src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRF_VdwLJ_GeomP1P1_sse4_1_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs sse4_1_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_sse4_1_double.h"
  34 #include "kernelutil_x86_sse4_1_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sse4_1_double
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRF_VdwLJ_GeomP1P1_VF_sse4_1_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     krf              = _mm_set1_pd(fr->ic->k_rf);
  97     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
  98     crf              = _mm_set1_pd(fr->ic->c_rf);
  99     nvdwtype         = fr->ntype;
 100     vdwparam         = fr->nbfp;
 101     vdwtype          = mdatoms->typeA;
 102
 103     /* Avoid stupid compiler warnings */
 104     jnrA = jnrB = 0;
 105     j_coord_offsetA = 0;
 106     j_coord_offsetB = 0;
 107
 108     outeriter        = 0;
 109     inneriter        = 0;
 110
 111     /* Start outer loop over neighborlists */
 112     for(iidx=0; iidx<nri; iidx++)
 113     {
 114         /* Load shift vector for this list */
 115         i_shift_offset   = DIM*shiftidx[iidx];
 116
 117         /* Load limits for loop over neighbors */
 118         j_index_start    = jindex[iidx];
 119         j_index_end      = jindex[iidx+1];
 120
 121         /* Get outer coordinate index */
 122         inr              = iinr[iidx];
 123         i_coord_offset   = DIM*inr;
 124
 125         /* Load i particle coords and add shift vector */
 126         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 127
 128         fix0             = _mm_setzero_pd();
 129         fiy0             = _mm_setzero_pd();
 130         fiz0             = _mm_setzero_pd();
 131
 132         /* Load parameters for i particles */
 133         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 134         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136         /* Reset potential sums */
 137         velecsum         = _mm_setzero_pd();
 138         vvdwsum          = _mm_setzero_pd();
 139
 140         /* Start inner kernel loop */
 141         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 142         {
 143
 144             /* Get j neighbor index, and coordinate index */
 145             jnrA             = jjnr[jidx];
 146             jnrB             = jjnr[jidx+1];
 147             j_coord_offsetA  = DIM*jnrA;
 148             j_coord_offsetB  = DIM*jnrB;
 149
 150             /* load j atom coordinates */
 151             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 152                                               &jx0,&jy0,&jz0);
 153
 154             /* Calculate displacement vector */
 155             dx00             = _mm_sub_pd(ix0,jx0);
 156             dy00             = _mm_sub_pd(iy0,jy0);
 157             dz00             = _mm_sub_pd(iz0,jz0);
 158
 159             /* Calculate squared distance and things based on it */
 160             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 161
 162             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 163
 164             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 165
 166             /* Load parameters for j particles */
 167             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 168             vdwjidx0A        = 2*vdwtype[jnrA+0];
 169             vdwjidx0B        = 2*vdwtype[jnrB+0];
 170
 171             /**************************
 172              * CALCULATE INTERACTIONS *
 173              **************************/
 174
 175             /* Compute parameters for interactions between i and j atoms */
 176             qq00             = _mm_mul_pd(iq0,jq0);
 177             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 178                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 179
 180             /* REACTION-FIELD ELECTROSTATICS */
 181             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 182             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 183
 184             /* LENNARD-JONES DISPERSION/REPULSION */
 185
 186             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 187             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 188             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 189             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 190             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 191
 192             /* Update potential sum for this i atom from the interaction with this j atom. */
 193             velecsum         = _mm_add_pd(velecsum,velec);
 194             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 195
 196             fscal            = _mm_add_pd(felec,fvdw);
 197
 198             /* Calculate temporary vectorial force */
 199             tx               = _mm_mul_pd(fscal,dx00);
 200             ty               = _mm_mul_pd(fscal,dy00);
 201             tz               = _mm_mul_pd(fscal,dz00);
 202
 203             /* Update vectorial force */
 204             fix0             = _mm_add_pd(fix0,tx);
 205             fiy0             = _mm_add_pd(fiy0,ty);
 206             fiz0             = _mm_add_pd(fiz0,tz);
 207
 208             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 209
 210             /* Inner loop uses 44 flops */
 211         }
 212
 213         if(jidx<j_index_end)
 214         {
 215
 216             jnrA             = jjnr[jidx];
 217             j_coord_offsetA  = DIM*jnrA;
 218
 219             /* load j atom coordinates */
 220             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 221                                               &jx0,&jy0,&jz0);
 222
 223             /* Calculate displacement vector */
 224             dx00             = _mm_sub_pd(ix0,jx0);
 225             dy00             = _mm_sub_pd(iy0,jy0);
 226             dz00             = _mm_sub_pd(iz0,jz0);
 227
 228             /* Calculate squared distance and things based on it */
 229             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 230
 231             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 232
 233             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 234
 235             /* Load parameters for j particles */
 236             jq0              = _mm_load_sd(charge+jnrA+0);
 237             vdwjidx0A        = 2*vdwtype[jnrA+0];
 238
 239             /**************************
 240              * CALCULATE INTERACTIONS *
 241              **************************/
 242
