &jx0,&jy0,&jz0);
 235
 236             /* Calculate displacement vector */
 237             dx00             = _mm_sub_pd(ix0,jx0);
 238             dy00             = _mm_sub_pd(iy0,jy0);
 239             dz00             = _mm_sub_pd(iz0,jz0);
 240
 241             /* Calculate squared distance and things based on it */
 242             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 243
 244             rinvsq00         = sse41_inv_d(rsq00);
 245
 246             /* Load parameters for j particles */
 247             vdwjidx0A        = 2*vdwtype[jnrA+0];
 248
 249             /**************************
 250              * CALCULATE INTERACTIONS *
 251              **************************/
 252
 253             if (gmx_mm_any_lt(rsq00,rcutoff2))
 254             {
 255
 256             /* Compute parameters for interactions between i and j atoms */
 257             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 258
 259             /* LENNARD-JONES DISPERSION/REPULSION */
 260
 261             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 262             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 263             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 264             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) ,
 265                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 266             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 267
 268             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 269
 270             /* Update potential sum for this i atom from the interaction with this j atom. */
 271             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 272             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 273             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 274
 275             fscal            = fvdw;
 276
 277             fscal            = _mm_and_pd(fscal,cutoff_mask);
 278
 279             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 280
 281             /* Calculate temporary vectorial force */
 282             tx               = _mm_mul_pd(fscal,dx00);
 283             ty               = _mm_mul_pd(fscal,dy00);
 284             tz               = _mm_mul_pd(fscal,dz00);
 285
 286             /* Update vectorial force */
 287             fix0             = _mm_add_pd(fix0,tx);
 288             fiy0             = _mm_add_pd(fiy0,ty);
 289             fiz0             = _mm_add_pd(fiz0,tz);
 290
 291             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 292
 293             }
 294
 295             /* Inner loop uses 41 flops */
 296         }
 297
 298         /* End of innermost loop */
 299
 300         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 301                                               f+i_coord_offset,fshift+i_shift_offset);
 302
 303         ggid                        = gid[iidx];
 304         /* Update potential energies */
 305         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 306
 307         /* Increment number of inner iterations */
 308         inneriter                  += j_index_end - j_index_start;
 309
 310         /* Outer loop uses 7 flops */
 311     }
 312
 313     /* Increment number of outer iterations */
 314     outeriter        += nri;
 315
 316     /* Update outer/inner flops */
 317
 318     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*41);
 319 }
 320 /*
 321  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 322  * Electrostatics interaction: None
 323  * VdW interaction:            LennardJones
 324  * Geometry:                   Particle-Particle
 325  * Calculate force/pot:        Force
 326  */
 327 void
 328 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 329                     (t_nblist                    * gmx_restrict       nlist,
 330                      rvec                        * gmx_restrict          xx,
 331                      rvec                        * gmx_restrict          ff,
 332                      struct t_forcerec           * gmx_restrict          fr,
 333                      t_mdatoms                   * gmx_restrict     mdatoms,
 334                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 335                      t_nrnb                      * gmx_restrict        nrnb)
 336 {
 337     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 338      * just 0 for non-waters.
 339      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 340      * jnr indices corresponding to data put in the four positions in the SIMD register.
 341      */
 342     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 343     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 344     int              jnrA,jnrB;
 345     int              j_coord_offsetA,j_coord_offsetB;
 346     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 347     real             rcutoff_scalar;
 348     real             *shiftvec,*fshift,*x,*f;
 349     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 350     int              vdwioffset0;
 351     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 352     int              vdwjidx0A,vdwjidx0B;
 353     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 354     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 355     int              nvdwtype;
 356     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 357     int              *vdwtype;
 358     real             *vdwparam;
 359     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 360     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 361     __m128d          dummy_mask,cutoff_mask;
 362     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 363     __m128d          one     = _mm_set1_pd(1.0);
 364     __m128d          two     = _mm_set1_pd(2.0);
 365     x                = xx[0];
 366     f                = ff[0];
 367
 368     nri              = nlist->nri;
 369     iinr             = nlist->iinr;
 370     jindex           = nlist->jindex;
 371     jjnr             = nlist->jjnr;
 372     shiftidx         = nlist->shift;
 373     gid              = nlist->gid;
 374     shiftvec         = fr->shift_vec[0];
 375     fshift           = fr->fshift[0];
 376     nvdwtype         = fr->ntype;
 377     vdwparam         = fr->nbfp;
 378     vdwtype          = mdatoms->typeA;
 379
 380     rcutoff_scalar   = fr->ic->rvdw;
 381     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 382     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 383
