src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecNone_VdwLJSh_GeomP1P1_avx_128_fma_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 avx_128_fma_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 "gromacs/simd/math_x86_avx_128_fma_double.h"
  50 #include "kernelutil_x86_avx_128_fma_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_avx_128_fma_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_avx_128_fma_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_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 199                                            _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),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             /* Update vectorial force */
 213             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 214             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 215             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 216
 217             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 218                                                    _mm_mul_pd(dx00,fscal),
 219                                                    _mm_mul_pd(dy00,fscal),
 220                                                    _mm_mul_pd(dz00,fscal));
 221
 222             }
 223
 224             /* Inner loop uses 44 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_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 266                                            _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),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             /* Update vectorial force */
 283             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 284             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 285             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 286
 287             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 288                                                    _mm_mul_pd(dx00,fscal),
 289                                                    _mm_mul_pd(dy00,fscal),
 290                                                    _mm_mul_pd(dz00,fscal));
 291
 292             }
 293
 294             /* Inner loop uses 44 flops */
 295         }
 296
 297         /* End of innermost loop */
 298
 299         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 300                                               f+i_coord_offset,fshift+i_shift_offset);
 301
 302         ggid                        = gid[iidx];
 303         /* Update potential energies */
 304         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 305
 306         /* Increment number of inner iterations */
 307         inneriter                  += j_index_end - j_index_start;
 308
 309         /* Outer loop uses 7 flops */
 310     }
 311
 312     /* Increment number of outer iterations */
 313     outeriter        += nri;
 314
 315     /* Update outer/inner flops */
 316
 317     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*44);
 318 }
 319 /*
 320  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_avx_128_fma_double
 321  * Electrostatics interaction: None
 322  * VdW interaction:            LennardJones
 323  * Geometry:                   Particle-Particle
 324  * Calculate force/pot:        Force
 325  */
 326 void
 327 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_avx_128_fma_double
 328                     (t_nblist                    * gmx_restrict       nlist,
 329                      rvec                        * gmx_restrict          xx,
 330                      rvec                        * gmx_restrict          ff,
 331                      t_forcerec                  * gmx_restrict          fr,
 332                      t_mdatoms                   * gmx_restrict     mdatoms,
 333                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 334                      t_nrnb                      * gmx_restrict        nrnb)
 335 {
 336     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 337      * just 0 for non-waters.
 338      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 339      * jnr indices corresponding to data put in the four positions in the SIMD register.
 340      */
 341     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 342     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 343     int              jnrA,jnrB;
 344     int              j_coord_offsetA,j_coord_offsetB;
 345     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 346     real             rcutoff_scalar;
 347     real             *shiftvec,*fshift,*x,*f;
 348     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 349     int              vdwioffset0;
 350     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 351     int              vdwjidx0A,vdwjidx0B;
 352     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 353     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 354     int              nvdwtype;
 355     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 356     int              *vdwtype;
 357     real             *vdwparam;
 358     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 359     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 360     __m128d          dummy_mask,cutoff_mask;
 361     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 362     __m128d          one     = _mm_set1_pd(1.0);
 363     __m128d          two     = _mm_set1_pd(2.0);
 364     x                = xx[0];
 365     f                = ff[0];
 366
 367     nri              = nlist->nri;
 368     iinr             = nlist->iinr;
 369     jindex           = nlist->jindex;
 370     jjnr             = nlist->jjnr;
 371     shiftidx         = nlist->shift;
 372     gid              = nlist->gid;
 373     shiftvec         = fr->shift_vec[0];
 374     fshift           = fr->fshift[0];
 375     nvdwtype         = fr->ntype;
 376     vdwparam         = fr->nbfp;
 377     vdwtype          = mdatoms->typeA;
 378
 379     rcutoff_scalar   = fr->rvdw;
 380     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 381     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 382
 383     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 384     rvdw             = _mm_set1_pd(fr->rvdw);
 385
 386     /* Avoid stupid compiler warnings */
 387     jnrA = jnrB = 0;
 388     j_coord_offsetA = 0;
 389     j_coord_offsetB = 0;
 390
 391     outeriter        = 0;
 392     inneriter        = 0;
 393
