src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_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_ElecCoul_VdwCSTab_GeomW3P1_VF_sse4_1_double
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            CubicSplineTable
  40  * Geometry:                   Water3-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_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              vdwioffset1;
  69     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  70     int              vdwioffset2;
  71     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  72     int              vdwjidx0A,vdwjidx0B;
  73     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  74     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  75     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  76     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  77     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  78     real             *charge;
  79     int              nvdwtype;
  80     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  81     int              *vdwtype;
  82     real             *vdwparam;
  83     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  84     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  85     __m128i          vfitab;
  86     __m128i          ifour       = _mm_set1_epi32(4);
  87     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
  88     real             *vftab;
  89     __m128d          dummy_mask,cutoff_mask;
  90     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  91     __m128d          one     = _mm_set1_pd(1.0);
  92     __m128d          two     = _mm_set1_pd(2.0);
  93     x                = xx[0];
  94     f                = ff[0];
  95
  96     nri              = nlist->nri;
  97     iinr             = nlist->iinr;
  98     jindex           = nlist->jindex;
  99     jjnr             = nlist->jjnr;
 100     shiftidx         = nlist->shift;
 101     gid              = nlist->gid;
 102     shiftvec         = fr->shift_vec[0];
 103     fshift           = fr->fshift[0];
 104     facel            = _mm_set1_pd(fr->epsfac);
 105     charge           = mdatoms->chargeA;
 106     nvdwtype         = fr->ntype;
 107     vdwparam         = fr->nbfp;
 108     vdwtype          = mdatoms->typeA;
 109
 110     vftab            = kernel_data->table_vdw->data;
 111     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 112
 113     /* Setup water-specific parameters */
 114     inr              = nlist->iinr[0];
 115     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 116     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 117     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 118     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 119
 120     /* Avoid stupid compiler warnings */
 121     jnrA = jnrB = 0;
 122     j_coord_offsetA = 0;
 123     j_coord_offsetB = 0;
 124
 125     outeriter        = 0;
 126     inneriter        = 0;
 127
 128     /* Start outer loop over neighborlists */
 129     for(iidx=0; iidx<nri; iidx++)
 130     {
 131         /* Load shift vector for this list */
 132         i_shift_offset   = DIM*shiftidx[iidx];
 133
 134         /* Load limits for loop over neighbors */
 135         j_index_start    = jindex[iidx];
 136         j_index_end      = jindex[iidx+1];
 137
 138         /* Get outer coordinate index */
 139         inr              = iinr[iidx];
 140         i_coord_offset   = DIM*inr;
 141
 142         /* Load i particle coords and add shift vector */
 143         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 144                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 145
 146         fix0             = _mm_setzero_pd();
 147         fiy0             = _mm_setzero_pd();
 148         fiz0             = _mm_setzero_pd();
 149         fix1             = _mm_setzero_pd();
 150         fiy1             = _mm_setzero_pd();
 151         fiz1             = _mm_setzero_pd();
 152         fix2             = _mm_setzero_pd();
 153         fiy2             = _mm_setzero_pd();
 154         fiz2             = _mm_setzero_pd();
 155
 156         /* Reset potential sums */
 157         velecsum         = _mm_setzero_pd();
 158         vvdwsum          = _mm_setzero_pd();
 159
 160         /* Start inner kernel loop */
 161         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 162         {
 163
 164             /* Get j neighbor index, and coordinate index */
 165             jnrA             = jjnr[jidx];
 166             jnrB             = jjnr[jidx+1];
 167             j_coord_offsetA  = DIM*jnrA;
 168             j_coord_offsetB  = DIM*jnrB;
 169
 170             /* load j atom coordinates */
 171             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 172                                               &jx0,&jy0,&jz0);
 173
 174             /* Calculate displacement vector */
 175             dx00             = _mm_sub_pd(ix0,jx0);
 176             dy00             = _mm_sub_pd(iy0,jy0);
 177             dz00             = _mm_sub_pd(iz0,jz0);
 178             dx10             = _mm_sub_pd(ix1,jx0);
 179             dy10             = _mm_sub_pd(iy1,jy0);
 180             dz10             = _mm_sub_pd(iz1,jz0);
 181             dx20             = _mm_sub_pd(ix2,jx0);
 182             dy20             = _mm_sub_pd(iy2,jy0);
 183             dz20             = _mm_sub_pd(iz2,jz0);
 184
 185             /* Calculate squared distance and things based on it */
 186             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 187             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 188             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 189
 190             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 191             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 192             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 193