 384     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 385     rvdw             = _mm_set1_pd(fr->ic->rvdw);
 386
 387     /* Avoid stupid compiler warnings */
 388     jnrA = jnrB = 0;
 389     j_coord_offsetA = 0;
 390     j_coord_offsetB = 0;
 391
 392     outeriter        = 0;
 393     inneriter        = 0;
 394
 395     /* Start outer loop over neighborlists */
 396     for(iidx=0; iidx<nri; iidx++)
 397     {
 398         /* Load shift vector for this list */
 399         i_shift_offset   = DIM*shiftidx[iidx];
 400
 401         /* Load limits for loop over neighbors */
 402         j_index_start    = jindex[iidx];
 403         j_index_end      = jindex[iidx+1];
 404
 405         /* Get outer coordinate index */
 406         inr              = iinr[iidx];
 407         i_coord_offset   = DIM*inr;
 408
 409         /* Load i particle coords and add shift vector */
 410         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 411
 412         fix0             = _mm_setzero_pd();
 413         fiy0             = _mm_setzero_pd();
 414         fiz0             = _mm_setzero_pd();
 415
 416         /* Load parameters for i particles */
 417         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 418
 419         /* Start inner kernel loop */
 420         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 421         {
 422
 423             /* Get j neighbor index, and coordinate index */
 424             jnrA             = jjnr[jidx];
 425             jnrB             = jjnr[jidx+1];
 426             j_coord_offsetA  = DIM*jnrA;
 427             j_coord_offsetB  = DIM*jnrB;
 428
 429             /* load j atom coordinates */
 430             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 431                                               &jx0,&jy0,&jz0);
 432
 433             /* Calculate displacement vector */
 434             dx00             = _mm_sub_pd(ix0,jx0);
 435             dy00             = _mm_sub_pd(iy0,jy0);
 436             dz00             = _mm_sub_pd(iz0,jz0);
 437
 438             /* Calculate squared distance and things based on it */
 439             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 440
 441             rinvsq00         = sse41_inv_d(rsq00);
 442
 443             /* Load parameters for j particles */
 444             vdwjidx0A        = 2*vdwtype[jnrA+0];
 445             vdwjidx0B        = 2*vdwtype[jnrB+0];
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             if (gmx_mm_any_lt(rsq00,rcutoff2))
 452             {
 453
 454             /* Compute parameters for interactions between i and j atoms */
 455             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 456                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 457
 458             /* LENNARD-JONES DISPERSION/REPULSION */
 459
 460             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 461             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 462
 463             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 464
 465             fscal            = fvdw;
 466
 467             fscal            = _mm_and_pd(fscal,cutoff_mask);
 468
 469             /* Calculate temporary vectorial force */
 470             tx               = _mm_mul_pd(fscal,dx00);
 471             ty               = _mm_mul_pd(fscal,dy00);
 472             tz               = _mm_mul_pd(fscal,dz00);
 473
 474             /* Update vectorial force */
 475             fix0             = _mm_add_pd(fix0,tx);
 476             fiy0             = _mm_add_pd(fiy0,ty);
 477             fiz0             = _mm_add_pd(fiz0,tz);
 478
 479             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 480
 481             }
 482
 483             /* Inner loop uses 30 flops */
 484         }
 485
 486         if(jidx<j_index_end)
 487         {
 488
 489             jnrA             = jjnr[jidx];
 490             j_coord_offsetA  = DIM*jnrA;
 491
 492             /* load j atom coordinates */
 493             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 494                                               &jx0,&jy0,&jz0);
 495
 496             /* Calculate displacement vector */
 497             dx00             = _mm_sub_pd(ix0,jx0);
 498             dy00             = _mm_sub_pd(iy0,jy0);
 499             dz00             = _mm_sub_pd(iz0,jz0);
 500
 501             /* Calculate squared distance and things based on it */
 502             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 503
 504             rinvsq00         = sse41_inv_d(rsq00);
 505
 506             /* Load parameters for j particles */
 507             vdwjidx0A        = 2*vdwtype[jnrA+0];
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             if (gmx_mm_any_lt(rsq00,rcutoff2))
 514             {
 515
 516             /* Compute parameters for interactions between i and j atoms */
 517             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 518
 519             /* LENNARD-JONES DISPERSION/REPULSION */
 520
 521             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 522             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 523
 524             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 525
 526             fscal            = fvdw;
 527
 528             fscal            = _mm_and_pd(fscal,cutoff_mask);
 529
 530             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm_mul_pd(fscal,dx00);
 534             ty               = _mm_mul_pd(fscal,dy00);
 535             tz               = _mm_mul_pd(fscal,dz00);
 536
 537             /* Update vectorial force */
 538             fix0             = _mm_add_pd(fix0,tx);
 539             fiy0             = _mm_add_pd(fiy0,ty);
 540             fiz0             = _mm_add_pd(fiz0,tz);
 541
 542             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 543
 544             }
 545
 546             /* Inner loop uses 30 flops */
 547         }
 548
 549         /* End of innermost loop */
 550
 551         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 552                                               f+i_coord_offset,fshift+i_shift_offset);
 553
 554         /* Increment number of inner iterations */
 555         inneriter                  += j_index_end - j_index_start;
 556
 557         /* Outer loop uses 6 flops */
 558     }
 559
 560     /* Increment number of outer iterations */
 561     outeriter        += nri;
 562
 563     /* Update outer/inner flops */
 564
 565     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*30);
 566 }