 394     /* Start outer loop over neighborlists */
 395     for(iidx=0; iidx<nri; iidx++)
 396     {
 397         /* Load shift vector for this list */
 398         i_shift_offset   = DIM*shiftidx[iidx];
 399
 400         /* Load limits for loop over neighbors */
 401         j_index_start    = jindex[iidx];
 402         j_index_end      = jindex[iidx+1];
 403
 404         /* Get outer coordinate index */
 405         inr              = iinr[iidx];
 406         i_coord_offset   = DIM*inr;
 407
 408         /* Load i particle coords and add shift vector */
 409         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 410
 411         fix0             = _mm_setzero_pd();
 412         fiy0             = _mm_setzero_pd();
 413         fiz0             = _mm_setzero_pd();
 414
 415         /* Load parameters for i particles */
 416         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 417
 418         /* Start inner kernel loop */
 419         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 420         {
 421
 422             /* Get j neighbor index, and coordinate index */
 423             jnrA             = jjnr[jidx];
 424             jnrB             = jjnr[jidx+1];
 425             j_coord_offsetA  = DIM*jnrA;
 426             j_coord_offsetB  = DIM*jnrB;
 427
 428             /* load j atom coordinates */
 429             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 430                                               &jx0,&jy0,&jz0);
 431
 432             /* Calculate displacement vector */
 433             dx00             = _mm_sub_pd(ix0,jx0);
 434             dy00             = _mm_sub_pd(iy0,jy0);
 435             dz00             = _mm_sub_pd(iz0,jz0);
 436
 437             /* Calculate squared distance and things based on it */
 438             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 439
 440             rinvsq00         = gmx_mm_inv_pd(rsq00);
 441
 442             /* Load parameters for j particles */
 443             vdwjidx0A        = 2*vdwtype[jnrA+0];
 444             vdwjidx0B        = 2*vdwtype[jnrB+0];
 445
 446             /**************************
 447              * CALCULATE INTERACTIONS *
 448              **************************/
 449
 450             if (gmx_mm_any_lt(rsq00,rcutoff2))
 451             {
 452
 453             /* Compute parameters for interactions between i and j atoms */
 454             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 455                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 456
 457             /* LENNARD-JONES DISPERSION/REPULSION */
 458
 459             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 460             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 461
 462             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 463
 464             fscal            = fvdw;
 465
 466             fscal            = _mm_and_pd(fscal,cutoff_mask);
 467
 468             /* Update vectorial force */
 469             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 470             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 471             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 472
 473             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 474                                                    _mm_mul_pd(dx00,fscal),
 475                                                    _mm_mul_pd(dy00,fscal),
 476                                                    _mm_mul_pd(dz00,fscal));
 477
 478             }
 479
 480             /* Inner loop uses 33 flops */
 481         }
 482
 483         if(jidx<j_index_end)
 484         {
 485
 486             jnrA             = jjnr[jidx];
 487             j_coord_offsetA  = DIM*jnrA;
 488
 489             /* load j atom coordinates */
 490             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 491                                               &jx0,&jy0,&jz0);
 492
 493             /* Calculate displacement vector */
 494             dx00             = _mm_sub_pd(ix0,jx0);
 495             dy00             = _mm_sub_pd(iy0,jy0);
 496             dz00             = _mm_sub_pd(iz0,jz0);
 497
 498             /* Calculate squared distance and things based on it */
 499             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 500
 501             rinvsq00         = gmx_mm_inv_pd(rsq00);
 502
 503             /* Load parameters for j particles */
 504             vdwjidx0A        = 2*vdwtype[jnrA+0];
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             if (gmx_mm_any_lt(rsq00,rcutoff2))
 511             {
 512
 513             /* Compute parameters for interactions between i and j atoms */
 514             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 515
 516             /* LENNARD-JONES DISPERSION/REPULSION */
 517
 518             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 519             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 520
 521             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 522
 523             fscal            = fvdw;
 524
 525             fscal            = _mm_and_pd(fscal,cutoff_mask);
 526
 527             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 528
 529             /* Update vectorial force */
 530             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 531             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 532             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 533
 534             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 535                                                    _mm_mul_pd(dx00,fscal),
 536                                                    _mm_mul_pd(dy00,fscal),
 537                                                    _mm_mul_pd(dz00,fscal));
 538
 539             }
 540
 541             /* Inner loop uses 33 flops */
 542         }
 543
 544         /* End of innermost loop */
 545
 546         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 547                                               f+i_coord_offset,fshift+i_shift_offset);
 548
 549         /* Increment number of inner iterations */
 550         inneriter                  += j_index_end - j_index_start;
 551
 552         /* Outer loop uses 6 flops */
 553     }
 554
 555     /* Increment number of outer iterations */
 556     outeriter        += nri;
 557
 558     /* Update outer/inner flops */
 559
 560     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*33);
 561 }