src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecCoul_VdwLJ_GeomP1P1_avx_128_fma_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_128_fma_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_128_fma_double.h"
  34 #include "kernelutil_x86_avx_128_fma_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomP1P1_VF_avx_128_fma_double
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_VF_avx_128_fma_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB;
  61     int              j_coord_offsetA,j_coord_offsetB;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwjidx0A,vdwjidx0B;
  69     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  70     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  71     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  72     real             *charge;
  73     int              nvdwtype;
  74     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  75     int              *vdwtype;
  76     real             *vdwparam;
  77     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  78     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  79     __m128d          dummy_mask,cutoff_mask;
  80     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  81     __m128d          one     = _mm_set1_pd(1.0);
  82     __m128d          two     = _mm_set1_pd(2.0);
  83     x                = xx[0];
  84     f                = ff[0];
  85
  86     nri              = nlist->nri;
  87     iinr             = nlist->iinr;
  88     jindex           = nlist->jindex;
  89     jjnr             = nlist->jjnr;
  90     shiftidx         = nlist->shift;
  91     gid              = nlist->gid;
  92     shiftvec         = fr->shift_vec[0];
  93     fshift           = fr->fshift[0];
  94     facel            = _mm_set1_pd(fr->epsfac);
  95     charge           = mdatoms->chargeA;
  96     nvdwtype         = fr->ntype;
  97     vdwparam         = fr->nbfp;
  98     vdwtype          = mdatoms->typeA;
  99
 100     /* Avoid stupid compiler warnings */
 101     jnrA = jnrB = 0;
 102     j_coord_offsetA = 0;
 103     j_coord_offsetB = 0;
 104
 105     outeriter        = 0;
 106     inneriter        = 0;
 107
 108     /* Start outer loop over neighborlists */
 109     for(iidx=0; iidx<nri; iidx++)
 110     {
 111         /* Load shift vector for this list */
 112         i_shift_offset   = DIM*shiftidx[iidx];
 113
 114         /* Load limits for loop over neighbors */
 115         j_index_start    = jindex[iidx];
 116         j_index_end      = jindex[iidx+1];
 117
 118         /* Get outer coordinate index */
 119         inr              = iinr[iidx];
 120         i_coord_offset   = DIM*inr;
 121
 122         /* Load i particle coords and add shift vector */
 123         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 124
 125         fix0             = _mm_setzero_pd();
 126         fiy0             = _mm_setzero_pd();
 127         fiz0             = _mm_setzero_pd();
 128
 129         /* Load parameters for i particles */
 130         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 131         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133         /* Reset potential sums */
 134         velecsum         = _mm_setzero_pd();
 135         vvdwsum          = _mm_setzero_pd();
 136
 137         /* Start inner kernel loop */
 138         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 139         {
 140
 141             /* Get j neighbor index, and coordinate index */
 142             jnrA             = jjnr[jidx];
 143             jnrB             = jjnr[jidx+1];
 144             j_coord_offsetA  = DIM*jnrA;
 145             j_coord_offsetB  = DIM*jnrB;
 146
 147             /* load j atom coordinates */
 148             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 149                                               &jx0,&jy0,&jz0);
 150
 151             /* Calculate displacement vector */
 152             dx00             = _mm_sub_pd(ix0,jx0);
 153             dy00             = _mm_sub_pd(iy0,jy0);
 154             dz00             = _mm_sub_pd(iz0,jz0);
 155
 156             /* Calculate squared distance and things based on it */
 157             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 158
 159             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 160
 161             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 162
 163             /* Load parameters for j particles */
 164             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 165             vdwjidx0A        = 2*vdwtype[jnrA+0];
 166             vdwjidx0B        = 2*vdwtype[jnrB+0];
 167
 168             /**************************
 169              * CALCULATE INTERACTIONS *
 170              **************************/
 171
 172             /* Compute parameters for interactions between i and j atoms */
 173             qq00             = _mm_mul_pd(iq0,jq0);
 174             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 175                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 176
 177             /* COULOMB ELECTROSTATICS */
 178             velec            = _mm_mul_pd(qq00,rinv00);
 179             felec            = _mm_mul_pd(velec,rinvsq00);
 180
 181             /* LENNARD-JONES DISPERSION/REPULSION */
 182
 183             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 184             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 185             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 186             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 187             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 188
 189             /* Update potential sum for this i atom from the interaction with this j atom. */
 190             velecsum         = _mm_add_pd(velecsum,velec);
 191             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 192
 193             fscal            = _mm_add_pd(felec,fvdw);
 194
 195             /* Update vectorial force */
 196             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 197             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 198             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 199
 200             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 201                                                    _mm_mul_pd(dx00,fscal),
 202                                                    _mm_mul_pd(dy00,fscal),
 203                                                    _mm_mul_pd(dz00,fscal));
 204
 205             /* Inner loop uses 43 flops */
 206         }
 207
 208         if(jidx<j_index_end)
 209         {
 210
 211             jnrA             = jjnr[jidx];
 212             j_coord_offsetA  = DIM*jnrA;
 213
 214             /* load j atom coordinates */
 215             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 216                                               &jx0,&jy0,&jz0);
