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) 2012,2013,2014, 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 sse4_1_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 "gromacs/simd/math_x86_sse4_1_double.h"
  48 #include "kernelutil_x86_sse4_1_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sse4_1_double
  52  * Electrostatics interaction: None
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_sse4_1_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwjidx0A,vdwjidx0B;
  83     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  84     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     int              nvdwtype;
  86     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  87     int              *vdwtype;
  88     real             *vdwparam;
  89     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  90     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  91     __m128d          dummy_mask,cutoff_mask;
  92     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  93     __m128d          one     = _mm_set1_pd(1.0);
  94     __m128d          two     = _mm_set1_pd(2.0);
  95     x                = xx[0];
  96     f                = ff[0];
  97
  98     nri              = nlist->nri;
  99     iinr             = nlist->iinr;
 100     jindex           = nlist->jindex;
 101     jjnr             = nlist->jjnr;
 102     shiftidx         = nlist->shift;
 103     gid              = nlist->gid;
 104     shiftvec         = fr->shift_vec[0];
 105     fshift           = fr->fshift[0];
 106     nvdwtype         = fr->ntype;
 107     vdwparam         = fr->nbfp;
 108     vdwtype          = mdatoms->typeA;
 109
 110     rcutoff_scalar   = fr->rvdw;
 111     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 112     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 113
 114     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 115     rvdw             = _mm_set1_pd(fr->rvdw);
 116
 117     /* Avoid stupid compiler warnings */
 118     jnrA = jnrB = 0;
 119     j_coord_offsetA = 0;
 120     j_coord_offsetB = 0;
 121
 122     outeriter        = 0;
 123     inneriter        = 0;
 124
 125     /* Start outer loop over neighborlists */
 126     for(iidx=0; iidx<nri; iidx++)
 127     {
 128         /* Load shift vector for this list */
 129         i_shift_offset   = DIM*shiftidx[iidx];
 130
 131         /* Load limits for loop over neighbors */
 132         j_index_start    = jindex[iidx];
 133         j_index_end      = jindex[iidx+1];
 134
 135         /* Get outer coordinate index */
 136         inr              = iinr[iidx];
 137         i_coord_offset   = DIM*inr;
 138
 139         /* Load i particle coords and add shift vector */
 140         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 141
 142         fix0             = _mm_setzero_pd();
 143         fiy0             = _mm_setzero_pd();
 144         fiz0             = _mm_setzero_pd();
 145
 146         /* Load parameters for i particles */
 147         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 148
 149         /* Reset potential sums */
 150         vvdwsum          = _mm_setzero_pd();
 151
 152         /* Start inner kernel loop */
 153         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 154         {
 155
 156             /* Get j neighbor index, and coordinate index */
 157             jnrA             = jjnr[jidx];
 158             jnrB             = jjnr[jidx+1];
 159             j_coord_offsetA  = DIM*jnrA;
 160             j_coord_offsetB  = DIM*jnrB;
 161
 162             /* load j atom coordinates */
 163             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 164                                               &jx0,&jy0,&jz0);
 165
 166             /* Calculate displacement vector */
 167             dx00             = _mm_sub_pd(ix0,jx0);
 168             dy00             = _mm_sub_pd(iy0,jy0);
 169             dz00             = _mm_sub_pd(iz0,jz0);
 170
 171             /* Calculate squared distance and things based on it */
 172             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 173
 174             rinvsq00         = gmx_mm_inv_pd(rsq00);
 175
 176             /* Load parameters for j particles */
 177             vdwjidx0A        = 2*vdwtype[jnrA+0];
 178             vdwjidx0B        = 2*vdwtype[jnrB+0];
 179
 180             /**************************
 181              * CALCULATE INTERACTIONS *
 182              **************************/
 183
 184             if (gmx_mm_any_lt(rsq00,rcutoff2))
 185             {
 186
 187             /* Compute parameters for interactions between i and j atoms */
 188             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 189                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 190
 191             /* LENNARD-JONES DISPERSION/REPULSION */
 192
 193             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 194             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 195             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 196             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) ,
 197                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 198             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 199
 200             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 201
 202             /* Update potential sum for this i atom from the interaction with this j atom. */
 203             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 204             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 205
