src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEw_VdwCSTab_GeomW4P1_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4P1_VF_sse4_1_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water4-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwCSTab_GeomW4P1_VF_sse4_1_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B;
  91     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  94     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  95     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  96     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     int              nvdwtype;
  99     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 100     int              *vdwtype;
 101     real             *vdwparam;
 102     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 103     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 104     __m128i          vfitab;
 105     __m128i          ifour       = _mm_set1_epi32(4);
 106     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 107     real             *vftab;
 108     __m128i          ewitab;
 109     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 110     real             *ewtab;
 111     __m128d          dummy_mask,cutoff_mask;
 112     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 113     __m128d          one     = _mm_set1_pd(1.0);
 114     __m128d          two     = _mm_set1_pd(2.0);
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = _mm_set1_pd(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128     nvdwtype         = fr->ntype;
 129     vdwparam         = fr->nbfp;
 130     vdwtype          = mdatoms->typeA;
 131
 132     vftab            = kernel_data->table_vdw->data;
 133     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 134
 135     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 136     ewtab            = fr->ic->tabq_coul_FDV0;
 137     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 138     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 143     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 144     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     /* Avoid stupid compiler warnings */
 148     jnrA = jnrB = 0;
 149     j_coord_offsetA = 0;
 150     j_coord_offsetB = 0;
 151
 152     outeriter        = 0;
 153     inneriter        = 0;
 154
 155     /* Start outer loop over neighborlists */
 156     for(iidx=0; iidx<nri; iidx++)
 157     {
 158         /* Load shift vector for this list */
 159         i_shift_offset   = DIM*shiftidx[iidx];
 160
 161         /* Load limits for loop over neighbors */
 162         j_index_start    = jindex[iidx];
 163         j_index_end      = jindex[iidx+1];
 164
 165         /* Get outer coordinate index */
 166         inr              = iinr[iidx];
 167         i_coord_offset   = DIM*inr;
 168
 169         /* Load i particle coords and add shift vector */
 170         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 171                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 172
 173         fix0             = _mm_setzero_pd();
 174         fiy0             = _mm_setzero_pd();
 175         fiz0             = _mm_setzero_pd();
 176         fix1             = _mm_setzero_pd();
 177         fiy1             = _mm_setzero_pd();
 178         fiz1             = _mm_setzero_pd();
 179         fix2             = _mm_setzero_pd();
 180         fiy2             = _mm_setzero_pd();
 181         fiz2             = _mm_setzero_pd();
 182         fix3             = _mm_setzero_pd();
 183         fiy3             = _mm_setzero_pd();
 184         fiz3             = _mm_setzero_pd();
 185
 186         /* Reset potential sums */
 187         velecsum         = _mm_setzero_pd();
 188         vvdwsum          = _mm_setzero_pd();
 189
 190         /* Start inner kernel loop */
 191         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 192         {
 193
 194             /* Get j neighbor index, and coordinate index */
 195             jnrA             = jjnr[jidx];
 196             jnrB             = jjnr[jidx+1];
 197             j_coord_offsetA  = DIM*jnrA;
 198             j_coord_offsetB  = DIM*jnrB;
 199
 200             /* load j atom coordinates */
 201             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 202                                               &jx0,&jy0,&jz0);
 203
 204             /* Calculate displacement vector */
 205             dx00             = _mm_sub_pd(ix0,jx0);
 206             dy00             = _mm_sub_pd(iy0,jy0);
 207             dz00             = _mm_sub_pd(iz0,jz0);
 208             dx10             = _mm_sub_pd(ix1,jx0);
 209             dy10             = _mm_sub_pd(iy1,jy0);
 210             dz10             = _mm_sub_pd(iz1,jz0);
 211             dx20             = _mm_sub_pd(ix2,jx0);
 212             dy20             = _mm_sub_pd(iy2,jy0);
 213             dz20             = _mm_sub_pd(iz2,jz0);
 214             dx30             = _mm_sub_pd(ix3,jx0);
 215             dy30             = _mm_sub_pd(iy3,jy0);
 216             dz30             = _mm_sub_pd(iz3,jz0);
 217
 218             /* Calculate squared distance and things based on it */
 219             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 220             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 221             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 222             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 223
 224             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 225             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 226             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 227             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 228
 229             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 230             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 231             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 232
 233             /* Load parameters for j particles */
 234             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 235             vdwjidx0A        = 2*vdwtype[jnrA+0];
 236             vdwjidx0B        = 2*vdwtype[jnrB+0];
 237
 238             fjx0             = _mm_setzero_pd();