 217
 218             /* Calculate displacement vector */
 219             dx00             = _mm_sub_pd(ix0,jx0);
 220             dy00             = _mm_sub_pd(iy0,jy0);
 221             dz00             = _mm_sub_pd(iz0,jz0);
 222
 223             /* Calculate squared distance and things based on it */
 224             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 225
 226             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 227
 228             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 229
 230             /* Load parameters for j particles */
 231             jq0              = _mm_load_sd(charge+jnrA+0);
 232             vdwjidx0A        = 2*vdwtype[jnrA+0];
 233
 234             /**************************
 235              * CALCULATE INTERACTIONS *
 236              **************************/
 237
 238             /* Compute parameters for interactions between i and j atoms */
 239             qq00             = _mm_mul_pd(iq0,jq0);
 240             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 241
 242             /* COULOMB ELECTROSTATICS */
 243             velec            = _mm_mul_pd(qq00,rinv00);
 244             felec            = _mm_mul_pd(velec,rinvsq00);
 245
 246             /* LENNARD-JONES DISPERSION/REPULSION */
 247
 248             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 249             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 250             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 251             vvdw             = _mm_msub_pd( vvdw12,one_twelfth, _mm_mul_pd(vvdw6,one_sixth) );
 252             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 253
 254             /* Update potential sum for this i atom from the interaction with this j atom. */
 255             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 256             velecsum         = _mm_add_pd(velecsum,velec);
 257             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 258             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 259
 260             fscal            = _mm_add_pd(felec,fvdw);
 261
 262             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 263
 264             /* Update vectorial force */
 265             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 266             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 267             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 268
 269             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 270                                                    _mm_mul_pd(dx00,fscal),
 271                                                    _mm_mul_pd(dy00,fscal),
 272                                                    _mm_mul_pd(dz00,fscal));
 273
 274             /* Inner loop uses 43 flops */
 275         }
 276
 277         /* End of innermost loop */
 278
 279         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 280                                               f+i_coord_offset,fshift+i_shift_offset);
 281
 282         ggid                        = gid[iidx];
 283         /* Update potential energies */
 284         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 285         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 286
 287         /* Increment number of inner iterations */
 288         inneriter                  += j_index_end - j_index_start;
 289
 290         /* Outer loop uses 9 flops */
 291     }
 292
 293     /* Increment number of outer iterations */
 294     outeriter        += nri;
 295
 296     /* Update outer/inner flops */
 297
 298     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*43);
 299 }
 300 /*
 301  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_avx_128_fma_double
 302  * Electrostatics interaction: Coulomb
 303  * VdW interaction:            LennardJones
 304  * Geometry:                   Particle-Particle
 305  * Calculate force/pot:        Force
 306  */
 307 void
 308 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_avx_128_fma_double
 309                     (t_nblist * gmx_restrict                nlist,
 310                      rvec * gmx_restrict                    xx,
 311                      rvec * gmx_restrict                    ff,
 312                      t_forcerec * gmx_restrict              fr,
 313                      t_mdatoms * gmx_restrict               mdatoms,
 314                      nb_kernel_data_t * gmx_restrict        kernel_data,
 315                      t_nrnb * gmx_restrict                  nrnb)
 316 {
 317     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 318      * just 0 for non-waters.
 319      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 320      * jnr indices corresponding to data put in the four positions in the SIMD register.
 321      */
 322     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 323     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 324     int              jnrA,jnrB;
 325     int              j_coord_offsetA,j_coord_offsetB;
 326     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 327     real             rcutoff_scalar;
 328     real             *shiftvec,*fshift,*x,*f;
 329     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 330     int              vdwioffset0;
 331     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 332     int              vdwjidx0A,vdwjidx0B;
 333     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 334     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 335     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 336     real             *charge;
 337     int              nvdwtype;
 338     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 339     int              *vdwtype;
 340     real             *vdwparam;
 341     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 342     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 343     __m128d          dummy_mask,cutoff_mask;
 344     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 345     __m128d          one     = _mm_set1_pd(1.0);
 346     __m128d          two     = _mm_set1_pd(2.0);
 347     x                = xx[0];
 348     f                = ff[0];
 349
 350     nri              = nlist->nri;
 351     iinr             = nlist->iinr;
 352     jindex           = nlist->jindex;
 353     jjnr             = nlist->jjnr;
 354     shiftidx         = nlist->shift;
 355     gid              = nlist->gid;
 356     shiftvec         = fr->shift_vec[0];
 357     fshift           = fr->fshift[0];
 358     facel            = _mm_set1_pd(fr->epsfac);
 359     charge           = mdatoms->chargeA;
 360     nvdwtype         = fr->ntype;
 361     vdwparam         = fr->nbfp;
 362     vdwtype          = mdatoms->typeA;
 363
 364     /* Avoid stupid compiler warnings */
 365     jnrA = jnrB = 0;
 366     j_coord_offsetA = 0;
 367     j_coord_offsetB = 0;
 368