 194             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 195             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 196             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 197
 198             /* Load parameters for j particles */
 199             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 200             vdwjidx0A        = 2*vdwtype[jnrA+0];
 201             vdwjidx0B        = 2*vdwtype[jnrB+0];
 202
 203             fjx0             = _mm_setzero_pd();
 204             fjy0             = _mm_setzero_pd();
 205             fjz0             = _mm_setzero_pd();
 206
 207             /**************************
 208              * CALCULATE INTERACTIONS *
 209              **************************/
 210
 211             r00              = _mm_mul_pd(rsq00,rinv00);
 212
 213             /* Compute parameters for interactions between i and j atoms */
 214             qq00             = _mm_mul_pd(iq0,jq0);
 215             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 216                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 217
 218             /* Calculate table index by multiplying r with table scale and truncate to integer */
 219             rt               = _mm_mul_pd(r00,vftabscale);
 220             vfitab           = _mm_cvttpd_epi32(rt);
 221             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 222             vfitab           = _mm_slli_epi32(vfitab,3);
 223
 224             /* COULOMB ELECTROSTATICS */
 225             velec            = _mm_mul_pd(qq00,rinv00);
 226             felec            = _mm_mul_pd(velec,rinvsq00);
 227
 228             /* CUBIC SPLINE TABLE DISPERSION */
 229             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 230             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 231             GMX_MM_TRANSPOSE2_PD(Y,F);
 232             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 233             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 234             GMX_MM_TRANSPOSE2_PD(G,H);
 235             Heps             = _mm_mul_pd(vfeps,H);
 236             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 237             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 238             vvdw6            = _mm_mul_pd(c6_00,VV);
 239             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 240             fvdw6            = _mm_mul_pd(c6_00,FF);
 241
 242             /* CUBIC SPLINE TABLE REPULSION */
 243             vfitab           = _mm_add_epi32(vfitab,ifour);
 244             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 245             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 246             GMX_MM_TRANSPOSE2_PD(Y,F);
 247             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 248             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 249             GMX_MM_TRANSPOSE2_PD(G,H);
 250             Heps             = _mm_mul_pd(vfeps,H);
 251             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 252             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 253             vvdw12           = _mm_mul_pd(c12_00,VV);
 254             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 255             fvdw12           = _mm_mul_pd(c12_00,FF);
 256             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 257             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 258
 259             /* Update potential sum for this i atom from the interaction with this j atom. */
 260             velecsum         = _mm_add_pd(velecsum,velec);
 261             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 262
 263             fscal            = _mm_add_pd(felec,fvdw);
 264
 265             /* Calculate temporary vectorial force */
 266             tx               = _mm_mul_pd(fscal,dx00);
 267             ty               = _mm_mul_pd(fscal,dy00);
 268             tz               = _mm_mul_pd(fscal,dz00);
 269
 270             /* Update vectorial force */
 271             fix0             = _mm_add_pd(fix0,tx);
 272             fiy0             = _mm_add_pd(fiy0,ty);
 273             fiz0             = _mm_add_pd(fiz0,tz);
 274
 275             fjx0             = _mm_add_pd(fjx0,tx);
 276             fjy0             = _mm_add_pd(fjy0,ty);
 277             fjz0             = _mm_add_pd(fjz0,tz);
 278
 279             /**************************
 280              * CALCULATE INTERACTIONS *
 281              **************************/
 282
 283             /* Compute parameters for interactions between i and j atoms */
 284             qq10             = _mm_mul_pd(iq1,jq0);
 285
 286             /* COULOMB ELECTROSTATICS */
 287             velec            = _mm_mul_pd(qq10,rinv10);
 288             felec            = _mm_mul_pd(velec,rinvsq10);
 289
 290             /* Update potential sum for this i atom from the interaction with this j atom. */
 291             velecsum         = _mm_add_pd(velecsum,velec);
 292
 293             fscal            = felec;
 294
 295             /* Calculate temporary vectorial force */
 296             tx               = _mm_mul_pd(fscal,dx10);
 297             ty               = _mm_mul_pd(fscal,dy10);
 298             tz               = _mm_mul_pd(fscal,dz10);
 299
 300             /* Update vectorial force */
 301             fix1             = _mm_add_pd(fix1,tx);
 302             fiy1             = _mm_add_pd(fiy1,ty);
 303             fiz1             = _mm_add_pd(fiz1,tz);
 304
 305             fjx0             = _mm_add_pd(fjx0,tx);
 306             fjy0             = _mm_add_pd(fjy0,ty);
 307             fjz0             = _mm_add_pd(fjz0,tz);
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             /* Compute parameters for interactions between i and j atoms */
 314             qq20             = _mm_mul_pd(iq2,jq0);
 315
 316             /* COULOMB ELECTROSTATICS */