 206             fscal            = fvdw;
 207
 208             fscal            = _mm_and_pd(fscal,cutoff_mask);
 209
 210             /* Calculate temporary vectorial force */
 211             tx               = _mm_mul_pd(fscal,dx00);
 212             ty               = _mm_mul_pd(fscal,dy00);
 213             tz               = _mm_mul_pd(fscal,dz00);
 214
 215             /* Update vectorial force */
 216             fix0             = _mm_add_pd(fix0,tx);
 217             fiy0             = _mm_add_pd(fiy0,ty);
 218             fiz0             = _mm_add_pd(fiz0,tz);
 219
 220             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 221
 222             }
 223
 224             /* Inner loop uses 41 flops */
 225         }
 226
 227         if(jidx<j_index_end)
 228         {
 229
 230             jnrA             = jjnr[jidx];
 231             j_coord_offsetA  = DIM*jnrA;
 232
 233             /* load j atom coordinates */
 234             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 235                                               &jx0,&jy0,&jz0);
 236
 237             /* Calculate displacement vector */
 238             dx00             = _mm_sub_pd(ix0,jx0);
 239             dy00             = _mm_sub_pd(iy0,jy0);
 240             dz00             = _mm_sub_pd(iz0,jz0);
 241
 242             /* Calculate squared distance and things based on it */
 243             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 244
 245             rinvsq00         = gmx_mm_inv_pd(rsq00);
 246
 247             /* Load parameters for j particles */
 248             vdwjidx0A        = 2*vdwtype[jnrA+0];
 249
 250             /**************************
 251              * CALCULATE INTERACTIONS *
 252              **************************/
 253
 254             if (gmx_mm_any_lt(rsq00,rcutoff2))
 255             {
 256
 257             /* Compute parameters for interactions between i and j atoms */
 258             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 259
 260             /* LENNARD-JONES DISPERSION/REPULSION */
 261
 262             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 263             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 264             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 265             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) ,
 266                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 267             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 268
 269             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 270
 271             /* Update potential sum for this i atom from the interaction with this j atom. */
 272             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 273             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 274             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 275
 276             fscal            = fvdw;
 277
 278             fscal            = _mm_and_pd(fscal,cutoff_mask);
 279
 280             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 281
 282             /* Calculate temporary vectorial force */
 283             tx               = _mm_mul_pd(fscal,dx00);
 284             ty               = _mm_mul_pd(fscal,dy00);
 285             tz               = _mm_mul_pd(fscal,dz00);
 286
 287             /* Update vectorial force */
 288             fix0             = _mm_add_pd(fix0,tx);
 289             fiy0             = _mm_add_pd(fiy0,ty);
 290             fiz0             = _mm_add_pd(fiz0,tz);
 291
 292             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 293
 294             }
 295
 296             /* Inner loop uses 41 flops */
 297         }
 298
 299         /* End of innermost loop */
 300
 301         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 302                                               f+i_coord_offset,fshift+i_shift_offset);
 303
 304         ggid                        = gid[iidx];
 305         /* Update potential energies */
 306         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 307
 308         /* Increment number of inner iterations */
 309         inneriter                  += j_index_end - j_index_start;
 310
 311         /* Outer loop uses 7 flops */
 312     }
 313
 314     /* Increment number of outer iterations */
 315     outeriter        += nri;
 316
 317     /* Update outer/inner flops */
 318
 319     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*41);
 320 }
 321 /*
 322  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 323  * Electrostatics interaction: None
 324  * VdW interaction:            LennardJones
 325  * Geometry:                   Particle-Particle
 326  * Calculate force/pot:        Force
 327  */
 328 void
 329 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_sse4_1_double
 330                     (t_nblist                    * gmx_restrict       nlist,
 331                      rvec                        * gmx_restrict          xx,
 332                      rvec                        * gmx_restrict          ff,
 333                      t_forcerec                  * gmx_restrict          fr,
 334                      t_mdatoms                   * gmx_restrict     mdatoms,
 335                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 336                      t_nrnb                      * gmx_restrict        nrnb)
 337 {
 338     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 339      * just 0 for non-waters.
