src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecRF_VdwLJ_GeomP1P1_avx_128_fma_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015, 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/math/vec.h"
  46 #include "gromacs/legacyheaders/nrnb.h"
  47
  48 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  49 #include "kernelutil_x86_avx_128_fma_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_avx_128_fma_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_avx_128_fma_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwjidx0A,vdwjidx0B;
  84     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  85     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  87     real             *charge;
  88     int              nvdwtype;
  89     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  90     int              *vdwtype;
  91     real             *vdwparam;
  92     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  93     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  94     __m128d          dummy_mask,cutoff_mask;
  95     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  96     __m128d          one     = _mm_set1_pd(1.0);
  97     __m128d          two     = _mm_set1_pd(2.0);
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = _mm_set1_pd(fr->epsfac);
 110     charge           = mdatoms->chargeA;
 111     krf              = _mm_set1_pd(fr->ic->k_rf);
 112     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 113     crf              = _mm_set1_pd(fr->ic->c_rf);
 114     nvdwtype         = fr->ntype;
 115     vdwparam         = fr->nbfp;
 116     vdwtype          = mdatoms->typeA;
 117
 118     /* Avoid stupid compiler warnings */
 119     jnrA = jnrB = 0;
 120     j_coord_offsetA = 0;
 121     j_coord_offsetB = 0;
 122
 123     outeriter        = 0;
 124     inneriter        = 0;
 125
 126     /* Start outer loop over neighborlists */
 127     for(iidx=0; iidx<nri; iidx++)
 128     {
 129         /* Load shift vector for this list */
 130         i_shift_offset   = DIM*shiftidx[iidx];
 131
 132         /* Load limits for loop over neighbors */
 133         j_index_start    = jindex[iidx];
 134         j_index_end      = jindex[iidx+1];
 135
 136         /* Get outer coordinate index */
 137         inr              = iinr[iidx];
 138         i_coord_offset   = DIM*inr;
 139
 140         /* Load i particle coords and add shift vector */
 141         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 142
 143         fix0             = _mm_setzero_pd();
 144         fiy0             = _mm_setzero_pd();
 145         fiz0             = _mm_setzero_pd();
 146
 147         /* Load parameters for i particles */
 148         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 149         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 150
 151         /* Reset potential sums */
 152         velecsum         = _mm_setzero_pd();
 153         vvdwsum          = _mm_setzero_pd();
 154
 155         /* Start inner kernel loop */
 156         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 157         {
 158
 159             /* Get j neighbor index, and coordinate index */
 160             jnrA             = jjnr[jidx];
 161             jnrB             = jjnr[jidx+1];
 162             j_coord_offsetA  = DIM*jnrA;
 163             j_coord_offsetB  = DIM*jnrB;
 164
 165             /* load j atom coordinates */
 166             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 167                                               &jx0,&jy0,&jz0);
 168
 169             /* Calculate displacement vector */
 170             dx00             = _mm_sub_pd(ix0,jx0);
 171             dy00             = _mm_sub_pd(iy0,jy0);
 172             dz00             = _mm_sub_pd(iz0,jz0);
 173
 174             /* Calculate squared distance and things based on it */
 175             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 176
 177             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 178
 179             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 180
 181             /* Load parameters for j particles */
 182             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 183             vdwjidx0A        = 2*vdwtype[jnrA+0];
 184             vdwjidx0B        = 2*vdwtype[jnrB+0];
 185
 186             /**************************
 187              * CALCULATE INTERACTIONS *
 188              **************************/
 189
 190             /* Compute parameters for interactions between i and j atoms */
 191             qq00             = _mm_mul_pd(iq0,jq0);
 192             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 193                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 194
 195             /* REACTION-FIELD ELECTROSTATICS */
 196             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_macc_pd(krf,rsq00,rinv00),crf));
 197             felec            = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
 198
 199             /* LENNARD-JONES DISPERSION/REPULSION */
 200
 201             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 202             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 203             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 204             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 205             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 206
 207             /* Update potential sum for this i atom from the interaction with this j atom. */
 208             velecsum         = _mm_add_pd(velecsum,velec);
 209             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 210
 211             fscal            = _mm_add_pd(felec,fvdw);
 212
 213             /* Update vectorial force */
 214             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 215             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 216             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 217
 218             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 219                                                    _mm_mul_pd(dx00,fscal),
 220                                                    _mm_mul_pd(dy00,fscal),
 221                                                    _mm_mul_pd(dz00,fscal));
 222
 223             /* Inner loop uses 47 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             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 245
 246             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 247
 248             /* Load parameters for j particles */
 249             jq0              = _mm_load_sd(charge+jnrA+0);
 250             vdwjidx0A        = 2*vdwtype[jnrA+0];
 251
 252             /**************************
 253              * CALCULATE INTERACTIONS *
 254              **************************/
 255
 256             /* Compute parameters for interactions between i and j atoms */
 257             qq00             = _mm_mul_pd(iq0,jq0);
 258             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 259
 260             /* REACTION-FIELD ELECTROSTATICS */
 261             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_macc_pd(krf,rsq00,rinv00),crf));
 262             felec            = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
 263
 264             /* LENNARD-JONES DISPERSION/REPULSION */
 265
 266             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 267             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 268             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 269             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 270             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 271
 272             /* Update potential sum for this i atom from the interaction with this j atom. */
 273             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 274             velecsum         = _mm_add_pd(velecsum,velec);
 275             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 276             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 277
 278             fscal            = _mm_add_pd(felec,fvdw);
 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             /* Inner loop uses 47 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(velecsum,kernel_data->energygrp_elec+ggid);
 303         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 304
 305         /* Increment number of inner iterations */
 306         inneriter                  += j_index_end - j_index_start;
 307
 308         /* Outer loop uses 9 flops */
 309     }
 310
 311     /* Increment number of outer iterations */
 312     outeriter        += nri;
 313
 314     /* Update outer/inner flops */
 315
 316     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*47);
 317 }
 318 /*
 319  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_avx_128_fma_double
 320  * Electrostatics interaction: ReactionField
 321  * VdW interaction:            LennardJones
 322  * Geometry:                   Particle-Particle
 323  * Calculate force/pot:        Force
 324  */
 325 void
 326 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_avx_128_fma_double
 327                     (t_nblist                    * gmx_restrict       nlist,
 328                      rvec                        * gmx_restrict          xx,
 329                      rvec                        * gmx_restrict          ff,
 330                      t_forcerec                  * gmx_restrict          fr,
 331                      t_mdatoms                   * gmx_restrict     mdatoms,
 332                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 333                      t_nrnb                      * gmx_restrict        nrnb)
 334 {
 335     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 336      * just 0 for non-waters.