 243             /* Compute parameters for interactions between i and j atoms */
 244             qq00             = _mm_mul_pd(iq0,jq0);
 245             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 246
 247             /* REACTION-FIELD ELECTROSTATICS */
 248             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 249             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 250
 251             /* LENNARD-JONES DISPERSION/REPULSION */
 252
 253             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 254             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 255             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 256             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 257             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 258
 259             /* Update potential sum for this i atom from the interaction with this j atom. */
 260             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 261             velecsum         = _mm_add_pd(velecsum,velec);
 262             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 263             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 264
 265             fscal            = _mm_add_pd(felec,fvdw);
 266
 267             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 268
 269             /* Calculate temporary vectorial force */
 270             tx               = _mm_mul_pd(fscal,dx00);
 271             ty               = _mm_mul_pd(fscal,dy00);
 272             tz               = _mm_mul_pd(fscal,dz00);
 273
 274             /* Update vectorial force */
 275             fix0             = _mm_add_pd(fix0,tx);
 276             fiy0             = _mm_add_pd(fiy0,ty);
 277             fiz0             = _mm_add_pd(fiz0,tz);
 278
 279             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 280
 281             /* Inner loop uses 44 flops */
 282         }
 283
 284         /* End of innermost loop */
 285
 286         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 287                                               f+i_coord_offset,fshift+i_shift_offset);
 288
 289         ggid                        = gid[iidx];
 290         /* Update potential energies */
 291         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 292         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 293
 294         /* Increment number of inner iterations */
 295         inneriter                  += j_index_end - j_index_start;
 296
 297         /* Outer loop uses 9 flops */
 298     }
 299
 300     /* Increment number of outer iterations */
 301     outeriter        += nri;
 302
 303     /* Update outer/inner flops */
 304
 305     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*44);
 306 }
 307 /*
 308  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sse4_1_double
 309  * Electrostatics interaction: ReactionField
 310  * VdW interaction:            LennardJones
 311  * Geometry:                   Particle-Particle
 312  * Calculate force/pot:        Force
 313  */
 314 void
 315 nb_kernel_ElecRF_VdwLJ_GeomP1P1_F_sse4_1_double
 316                     (t_nblist * gmx_restrict                nlist,
 317                      rvec * gmx_restrict                    xx,
 318                      rvec * gmx_restrict                    ff,
 319                      t_forcerec * gmx_restrict              fr,
 320                      t_mdatoms * gmx_restrict               mdatoms,
 321                      nb_kernel_data_t * gmx_restrict        kernel_data,
 322                      t_nrnb * gmx_restrict                  nrnb)
 323 {
 324     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 325      * just 0 for non-waters.
 326      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 327      * jnr indices corresponding to data put in the four positions in the SIMD register.
 328      */
 329     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 330     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 331     int              jnrA,jnrB;
 332     int              j_coord_offsetA,j_coord_offsetB;
 333     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 334     real             rcutoff_scalar;
 335     real             *shiftvec,*fshift,*x,*f;
 336     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 337     int              vdwioffset0;
 338     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 339     int              vdwjidx0A,vdwjidx0B;
 340     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 341     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 342     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 343     real             *charge;
 344     int              nvdwtype;
 345     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 346     int              *vdwtype;
 347     real             *vdwparam;
 348     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 349     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 350     __m128d          dummy_mask,cutoff_mask;
 351     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 352     __m128d          one     = _mm_set1_pd(1.0);
 353     __m128d          two     = _mm_set1_pd(2.0);
 354     x                = xx[0];
 355     f                = ff[0];
 356
 357     nri              = nlist->nri;
 358     iinr             = nlist->iinr;
 359     jindex           = nlist->jindex;
 360     jjnr             = nlist->jjnr;
 361     shiftidx         = nlist->shift;
 362     gid              = nlist->gid;
 363     shiftvec         = fr->shift_vec[0];
 364     fshift           = fr->fshift[0];
 365     facel            = _mm_set1_pd(fr->epsfac);
 366     charge           = mdatoms->chargeA;
 367     krf              = _mm_set1_pd(fr->ic->k_rf);
 368     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 369     crf              = _mm_set1_pd(fr->ic->c_rf);
 370     nvdwtype         = fr->ntype;
 371     vdwparam         = fr->nbfp;
 372     vdwtype          = mdatoms->typeA;
 373
 374     /* Avoid stupid compiler warnings */
 375     jnrA = jnrB = 0;
 376     j_coord_offsetA = 0;
 377     j_coord_offsetB = 0;
 378
 379     outeriter        = 0;
 380     inneriter        = 0;
 381
 382     /* Start outer loop over neighborlists */
 383     for(iidx=0; iidx<nri; iidx++)
 384     {
 385         /* Load shift vector for this list */
 386         i_shift_offset   = DIM*shiftidx[iidx];
 387
 388         /* Load limits for loop over neighbors */
 389         j_index_start    = jindex[iidx];