 317             velec            = _mm_mul_pd(qq20,rinv20);
 318             felec            = _mm_mul_pd(velec,rinvsq20);
 319
 320             /* Update potential sum for this i atom from the interaction with this j atom. */
 321             velecsum         = _mm_add_pd(velecsum,velec);
 322
 323             fscal            = felec;
 324
 325             /* Calculate temporary vectorial force */
 326             tx               = _mm_mul_pd(fscal,dx20);
 327             ty               = _mm_mul_pd(fscal,dy20);
 328             tz               = _mm_mul_pd(fscal,dz20);
 329
 330             /* Update vectorial force */
 331             fix2             = _mm_add_pd(fix2,tx);
 332             fiy2             = _mm_add_pd(fiy2,ty);
 333             fiz2             = _mm_add_pd(fiz2,tz);
 334
 335             fjx0             = _mm_add_pd(fjx0,tx);
 336             fjy0             = _mm_add_pd(fjy0,ty);
 337             fjz0             = _mm_add_pd(fjz0,tz);
 338
 339             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 340
 341             /* Inner loop uses 122 flops */
 342         }
 343
 344         if(jidx<j_index_end)
 345         {
 346
 347             jnrA             = jjnr[jidx];
 348             j_coord_offsetA  = DIM*jnrA;
 349
 350             /* load j atom coordinates */
 351             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 352                                               &jx0,&jy0,&jz0);
 353
 354             /* Calculate displacement vector */
 355             dx00             = _mm_sub_pd(ix0,jx0);
 356             dy00             = _mm_sub_pd(iy0,jy0);
 357             dz00             = _mm_sub_pd(iz0,jz0);
 358             dx10             = _mm_sub_pd(ix1,jx0);
 359             dy10             = _mm_sub_pd(iy1,jy0);
 360             dz10             = _mm_sub_pd(iz1,jz0);
 361             dx20             = _mm_sub_pd(ix2,jx0);
 362             dy20             = _mm_sub_pd(iy2,jy0);
 363             dz20             = _mm_sub_pd(iz2,jz0);
 364
 365             /* Calculate squared distance and things based on it */
 366             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 367             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 368             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 369
 370             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 371             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 372             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 373
 374             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 375             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 376             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 377
 378             /* Load parameters for j particles */
 379             jq0              = _mm_load_sd(charge+jnrA+0);
 380             vdwjidx0A        = 2*vdwtype[jnrA+0];
 381
 382             fjx0             = _mm_setzero_pd();
 383             fjy0             = _mm_setzero_pd();
 384             fjz0             = _mm_setzero_pd();
 385
 386             /**************************
 387              * CALCULATE INTERACTIONS *
 388              **************************/
 389
 390             r00              = _mm_mul_pd(rsq00,rinv00);
 391
 392             /* Compute parameters for interactions between i and j atoms */
 393             qq00             = _mm_mul_pd(iq0,jq0);
 394             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 395
 396             /* Calculate table index by multiplying r with table scale and truncate to integer */
 397             rt               = _mm_mul_pd(r00,vftabscale);
 398             vfitab           = _mm_cvttpd_epi32(rt);
 399             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 400             vfitab           = _mm_slli_epi32(vfitab,3);
 401
 402             /* COULOMB ELECTROSTATICS */
 403             velec            = _mm_mul_pd(qq00,rinv00);
 404             felec            = _mm_mul_pd(velec,rinvsq00);
 405
 406             /* CUBIC SPLINE TABLE DISPERSION */
 407             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 408             F                = _mm_setzero_pd();
 409             GMX_MM_TRANSPOSE2_PD(Y,F);
 410             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 411             H                = _mm_setzero_pd();
 412             GMX_MM_TRANSPOSE2_PD(G,H);
 413             Heps             = _mm_mul_pd(vfeps,H);
 414             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 415             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 416             vvdw6            = _mm_mul_pd(c6_00,VV);
 417             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 418             fvdw6            = _mm_mul_pd(c6_00,FF);
 419
 420             /* CUBIC SPLINE TABLE REPULSION */
 421             vfitab           = _mm_add_epi32(vfitab,ifour);
 422             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 423             F                = _mm_setzero_pd();
 424             GMX_MM_TRANSPOSE2_PD(Y,F);
 425             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 426             H                = _mm_setzero_pd();
 427             GMX_MM_TRANSPOSE2_PD(G,H);
 428             Heps             = _mm_mul_pd(vfeps,H);
 429             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 430             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 431             vvdw12           = _mm_mul_pd(c12_00,VV);
 432             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 433             fvdw12           = _mm_mul_pd(c12_00,FF);
 434             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 435             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 436
 437             /* Update potential sum for this i atom from the interaction with this j atom. */
