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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  48 #include "kernelutil_x86_avx_128_fma_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_VF_avx_128_fma_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_avx_128_fma_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_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 197                                            _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),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             /* Update vectorial force */
 211             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 212             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 213             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 214
 215             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 216                                                    _mm_mul_pd(dx00,fscal),
 217                                                    _mm_mul_pd(dy00,fscal),
 218                                                    _mm_mul_pd(dz00,fscal));
 219
 220             }
 221
 222             /* Inner loop uses 44 flops */
 223         }
 224
 225         if(jidx<j_index_end)
 226         {
 227
 228             jnrA             = jjnr[jidx];
 229             j_coord_offsetA  = DIM*jnrA;
 230
 231             /* load j atom coordinates */
 232             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 233                                               &jx0,&jy0,&jz0);
 234
 235             /* Calculate displacement vector */
 236             dx00             = _mm_sub_pd(ix0,jx0);
 237             dy00             = _mm_sub_pd(iy0,jy0);
 238             dz00             = _mm_sub_pd(iz0,jz0);
 239
 240             /* Calculate squared distance and things based on it */
 241             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 242
 243             rinvsq00         = gmx_mm_inv_pd(rsq00);
 244
 245             /* Load parameters for j particles */
 246             vdwjidx0A        = 2*vdwtype[jnrA+0];
 247
 248             /**************************
 249              * CALCULATE INTERACTIONS *
 250              **************************/
 251
 252             if (gmx_mm_any_lt(rsq00,rcutoff2))
 253             {
 254
 255             /* Compute parameters for interactions between i and j atoms */
 256             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 257
 258             /* LENNARD-JONES DISPERSION/REPULSION */
 259
 260             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 261             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 262             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 263             vvdw             = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 264                                            _mm_mul_pd(_mm_nmacc_pd( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 265             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 266
 267             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 268
 269             /* Update potential sum for this i atom from the interaction with this j atom. */
 270             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 271             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 272             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 273
 274             fscal            = fvdw;
 275
 276             fscal            = _mm_and_pd(fscal,cutoff_mask);
 277
 278             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 279
 280             /* Update vectorial force */
 281             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 282             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 283             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 284
 285             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 286                                                    _mm_mul_pd(dx00,fscal),
 287                                                    _mm_mul_pd(dy00,fscal),
 288                                                    _mm_mul_pd(dz00,fscal));
 289
 290             }
 291
 292             /* Inner loop uses 44 flops */
 293         }
 294
 295         /* End of innermost loop */
 296
 297         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 298                                               f+i_coord_offset,fshift+i_shift_offset);
 299
 300         ggid                        = gid[iidx];
 301         /* Update potential energies */
 302         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 303
 304         /* Increment number of inner iterations */
 305         inneriter                  += j_index_end - j_index_start;
 306
 307         /* Outer loop uses 7 flops */
 308     }
 309
 310     /* Increment number of outer iterations */
 311     outeriter        += nri;
 312
 313     /* Update outer/inner flops */
 314
 315     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*44);
 316 }
 317 /*
 318  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_avx_128_fma_double
 319  * Electrostatics interaction: None
 320  * VdW interaction:            LennardJones
 321  * Geometry:                   Particle-Particle
 322  * Calculate force/pot:        Force
 323  */
 324 void
 325 nb_kernel_ElecNone_VdwLJSh_GeomP1P1_F_avx_128_fma_double
 326                     (t_nblist                    * gmx_restrict       nlist,
 327                      rvec                        * gmx_restrict          xx,
 328                      rvec                        * gmx_restrict          ff,
 329                      t_forcerec                  * gmx_restrict          fr,
 330                      t_mdatoms                   * gmx_restrict     mdatoms,
 331                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 332                      t_nrnb                      * gmx_restrict        nrnb)
 333 {
 334     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 335      * just 0 for non-waters.
 336      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 337      * jnr indices corresponding to data put in the four positions in the SIMD register.