 390         j_index_end      = jindex[iidx+1];
 391
 392         /* Get outer coordinate index */
 393         inr              = iinr[iidx];
 394         i_coord_offset   = DIM*inr;
 395
 396         /* Load i particle coords and add shift vector */
 397         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 398
 399         fix0             = _mm_setzero_pd();
 400         fiy0             = _mm_setzero_pd();
 401         fiz0             = _mm_setzero_pd();
 402
 403         /* Load parameters for i particles */
 404         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 405         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 406
 407         /* Start inner kernel loop */
 408         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 409         {
 410
 411             /* Get j neighbor index, and coordinate index */
 412             jnrA             = jjnr[jidx];
 413             jnrB             = jjnr[jidx+1];
 414             j_coord_offsetA  = DIM*jnrA;
 415             j_coord_offsetB  = DIM*jnrB;
 416
 417             /* load j atom coordinates */
 418             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 419                                               &jx0,&jy0,&jz0);
 420
 421             /* Calculate displacement vector */
 422             dx00             = _mm_sub_pd(ix0,jx0);
 423             dy00             = _mm_sub_pd(iy0,jy0);
 424             dz00             = _mm_sub_pd(iz0,jz0);
 425
 426             /* Calculate squared distance and things based on it */
 427             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 428
 429             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 430
 431             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 432
 433             /* Load parameters for j particles */
 434             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 435             vdwjidx0A        = 2*vdwtype[jnrA+0];
 436             vdwjidx0B        = 2*vdwtype[jnrB+0];
 437
 438             /**************************
 439              * CALCULATE INTERACTIONS *
 440              **************************/
 441
 442             /* Compute parameters for interactions between i and j atoms */
 443             qq00             = _mm_mul_pd(iq0,jq0);
 444             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 445                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 446
 447             /* REACTION-FIELD ELECTROSTATICS */
 448             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 449
 450             /* LENNARD-JONES DISPERSION/REPULSION */
 451
 452             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 453             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 454
 455             fscal            = _mm_add_pd(felec,fvdw);
 456
 457             /* Calculate temporary vectorial force */
 458             tx               = _mm_mul_pd(fscal,dx00);
 459             ty               = _mm_mul_pd(fscal,dy00);
 460             tz               = _mm_mul_pd(fscal,dz00);
 461
 462             /* Update vectorial force */
 463             fix0             = _mm_add_pd(fix0,tx);
 464             fiy0             = _mm_add_pd(fiy0,ty);
 465             fiz0             = _mm_add_pd(fiz0,tz);
 466
 467             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 468
 469             /* Inner loop uses 34 flops */
 470         }
 471
 472         if(jidx<j_index_end)
 473         {
 474
 475             jnrA             = jjnr[jidx];
 476             j_coord_offsetA  = DIM*jnrA;
 477
 478             /* load j atom coordinates */
 479             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 480                                               &jx0,&jy0,&jz0);
 481
 482             /* Calculate displacement vector */
 483             dx00             = _mm_sub_pd(ix0,jx0);
 484             dy00             = _mm_sub_pd(iy0,jy0);
 485             dz00             = _mm_sub_pd(iz0,jz0);
 486
 487             /* Calculate squared distance and things based on it */
 488             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 489
 490             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 491
 492             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 493
 494             /* Load parameters for j particles */
 495             jq0              = _mm_load_sd(charge+jnrA+0);
 496             vdwjidx0A        = 2*vdwtype[jnrA+0];
 497
 498             /**************************
 499              * CALCULATE INTERACTIONS *
 500              **************************/
 501
 502             /* Compute parameters for interactions between i and j atoms */
 503             qq00             = _mm_mul_pd(iq0,jq0);
 504             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 505
 506             /* REACTION-FIELD ELECTROSTATICS */
 507             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 508
 509             /* LENNARD-JONES DISPERSION/REPULSION */
 510
 511             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 512             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 513
 514             fscal            = _mm_add_pd(felec,fvdw);
 515
 516             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 517
 518             /* Calculate temporary vectorial force */
 519             tx               = _mm_mul_pd(fscal,dx00);
 520             ty               = _mm_mul_pd(fscal,dy00);
 521             tz               = _mm_mul_pd(fscal,dz00);
 522
 523             /* Update vectorial force */
 524             fix0             = _mm_add_pd(fix0,tx);
 525             fiy0             = _mm_add_pd(fiy0,ty);
 526             fiz0             = _mm_add_pd(fiz0,tz);
 527
 528             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 529
 530             /* Inner loop uses 34 flops */
 531         }
 532
 533         /* End of innermost loop */
 534
 535         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 536                                               f+i_coord_offset,fshift+i_shift_offset);
 537
 538         /* Increment number of inner iterations */
 539         inneriter                  += j_index_end - j_index_start;
 540
 541         /* Outer loop uses 7 flops */
 542     }
 543
 544     /* Increment number of outer iterations */
 545     outeriter        += nri;
 546
 547     /* Update outer/inner flops */
 548
 549     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*34);
 550 }