 438             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 439             velecsum         = _mm_add_pd(velecsum,velec);
 440             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 441             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 442
 443             fscal            = _mm_add_pd(felec,fvdw);
 444
 445             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 446
 447             /* Calculate temporary vectorial force */
 448             tx               = _mm_mul_pd(fscal,dx00);
 449             ty               = _mm_mul_pd(fscal,dy00);
 450             tz               = _mm_mul_pd(fscal,dz00);
 451
 452             /* Update vectorial force */
 453             fix0             = _mm_add_pd(fix0,tx);
 454             fiy0             = _mm_add_pd(fiy0,ty);
 455             fiz0             = _mm_add_pd(fiz0,tz);
 456
 457             fjx0             = _mm_add_pd(fjx0,tx);
 458             fjy0             = _mm_add_pd(fjy0,ty);
 459             fjz0             = _mm_add_pd(fjz0,tz);
 460
 461             /**************************
 462              * CALCULATE INTERACTIONS *
 463              **************************/
 464
 465             /* Compute parameters for interactions between i and j atoms */
 466             qq10             = _mm_mul_pd(iq1,jq0);
 467
 468             /* COULOMB ELECTROSTATICS */
 469             velec            = _mm_mul_pd(qq10,rinv10);
 470             felec            = _mm_mul_pd(velec,rinvsq10);
 471
 472             /* Update potential sum for this i atom from the interaction with this j atom. */
 473             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 474             velecsum         = _mm_add_pd(velecsum,velec);
 475
 476             fscal            = felec;
 477
 478             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 479
 480             /* Calculate temporary vectorial force */
 481             tx               = _mm_mul_pd(fscal,dx10);
 482             ty               = _mm_mul_pd(fscal,dy10);
 483             tz               = _mm_mul_pd(fscal,dz10);
 484
 485             /* Update vectorial force */
 486             fix1             = _mm_add_pd(fix1,tx);
 487             fiy1             = _mm_add_pd(fiy1,ty);
 488             fiz1             = _mm_add_pd(fiz1,tz);
 489
 490             fjx0             = _mm_add_pd(fjx0,tx);
 491             fjy0             = _mm_add_pd(fjy0,ty);
 492             fjz0             = _mm_add_pd(fjz0,tz);
 493
 494             /**************************
 495              * CALCULATE INTERACTIONS *
 496              **************************/
 497
 498             /* Compute parameters for interactions between i and j atoms */
 499             qq20             = _mm_mul_pd(iq2,jq0);
 500
 501             /* COULOMB ELECTROSTATICS */
 502             velec            = _mm_mul_pd(qq20,rinv20);
 503             felec            = _mm_mul_pd(velec,rinvsq20);
 504
 505             /* Update potential sum for this i atom from the interaction with this j atom. */
 506             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 507             velecsum         = _mm_add_pd(velecsum,velec);
 508
 509             fscal            = felec;
 510
 511             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 512
 513             /* Calculate temporary vectorial force */
 514             tx               = _mm_mul_pd(fscal,dx20);
 515             ty               = _mm_mul_pd(fscal,dy20);
 516             tz               = _mm_mul_pd(fscal,dz20);
 517
 518             /* Update vectorial force */
 519             fix2             = _mm_add_pd(fix2,tx);
 520             fiy2             = _mm_add_pd(fiy2,ty);
 521             fiz2             = _mm_add_pd(fiz2,tz);
 522
 523             fjx0             = _mm_add_pd(fjx0,tx);
 524             fjy0             = _mm_add_pd(fjy0,ty);
 525             fjz0             = _mm_add_pd(fjz0,tz);
 526
 527             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 528
 529             /* Inner loop uses 122 flops */
 530         }
 531
 532         /* End of innermost loop */
 533
 534         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 535                                               f+i_coord_offset,fshift+i_shift_offset);
 536
 537         ggid                        = gid[iidx];
 538         /* Update potential energies */
 539         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 540         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 541
 542         /* Increment number of inner iterations */
 543         inneriter                  += j_index_end - j_index_start;
 544
 545         /* Outer loop uses 20 flops */
 546     }
 547
 548     /* Increment number of outer iterations */
 549     outeriter        += nri;
 550
 551     /* Update outer/inner flops */
 552
 553     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*122);
 554 }
 555 /*
 556  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sse4_1_double
 557  * Electrostatics interaction: Coulomb
 558  * VdW interaction:            CubicSplineTable
 559  * Geometry:                   Water3-Particle
 560  * Calculate force/pot:        Force
 561  */
 562 void
 563 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sse4_1_double
 564                     (t_nblist * gmx_restrict                nlist,
 565                      rvec * gmx_restrict                    xx,
 566                      rvec * gmx_restrict                    ff,
 567                      t_forcerec * gmx_restrict              fr,
 568                      t_mdatoms * gmx_restrict               mdatoms,
 569                      nb_kernel_data_t * gmx_restrict        kernel_data,
 570                      t_nrnb * gmx_restrict                  nrnb)
 571 {
 572     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 573      * just 0 for non-waters.