 369     outeriter        = 0;
 370     inneriter        = 0;
 371
 372     /* Start outer loop over neighborlists */
 373     for(iidx=0; iidx<nri; iidx++)
 374     {
 375         /* Load shift vector for this list */
 376         i_shift_offset   = DIM*shiftidx[iidx];
 377
 378         /* Load limits for loop over neighbors */
 379         j_index_start    = jindex[iidx];
 380         j_index_end      = jindex[iidx+1];
 381
 382         /* Get outer coordinate index */
 383         inr              = iinr[iidx];
 384         i_coord_offset   = DIM*inr;
 385
 386         /* Load i particle coords and add shift vector */
 387         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 388
 389         fix0             = _mm_setzero_pd();
 390         fiy0             = _mm_setzero_pd();
 391         fiz0             = _mm_setzero_pd();
 392
 393         /* Load parameters for i particles */
 394         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 395         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 396
 397         /* Start inner kernel loop */
 398         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 399         {
 400
 401             /* Get j neighbor index, and coordinate index */
 402             jnrA             = jjnr[jidx];
 403             jnrB             = jjnr[jidx+1];
 404             j_coord_offsetA  = DIM*jnrA;
 405             j_coord_offsetB  = DIM*jnrB;
 406
 407             /* load j atom coordinates */
 408             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 409                                               &jx0,&jy0,&jz0);
 410
 411             /* Calculate displacement vector */
 412             dx00             = _mm_sub_pd(ix0,jx0);
 413             dy00             = _mm_sub_pd(iy0,jy0);
 414             dz00             = _mm_sub_pd(iz0,jz0);
 415
 416             /* Calculate squared distance and things based on it */
 417             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 418
 419             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 420
 421             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 422
 423             /* Load parameters for j particles */
 424             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 425             vdwjidx0A        = 2*vdwtype[jnrA+0];
 426             vdwjidx0B        = 2*vdwtype[jnrB+0];
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             /* Compute parameters for interactions between i and j atoms */
 433             qq00             = _mm_mul_pd(iq0,jq0);
 434             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 435                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 436
 437             /* COULOMB ELECTROSTATICS */
 438             velec            = _mm_mul_pd(qq00,rinv00);
 439             felec            = _mm_mul_pd(velec,rinvsq00);
 440
 441             /* LENNARD-JONES DISPERSION/REPULSION */
 442
 443             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 444             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 445
 446             fscal            = _mm_add_pd(felec,fvdw);
 447
 448             /* Update vectorial force */
 449             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 450             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 451             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 452
 453             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 454                                                    _mm_mul_pd(dx00,fscal),
 455                                                    _mm_mul_pd(dy00,fscal),
 456                                                    _mm_mul_pd(dz00,fscal));
 457
 458             /* Inner loop uses 37 flops */
 459         }
 460
 461         if(jidx<j_index_end)
 462         {
 463
 464             jnrA             = jjnr[jidx];
 465             j_coord_offsetA  = DIM*jnrA;
 466
 467             /* load j atom coordinates */
 468             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 469                                               &jx0,&jy0,&jz0);
 470
 471             /* Calculate displacement vector */
 472             dx00             = _mm_sub_pd(ix0,jx0);
 473             dy00             = _mm_sub_pd(iy0,jy0);
 474             dz00             = _mm_sub_pd(iz0,jz0);
 475
 476             /* Calculate squared distance and things based on it */
 477             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 478
 479             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 480
 481             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 482
 483             /* Load parameters for j particles */
 484             jq0              = _mm_load_sd(charge+jnrA+0);
 485             vdwjidx0A        = 2*vdwtype[jnrA+0];
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             /* Compute parameters for interactions between i and j atoms */
 492             qq00             = _mm_mul_pd(iq0,jq0);
 493             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 494
 495             /* COULOMB ELECTROSTATICS */
 496             velec            = _mm_mul_pd(qq00,rinv00);
 497             felec            = _mm_mul_pd(velec,rinvsq00);
 498
 499             /* LENNARD-JONES DISPERSION/REPULSION */
 500
 501             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 502             fvdw             = _mm_mul_pd(_mm_msub_pd(c12_00,rinvsix,c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 503
 504             fscal            = _mm_add_pd(felec,fvdw);
 505
 506             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 507
 508             /* Update vectorial force */
 509             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 510             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 511             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 512
 513             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 514                                                    _mm_mul_pd(dx00,fscal),
 515                                                    _mm_mul_pd(dy00,fscal),
 516                                                    _mm_mul_pd(dz00,fscal));
 517
 518             /* Inner loop uses 37 flops */
 519         }
 520
 521         /* End of innermost loop */
 522
 523         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 524                                               f+i_coord_offset,fshift+i_shift_offset);
 525
 526         /* Increment number of inner iterations */
 527         inneriter                  += j_index_end - j_index_start;
 528
 529         /* Outer loop uses 7 flops */
 530     }
 531
 532     /* Increment number of outer iterations */
 533     outeriter        += nri;
 534
 535     /* Update outer/inner flops */
 536
 537     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 538 }