 340      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 341      * jnr indices corresponding to data put in the four positions in the SIMD register.
 342      */
 343     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 344     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 345     int              jnrA,jnrB;
 346     int              j_coord_offsetA,j_coord_offsetB;
 347     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 348     real             rcutoff_scalar;
 349     real             *shiftvec,*fshift,*x,*f;
 350     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 351     int              vdwioffset0;
 352     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 353     int              vdwjidx0A,vdwjidx0B;
 354     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 355     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 356     int              nvdwtype;
 357     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 358     int              *vdwtype;
 359     real             *vdwparam;
 360     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 361     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 362     __m128d          dummy_mask,cutoff_mask;
 363     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 364     __m128d          one     = _mm_set1_pd(1.0);
 365     __m128d          two     = _mm_set1_pd(2.0);
 366     x                = xx[0];
 367     f                = ff[0];
 368
 369     nri              = nlist->nri;
 370     iinr             = nlist->iinr;
 371     jindex           = nlist->jindex;
 372     jjnr             = nlist->jjnr;
 373     shiftidx         = nlist->shift;
 374     gid              = nlist->gid;
 375     shiftvec         = fr->shift_vec[0];
 376     fshift           = fr->fshift[0];
 377     nvdwtype         = fr->ntype;
 378     vdwparam         = fr->nbfp;
 379     vdwtype          = mdatoms->typeA;
 380
 381     rcutoff_scalar   = fr->rvdw;
 382     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 383     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 384
 385     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 386     rvdw             = _mm_set1_pd(fr->rvdw);
 387
 388     /* Avoid stupid compiler warnings */
 389     jnrA = jnrB = 0;
 390     j_coord_offsetA = 0;
 391     j_coord_offsetB = 0;
 392
 393     outeriter        = 0;
 394     inneriter        = 0;
 395
 396     /* Start outer loop over neighborlists */
 397     for(iidx=0; iidx<nri; iidx++)
 398     {
 399         /* Load shift vector for this list */
 400         i_shift_offset   = DIM*shiftidx[iidx];
 401
 402         /* Load limits for loop over neighbors */
 403         j_index_start    = jindex[iidx];
 404         j_index_end      = jindex[iidx+1];
 405
 406         /* Get outer coordinate index */
 407         inr              = iinr[iidx];
 408         i_coord_offset   = DIM*inr;
 409
 410         /* Load i particle coords and add shift vector */
 411         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 412
 413         fix0             = _mm_setzero_pd();
 414         fiy0             = _mm_setzero_pd();
 415         fiz0             = _mm_setzero_pd();
 416
 417         /* Load parameters for i particles */
 418         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 419
 420         /* Start inner kernel loop */
 421         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 422         {
 423
 424             /* Get j neighbor index, and coordinate index */
 425             jnrA             = jjnr[jidx];
 426             jnrB             = jjnr[jidx+1];
 427             j_coord_offsetA  = DIM*jnrA;
 428             j_coord_offsetB  = DIM*jnrB;
 429
 430             /* load j atom coordinates */
 431             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 432                                               &jx0,&jy0,&jz0);
 433
 434             /* Calculate displacement vector */
 435             dx00             = _mm_sub_pd(ix0,jx0);
 436             dy00             = _mm_sub_pd(iy0,jy0);
 437             dz00             = _mm_sub_pd(iz0,jz0);
 438
 439             /* Calculate squared distance and things based on it */
 440             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 441
 442             rinvsq00         = gmx_mm_inv_pd(rsq00);
 443
 444             /* Load parameters for j particles */
 445             vdwjidx0A        = 2*vdwtype[jnrA+0];
 446             vdwjidx0B        = 2*vdwtype[jnrB+0];