 574      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 575      * jnr indices corresponding to data put in the four positions in the SIMD register.
 576      */
 577     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 578     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 579     int              jnrA,jnrB;
 580     int              j_coord_offsetA,j_coord_offsetB;
 581     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 582     real             rcutoff_scalar;
 583     real             *shiftvec,*fshift,*x,*f;
 584     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 585     int              vdwioffset0;
 586     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 587     int              vdwioffset1;
 588     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 589     int              vdwioffset2;
 590     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 591     int              vdwjidx0A,vdwjidx0B;
 592     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 593     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 594     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 595     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 596     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 597     real             *charge;
 598     int              nvdwtype;
 599     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 600     int              *vdwtype;
 601     real             *vdwparam;
 602     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 603     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 604     __m128i          vfitab;
 605     __m128i          ifour       = _mm_set1_epi32(4);
 606     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 607     real             *vftab;
 608     __m128d          dummy_mask,cutoff_mask;
 609     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 610     __m128d          one     = _mm_set1_pd(1.0);
 611     __m128d          two     = _mm_set1_pd(2.0);
 612     x                = xx[0];
 613     f                = ff[0];
 614
 615     nri              = nlist->nri;
 616     iinr             = nlist->iinr;
 617     jindex           = nlist->jindex;
 618     jjnr             = nlist->jjnr;
 619     shiftidx         = nlist->shift;
 620     gid              = nlist->gid;
 621     shiftvec         = fr->shift_vec[0];
 622     fshift           = fr->fshift[0];
 623     facel            = _mm_set1_pd(fr->epsfac);
 624     charge           = mdatoms->chargeA;
 625     nvdwtype         = fr->ntype;
 626     vdwparam         = fr->nbfp;
 627     vdwtype          = mdatoms->typeA;
 628
 629     vftab            = kernel_data->table_vdw->data;
 630     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 631
 632     /* Setup water-specific parameters */
 633     inr              = nlist->iinr[0];
 634     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 635     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 636     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 637     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 638
 639     /* Avoid stupid compiler warnings */
 640     jnrA = jnrB = 0;
 641     j_coord_offsetA = 0;
 642     j_coord_offsetB = 0;
 643
 644     outeriter        = 0;
 645     inneriter        = 0;
 646
 647     /* Start outer loop over neighborlists */
 648     for(iidx=0; iidx<nri; iidx++)
 649     {
 650         /* Load shift vector for this list */
 651         i_shift_offset   = DIM*shiftidx[iidx];
 652
 653         /* Load limits for loop over neighbors */
 654         j_index_start    = jindex[iidx];
 655         j_index_end      = jindex[iidx+1];
 656
 657         /* Get outer coordinate index */
 658         inr              = iinr[iidx];
 659         i_coord_offset   = DIM*inr;
 660
 661         /* Load i particle coords and add shift vector */
 662         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 663                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 664
 665         fix0             = _mm_setzero_pd();
 666         fiy0             = _mm_setzero_pd();
 667         fiz0             = _mm_setzero_pd();
 668         fix1             = _mm_setzero_pd();
 669         fiy1             = _mm_setzero_pd();
 670         fiz1             = _mm_setzero_pd();
 671         fix2             = _mm_setzero_pd();
 672         fiy2             = _mm_setzero_pd();
 673         fiz2             = _mm_setzero_pd();
 674
 675         /* Start inner kernel loop */
 676         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 677         {
 678
 679             /* Get j neighbor index, and coordinate index */
 680             jnrA             = jjnr[jidx];
 681             jnrB             = jjnr[jidx+1];
 682             j_coord_offsetA  = DIM*jnrA;
 683             j_coord_offsetB  = DIM*jnrB;
 684
 685             /* load j atom coordinates */
 686             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 687                                               &jx0,&jy0,&jz0);
 688
 689             /* Calculate displacement vector */
 690             dx00             = _mm_sub_pd(ix0,jx0);