 338      */
 339     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 340     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 341     int              jnrA,jnrB;
 342     int              j_coord_offsetA,j_coord_offsetB;
 343     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 344     real             rcutoff_scalar;
 345     real             *shiftvec,*fshift,*x,*f;
 346     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 347     int              vdwioffset0;
 348     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 349     int              vdwjidx0A,vdwjidx0B;
 350     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 351     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 352     int              nvdwtype;
 353     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 354     int              *vdwtype;
 355     real             *vdwparam;
 356     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 357     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 358     __m128d          dummy_mask,cutoff_mask;
 359     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 360     __m128d          one     = _mm_set1_pd(1.0);
 361     __m128d          two     = _mm_set1_pd(2.0);
 362     x                = xx[0];
 363     f                = ff[0];
 364
 365     nri              = nlist->nri;
 366     iinr             = nlist->iinr;
 367     jindex           = nlist->jindex;
 368     jjnr             = nlist->jjnr;
 369     shiftidx         = nlist->shift;
 370     gid              = nlist->gid;
 371     shiftvec         = fr->shift_vec[0];
 372     fshift           = fr->fshift[0];
 373     nvdwtype         = fr->ntype;
 374     vdwparam         = fr->nbfp;
 375     vdwtype          = mdatoms->typeA;
 376
 377     rcutoff_scalar   = fr->rvdw;
 378     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 379     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 380
 381     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 382     rvdw             = _mm_set1_pd(fr->rvdw);
 383
 384     /* Avoid stupid compiler warnings */
 385     jnrA = jnrB = 0;
 386     j_coord_offsetA = 0;
 387     j_coord_offsetB = 0;
 388
 389     outeriter        = 0;
 390     inneriter        = 0;
 391
 392     /* Start outer loop over neighborlists */
 393     for(iidx=0; iidx<nri; iidx++)
 394     {
 395         /* Load shift vector for this list */
 396         i_shift_offset   = DIM*shiftidx[iidx];
 397
 398         /* Load limits for loop over neighbors */
 399         j_index_start    = jindex[iidx];
 400         j_index_end      = jindex[iidx+1];
 401
 402         /* Get outer coordinate index */
 403         inr              = iinr[iidx];
 404         i_coord_offset   = DIM*inr;
 405
 406         /* Load i particle coords and add shift vector */
 407         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 408
 409         fix0             = _mm_setzero_pd();
 410         fiy0             = _mm_setzero_pd();
 411         fiz0             = _mm_setzero_pd();
 412
 413         /* Load parameters for i particles */
 414         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 415
 416         /* Start inner kernel loop */
 417         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 418         {
 419
 420             /* Get j neighbor index, and coordinate index */
 421             jnrA             = jjnr[jidx];
 422             jnrB             = jjnr[jidx+1];
 423             j_coord_offsetA  = DIM*jnrA;
 424             j_coord_offsetB  = DIM*jnrB;
 425
 426             /* load j atom coordinates */
 427             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 428                                               &jx0,&jy0,&jz0);
 429
 430             /* Calculate displacement vector */
 431             dx00             = _mm_sub_pd(ix0,jx0);
 432             dy00             = _mm_sub_pd(iy0,jy0);
 433             dz00             = _mm_sub_pd(iz0,jz0);
 434
 435             /* Calculate squared distance and things based on it */
 436             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 437
 438             rinvsq00         = gmx_mm_inv_pd(rsq00);
 439
 440             /* Load parameters for j particles */
 441             vdwjidx0A        = 2*vdwtype[jnrA+0];
 442             vdwjidx0B        = 2*vdwtype[jnrB+0];
 443
 444             /**************************
 445              * CALCULATE INTERACTIONS *
 446              **************************/
 447
 448             if (gmx_mm_any_lt(rsq00,rcutoff2))
 449             {
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 453                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 454
 455             /* LENNARD-JONES DISPERSION/REPULSION */
 456
 457             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 458             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 459
 460             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 461
 462             fscal            = fvdw;
 463
 464             fscal            = _mm_and_pd(fscal,cutoff_mask);
 465
 466             /* Update vectorial force */
 467             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 468             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 469             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 470
 471             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 472                                                    _mm_mul_pd(dx00,fscal),
 473                                                    _mm_mul_pd(dy00,fscal),
 474                                                    _mm_mul_pd(dz00,fscal));
 475
 476             }
 477
 478             /* Inner loop uses 33 flops */
 479         }
 480
 481         if(jidx<j_index_end)
 482         {
 483
 484             jnrA             = jjnr[jidx];
 485             j_coord_offsetA  = DIM*jnrA;
 486
 487             /* load j atom coordinates */
 488             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 489                                               &jx0,&jy0,&jz0);
 490
 491             /* Calculate displacement vector */
 492             dx00             = _mm_sub_pd(ix0,jx0);
 493             dy00             = _mm_sub_pd(iy0,jy0);
 494             dz00             = _mm_sub_pd(iz0,jz0);
 495
 496             /* Calculate squared distance and things based on it */
 497             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 498
 499             rinvsq00         = gmx_mm_inv_pd(rsq00);
 500
 501             /* Load parameters for j particles */
 502             vdwjidx0A        = 2*vdwtype[jnrA+0];
 503
 504             /**************************
 505              * CALCULATE INTERACTIONS *
 506              **************************/
 507
 508             if (gmx_mm_any_lt(rsq00,rcutoff2))
 509             {
 510
 511             /* Compute parameters for interactions between i and j atoms */
 512             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 513
 514             /* LENNARD-JONES DISPERSION/REPULSION */
 515
 516             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 517             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 518
 519             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 520
 521             fscal            = fvdw;
 522
 523             fscal            = _mm_and_pd(fscal,cutoff_mask);
 524
 525             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 526
 527             /* Update vectorial force */
 528             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 529             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 530             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 531
 532             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 533                                                    _mm_mul_pd(dx00,fscal),
 534                                                    _mm_mul_pd(dy00,fscal),
 535                                                    _mm_mul_pd(dz00,fscal));
 536
 537             }
 538
 539             /* Inner loop uses 33 flops */
 540         }
 541
 542         /* End of innermost loop */
 543
 544         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 545                                               f+i_coord_offset,fshift+i_shift_offset);
 546
 547         /* Increment number of inner iterations */
 548         inneriter                  += j_index_end - j_index_start;
 549
 550         /* Outer loop uses 6 flops */
 551     }
 552
 553     /* Increment number of outer iterations */
 554     outeriter        += nri;
 555
 556     /* Update outer/inner flops */
 557
 558     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*33);
 559 }