 447
 448             /**************************
 449              * CALCULATE INTERACTIONS *
 450              **************************/
 451
 452             if (gmx_mm_any_lt(rsq00,rcutoff2))
 453             {
 454
 455             /* Compute parameters for interactions between i and j atoms */
 456             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 457                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 458
 459             /* LENNARD-JONES DISPERSION/REPULSION */
 460
 461             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 462             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 463
 464             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 465
 466             fscal            = fvdw;
 467
 468             fscal            = _mm_and_pd(fscal,cutoff_mask);
 469
 470             /* Calculate temporary vectorial force */
 471             tx               = _mm_mul_pd(fscal,dx00);
 472             ty               = _mm_mul_pd(fscal,dy00);
 473             tz               = _mm_mul_pd(fscal,dz00);
 474
 475             /* Update vectorial force */
 476             fix0             = _mm_add_pd(fix0,tx);
 477             fiy0             = _mm_add_pd(fiy0,ty);
 478             fiz0             = _mm_add_pd(fiz0,tz);
 479
 480             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 481
 482             }
 483
 484             /* Inner loop uses 30 flops */
 485         }
 486
 487         if(jidx<j_index_end)
 488         {
 489
 490             jnrA             = jjnr[jidx];
 491             j_coord_offsetA  = DIM*jnrA;
 492
 493             /* load j atom coordinates */
 494             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 495                                               &jx0,&jy0,&jz0);
 496
 497             /* Calculate displacement vector */
 498             dx00             = _mm_sub_pd(ix0,jx0);
 499             dy00             = _mm_sub_pd(iy0,jy0);
 500             dz00             = _mm_sub_pd(iz0,jz0);
 501
 502             /* Calculate squared distance and things based on it */
 503             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 504
 505             rinvsq00         = gmx_mm_inv_pd(rsq00);
 506
 507             /* Load parameters for j particles */
 508             vdwjidx0A        = 2*vdwtype[jnrA+0];
 509
 510             /**************************
 511              * CALCULATE INTERACTIONS *
 512              **************************/
 513
 514             if (gmx_mm_any_lt(rsq00,rcutoff2))
 515             {
 516
 517             /* Compute parameters for interactions between i and j atoms */
 518             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 519
 520             /* LENNARD-JONES DISPERSION/REPULSION */
 521
 522             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 523             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 524
 525             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 526
 527             fscal            = fvdw;
 528
 529             fscal            = _mm_and_pd(fscal,cutoff_mask);
 530
 531             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 532
 533             /* Calculate temporary vectorial force */
 534             tx               = _mm_mul_pd(fscal,dx00);
 535             ty               = _mm_mul_pd(fscal,dy00);
 536             tz               = _mm_mul_pd(fscal,dz00);
 537
 538             /* Update vectorial force */
 539             fix0             = _mm_add_pd(fix0,tx);
 540             fiy0             = _mm_add_pd(fiy0,ty);
 541             fiz0             = _mm_add_pd(fiz0,tz);
 542
 543             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 544
 545             }
 546
 547             /* Inner loop uses 30 flops */
 548         }
 549
 550         /* End of innermost loop */
 551
 552         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 553                                               f+i_coord_offset,fshift+i_shift_offset);
 554
 555         /* Increment number of inner iterations */
 556         inneriter                  += j_index_end - j_index_start;
 557
 558         /* Outer loop uses 6 flops */
 559     }
 560
 561     /* Increment number of outer iterations */
 562     outeriter        += nri;
 563
 564     /* Update outer/inner flops */
 565
 566     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*30);
 567 }