 691             dy00             = _mm_sub_pd(iy0,jy0);
 692             dz00             = _mm_sub_pd(iz0,jz0);
 693             dx10             = _mm_sub_pd(ix1,jx0);
 694             dy10             = _mm_sub_pd(iy1,jy0);
 695             dz10             = _mm_sub_pd(iz1,jz0);
 696             dx20             = _mm_sub_pd(ix2,jx0);
 697             dy20             = _mm_sub_pd(iy2,jy0);
 698             dz20             = _mm_sub_pd(iz2,jz0);
 699
 700             /* Calculate squared distance and things based on it */
 701             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 702             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 703             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 704
 705             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 706             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 707             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 708
 709             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 710             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 711             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 712
 713             /* Load parameters for j particles */
 714             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 715             vdwjidx0A        = 2*vdwtype[jnrA+0];
 716             vdwjidx0B        = 2*vdwtype[jnrB+0];
 717
 718             fjx0             = _mm_setzero_pd();
 719             fjy0             = _mm_setzero_pd();
 720             fjz0             = _mm_setzero_pd();
 721
 722             /**************************
 723              * CALCULATE INTERACTIONS *
 724              **************************/
 725
 726             r00              = _mm_mul_pd(rsq00,rinv00);
 727
 728             /* Compute parameters for interactions between i and j atoms */
 729             qq00             = _mm_mul_pd(iq0,jq0);
 730             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 731                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 732
 733             /* Calculate table index by multiplying r with table scale and truncate to integer */
 734             rt               = _mm_mul_pd(r00,vftabscale);
 735             vfitab           = _mm_cvttpd_epi32(rt);
 736             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 737             vfitab           = _mm_slli_epi32(vfitab,3);
 738
 739             /* COULOMB ELECTROSTATICS */
 740             velec            = _mm_mul_pd(qq00,rinv00);
 741             felec            = _mm_mul_pd(velec,rinvsq00);
 742
 743             /* CUBIC SPLINE TABLE DISPERSION */
 744             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 745             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 746             GMX_MM_TRANSPOSE2_PD(Y,F);
 747             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 748             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 749             GMX_MM_TRANSPOSE2_PD(G,H);
 750             Heps             = _mm_mul_pd(vfeps,H);
 751             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 752             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 753             fvdw6            = _mm_mul_pd(c6_00,FF);
 754
 755             /* CUBIC SPLINE TABLE REPULSION */
 756             vfitab           = _mm_add_epi32(vfitab,ifour);
 757             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 758             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 759             GMX_MM_TRANSPOSE2_PD(Y,F);
 760             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 761             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 762             GMX_MM_TRANSPOSE2_PD(G,H);
 763             Heps             = _mm_mul_pd(vfeps,H);
 764             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 765             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 766             fvdw12           = _mm_mul_pd(c12_00,FF);
 767             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 768
 769             fscal            = _mm_add_pd(felec,fvdw);
 770
 771             /* Calculate temporary vectorial force */
 772             tx               = _mm_mul_pd(fscal,dx00);
 773             ty               = _mm_mul_pd(fscal,dy00);
 774             tz               = _mm_mul_pd(fscal,dz00);
 775
 776             /* Update vectorial force */
 777             fix0             = _mm_add_pd(fix0,tx);
 778             fiy0             = _mm_add_pd(fiy0,ty);
 779             fiz0             = _mm_add_pd(fiz0,tz);
 780
 781             fjx0             = _mm_add_pd(fjx0,tx);
 782             fjy0             = _mm_add_pd(fjy0,ty);
 783             fjz0             = _mm_add_pd(fjz0,tz);
 784
 785             /**************************
 786              * CALCULATE INTERACTIONS *
 787              **************************/
 788
 789             /* Compute parameters for interactions between i and j atoms */
 790             qq10             = _mm_mul_pd(iq1,jq0);
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm_mul_pd(qq10,rinv10);