 337      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 338      * jnr indices corresponding to data put in the four positions in the SIMD register.
 339      */
 340     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 341     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 342     int              jnrA,jnrB;
 343     int              j_coord_offsetA,j_coord_offsetB;
 344     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 345     real             rcutoff_scalar;
 346     real             *shiftvec,*fshift,*x,*f;
 347     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 348     int              vdwioffset0;
 349     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 350     int              vdwjidx0A,vdwjidx0B;
 351     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 352     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 353     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 354     real             *charge;
 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     facel            = _mm_set1_pd(fr->epsfac);
 377     charge           = mdatoms->chargeA;
 378     krf              = _mm_set1_pd(fr->ic->k_rf);
 379     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 380     crf              = _mm_set1_pd(fr->ic->c_rf);
 381     nvdwtype         = fr->ntype;
 382     vdwparam         = fr->nbfp;
 383     vdwtype          = mdatoms->typeA;
 384
 385     /* Avoid stupid compiler warnings */
 386     jnrA = jnrB = 0;
 387     j_coord_offsetA = 0;
 388     j_coord_offsetB = 0;
 389
 390     outeriter        = 0;
 391     inneriter        = 0;
 392
 393     /* Start outer loop over neighborlists */
 394     for(iidx=0; iidx<nri; iidx++)
 395     {
 396         /* Load shift vector for this list */
 397         i_shift_offset   = DIM*shiftidx[iidx];
 398
 399         /* Load limits for loop over neighbors */
 400         j_index_start    = jindex[iidx];
 401         j_index_end      = jindex[iidx+1];
 402
 403         /* Get outer coordinate index */
 404         inr              = iinr[iidx];
 405         i_coord_offset   = DIM*inr;
 406
 407         /* Load i particle coords and add shift vector */
 408         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 409
 410         fix0             = _mm_setzero_pd();
 411         fiy0             = _mm_setzero_pd();
 412         fiz0             = _mm_setzero_pd();
 413
 414         /* Load parameters for i particles */
 415         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 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             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 441
 442             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 443
 444             /* Load parameters for j particles */
 445             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 446             vdwjidx0A        = 2*vdwtype[jnrA+0];
 447             vdwjidx0B        = 2*vdwtype[jnrB+0];
 448
 449             /**************************
 450              * CALCULATE INTERACTIONS *
 451              **************************/
 452
 453             /* Compute parameters for interactions between i and j atoms */
 454             qq00             = _mm_mul_pd(iq0,jq0);
 455             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 456                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 457
 458             /* REACTION-FIELD ELECTROSTATICS */
 459             felec            = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
 460
 461             /* LENNARD-JONES DISPERSION/REPULSION */
 462
 463             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 464             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 465
 466             fscal            = _mm_add_pd(felec,fvdw);
 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             /* Inner loop uses 37 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             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 500
 501             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 502
 503             /* Load parameters for j particles */
 504             jq0              = _mm_load_sd(charge+jnrA+0);
 505             vdwjidx0A        = 2*vdwtype[jnrA+0];
 506
 507             /**************************
 508              * CALCULATE INTERACTIONS *
 509              **************************/
 510
 511             /* Compute parameters for interactions between i and j atoms */
 512             qq00             = _mm_mul_pd(iq0,jq0);
 513             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 514
 515             /* REACTION-FIELD ELECTROSTATICS */
 516             felec            = _mm_mul_pd(qq00,_mm_msub_pd(rinv00,rinvsq00,krf2));
 517
 518             /* LENNARD-JONES DISPERSION/REPULSION */
 519
 520             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 521             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 522
 523             fscal            = _mm_add_pd(felec,fvdw);
 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             /* Inner loop uses 37 flops */
 538         }
 539
 540         /* End of innermost loop */
 541
 542         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 543                                               f+i_coord_offset,fshift+i_shift_offset);
 544
 545         /* Increment number of inner iterations */
 546         inneriter                  += j_index_end - j_index_start;
 547
 548         /* Outer loop uses 7 flops */
 549     }
 550
 551     /* Increment number of outer iterations */
 552     outeriter        += nri;
 553
 554     /* Update outer/inner flops */
 555
 556     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 557 }