 794             felec            = _mm_mul_pd(velec,rinvsq10);
 795
 796             fscal            = felec;
 797
 798             /* Calculate temporary vectorial force */
 799             tx               = _mm_mul_pd(fscal,dx10);
 800             ty               = _mm_mul_pd(fscal,dy10);
 801             tz               = _mm_mul_pd(fscal,dz10);
 802
 803             /* Update vectorial force */
 804             fix1             = _mm_add_pd(fix1,tx);
 805             fiy1             = _mm_add_pd(fiy1,ty);
 806             fiz1             = _mm_add_pd(fiz1,tz);
 807
 808             fjx0             = _mm_add_pd(fjx0,tx);
 809             fjy0             = _mm_add_pd(fjy0,ty);
 810             fjz0             = _mm_add_pd(fjz0,tz);
 811
 812             /**************************
 813              * CALCULATE INTERACTIONS *
 814              **************************/
 815
 816             /* Compute parameters for interactions between i and j atoms */
 817             qq20             = _mm_mul_pd(iq2,jq0);
 818
 819             /* COULOMB ELECTROSTATICS */
 820             velec            = _mm_mul_pd(qq20,rinv20);
 821             felec            = _mm_mul_pd(velec,rinvsq20);
 822
 823             fscal            = felec;
 824
 825             /* Calculate temporary vectorial force */
 826             tx               = _mm_mul_pd(fscal,dx20);
 827             ty               = _mm_mul_pd(fscal,dy20);
 828             tz               = _mm_mul_pd(fscal,dz20);
 829
 830             /* Update vectorial force */
 831             fix2             = _mm_add_pd(fix2,tx);
 832             fiy2             = _mm_add_pd(fiy2,ty);
 833             fiz2             = _mm_add_pd(fiz2,tz);
 834
 835             fjx0             = _mm_add_pd(fjx0,tx);
 836             fjy0             = _mm_add_pd(fjy0,ty);
 837             fjz0             = _mm_add_pd(fjz0,tz);
 838
 839             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 840
 841             /* Inner loop uses 111 flops */
 842         }
 843
 844         if(jidx<j_index_end)
 845         {
 846
 847             jnrA             = jjnr[jidx];
 848             j_coord_offsetA  = DIM*jnrA;
 849
 850             /* load j atom coordinates */
 851             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 852                                               &jx0,&jy0,&jz0);
 853
 854             /* Calculate displacement vector */
 855             dx00             = _mm_sub_pd(ix0,jx0);
 856             dy00             = _mm_sub_pd(iy0,jy0);
 857             dz00             = _mm_sub_pd(iz0,jz0);
 858             dx10             = _mm_sub_pd(ix1,jx0);
 859             dy10             = _mm_sub_pd(iy1,jy0);
 860             dz10             = _mm_sub_pd(iz1,jz0);
 861             dx20             = _mm_sub_pd(ix2,jx0);
 862             dy20             = _mm_sub_pd(iy2,jy0);
 863             dz20             = _mm_sub_pd(iz2,jz0);
 864
 865             /* Calculate squared distance and things based on it */
 866             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 867             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 868             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 869
 870             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 871             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 872             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 873
 874             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 875             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 876             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 877
 878             /* Load parameters for j particles */
 879             jq0              = _mm_load_sd(charge+jnrA+0);
 880             vdwjidx0A        = 2*vdwtype[jnrA+0];
 881
 882             fjx0             = _mm_setzero_pd();
 883             fjy0             = _mm_setzero_pd();
 884             fjz0             = _mm_setzero_pd();
 885
 886             /**************************
 887              * CALCULATE INTERACTIONS *
 888              **************************/
 889
 890             r00              = _mm_mul_pd(rsq00,rinv00);
 891
 892             /* Compute parameters for interactions between i and j atoms */
 893             qq00             = _mm_mul_pd(iq0,jq0);
 894             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 895
 896             /* Calculate table index by multiplying r with table scale and truncate to integer */
 897             rt               = _mm_mul_pd(r00,vftabscale);
 898             vfitab           = _mm_cvttpd_epi32(rt);
 899             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 900             vfitab           = _mm_slli_epi32(vfitab,3);
 901
 902             /* COULOMB ELECTROSTATICS */
 903             velec            = _mm_mul_pd(qq00,rinv00);
 904             felec            = _mm_mul_pd(velec,rinvsq00);
 905
 906             /* CUBIC SPLINE TABLE DISPERSION */
 907             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 908             F                = _mm_setzero_pd();
 909             GMX_MM_TRANSPOSE2_PD(Y,F);
 910             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 911             H                = _mm_setzero_pd();
 912             GMX_MM_TRANSPOSE2_PD(G,H);
 913             Heps             = _mm_mul_pd(vfeps,H);
 914             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 915             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 916             fvdw6            = _mm_mul_pd(c6_00,FF);
 917
 918             /* CUBIC SPLINE TABLE REPULSION */
 919             vfitab           = _mm_add_epi32(vfitab,ifour);
 920             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 921             F                = _mm_setzero_pd();
 922             GMX_MM_TRANSPOSE2_PD(Y,F);
 923             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 924             H                = _mm_setzero_pd();
 925             GMX_MM_TRANSPOSE2_PD(G,H);
 926             Heps             = _mm_mul_pd(vfeps,H);
 927             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 928             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 929             fvdw12           = _mm_mul_pd(c12_00,FF);
 930             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 931
 932             fscal            = _mm_add_pd(felec,fvdw);
 933
 934             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 935
 936             /* Calculate temporary vectorial force */
 937             tx               = _mm_mul_pd(fscal,dx00);
 938             ty               = _mm_mul_pd(fscal,dy00);
 939             tz               = _mm_mul_pd(fscal,dz00);
 940
 941             /* Update vectorial force */
 942             fix0             = _mm_add_pd(fix0,tx);
 943             fiy0             = _mm_add_pd(fiy0,ty);
 944             fiz0             = _mm_add_pd(fiz0,tz);
 945
 946             fjx0             = _mm_add_pd(fjx0,tx);
 947             fjy0             = _mm_add_pd(fjy0,ty);
 948             fjz0             = _mm_add_pd(fjz0,tz);
 949
 950             /**************************
 951              * CALCULATE INTERACTIONS *
 952              **************************/
 953
 954             /* Compute parameters for interactions between i and j atoms */
 955             qq10             = _mm_mul_pd(iq1,jq0);
 956
 957             /* COULOMB ELECTROSTATICS */
 958             velec            = _mm_mul_pd(qq10,rinv10);
 959             felec            = _mm_mul_pd(velec,rinvsq10);
 960
 961             fscal            = felec;
 962
 963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 964
 965             /* Calculate temporary vectorial force */
 966             tx               = _mm_mul_pd(fscal,dx10);
 967             ty               = _mm_mul_pd(fscal,dy10);
 968             tz               = _mm_mul_pd(fscal,dz10);
 969
 970             /* Update vectorial force */
 971             fix1             = _mm_add_pd(fix1,tx);
 972             fiy1             = _mm_add_pd(fiy1,ty);
 973             fiz1             = _mm_add_pd(fiz1,tz);
 974
 975             fjx0             = _mm_add_pd(fjx0,tx);
 976             fjy0             = _mm_add_pd(fjy0,ty);
 977             fjz0             = _mm_add_pd(fjz0,tz);
 978
 979             /**************************
 980              * CALCULATE INTERACTIONS *
 981              **************************/
 982
 983             /* Compute parameters for interactions between i and j atoms */
 984             qq20             = _mm_mul_pd(iq2,jq0);
 985
 986             /* COULOMB ELECTROSTATICS */
 987             velec            = _mm_mul_pd(qq20,rinv20);
 988             felec            = _mm_mul_pd(velec,rinvsq20);
 989
 990             fscal            = felec;
 991
 992             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 993
 994             /* Calculate temporary vectorial force */
 995             tx               = _mm_mul_pd(fscal,dx20);
 996             ty               = _mm_mul_pd(fscal,dy20);
 997             tz               = _mm_mul_pd(fscal,dz20);
 998
 999             /* Update vectorial force */
1000             fix2             = _mm_add_pd(fix2,tx);
1001             fiy2             = _mm_add_pd(fiy2,ty);
1002             fiz2             = _mm_add_pd(fiz2,tz);
1003
1004             fjx0             = _mm_add_pd(fjx0,tx);
1005             fjy0             = _mm_add_pd(fjy0,ty);
1006             fjz0             = _mm_add_pd(fjz0,tz);
1007
1008             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1009
1010             /* Inner loop uses 111 flops */
1011         }
1012
1013         /* End of innermost loop */
1014
1015         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1016                                               f+i_coord_offset,fshift+i_shift_offset);
1017
1018         /* Increment number of inner iterations */
1019         inneriter                  += j_index_end - j_index_start;
1020
1021         /* Outer loop uses 18 flops */
1022     }
1023
1024     /* Increment number of outer iterations */
1025     outeriter        += nri;
1026
1027     /* Update outer/inner flops */
1028
1029     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*111);
1030 }