 239             fjy0             = _mm_setzero_pd();
 240             fjz0             = _mm_setzero_pd();
 241
 242             /**************************
 243              * CALCULATE INTERACTIONS *
 244              **************************/
 245
 246             r00              = _mm_mul_pd(rsq00,rinv00);
 247
 248             /* Compute parameters for interactions between i and j atoms */
 249             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 250                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 251
 252             /* Calculate table index by multiplying r with table scale and truncate to integer */
 253             rt               = _mm_mul_pd(r00,vftabscale);
 254             vfitab           = _mm_cvttpd_epi32(rt);
 255             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 256             vfitab           = _mm_slli_epi32(vfitab,3);
 257
 258             /* CUBIC SPLINE TABLE DISPERSION */
 259             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 260             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 261             GMX_MM_TRANSPOSE2_PD(Y,F);
 262             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 263             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 264             GMX_MM_TRANSPOSE2_PD(G,H);
 265             Heps             = _mm_mul_pd(vfeps,H);
 266             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 267             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 268             vvdw6            = _mm_mul_pd(c6_00,VV);
 269             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 270             fvdw6            = _mm_mul_pd(c6_00,FF);
 271
 272             /* CUBIC SPLINE TABLE REPULSION */
 273             vfitab           = _mm_add_epi32(vfitab,ifour);
 274             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 275             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 276             GMX_MM_TRANSPOSE2_PD(Y,F);
 277             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 278             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 279             GMX_MM_TRANSPOSE2_PD(G,H);
 280             Heps             = _mm_mul_pd(vfeps,H);
 281             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 282             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 283             vvdw12           = _mm_mul_pd(c12_00,VV);
 284             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 285             fvdw12           = _mm_mul_pd(c12_00,FF);
 286             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 287             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 291
 292             fscal            = fvdw;
 293
 294             /* Calculate temporary vectorial force */
 295             tx               = _mm_mul_pd(fscal,dx00);
 296             ty               = _mm_mul_pd(fscal,dy00);
 297             tz               = _mm_mul_pd(fscal,dz00);
 298
 299             /* Update vectorial force */
 300             fix0             = _mm_add_pd(fix0,tx);
 301             fiy0             = _mm_add_pd(fiy0,ty);
 302             fiz0             = _mm_add_pd(fiz0,tz);
 303
 304             fjx0             = _mm_add_pd(fjx0,tx);
 305             fjy0             = _mm_add_pd(fjy0,ty);
 306             fjz0             = _mm_add_pd(fjz0,tz);
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             r10              = _mm_mul_pd(rsq10,rinv10);
 313
 314             /* Compute parameters for interactions between i and j atoms */
 315             qq10             = _mm_mul_pd(iq1,jq0);
 316
 317             /* EWALD ELECTROSTATICS */
 318
 319             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 320             ewrt             = _mm_mul_pd(r10,ewtabscale);
 321             ewitab           = _mm_cvttpd_epi32(ewrt);
 322             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 323             ewitab           = _mm_slli_epi32(ewitab,2);
 324             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 325             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 326             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 327             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 328             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 329             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 330             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 331             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 332             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 333             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 334
 335             /* Update potential sum for this i atom from the interaction with this j atom. */
 336             velecsum         = _mm_add_pd(velecsum,velec);
 337
 338             fscal            = felec;
 339
 340             /* Calculate temporary vectorial force */
 341             tx               = _mm_mul_pd(fscal,dx10);
 342             ty               = _mm_mul_pd(fscal,dy10);
 343             tz               = _mm_mul_pd(fscal,dz10);
 344
 345             /* Update vectorial force */
 346             fix1             = _mm_add_pd(fix1,tx);
 347             fiy1             = _mm_add_pd(fiy1,ty);
 348             fiz1             = _mm_add_pd(fiz1,tz);
 349
 350             fjx0             = _mm_add_pd(fjx0,tx);
 351             fjy0             = _mm_add_pd(fjy0,ty);
 352             fjz0             = _mm_add_pd(fjz0,tz);
 353
 354             /**************************
 355              * CALCULATE INTERACTIONS *
 356              **************************/
 357
 358             r20              = _mm_mul_pd(rsq20,rinv20);
 359
 360             /* Compute parameters for interactions between i and j atoms */
 361             qq20             = _mm_mul_pd(iq2,jq0);
 362
 363             /* EWALD ELECTROSTATICS */
 364
 365             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 366             ewrt             = _mm_mul_pd(r20,ewtabscale);
 367             ewitab           = _mm_cvttpd_epi32(ewrt);
 368             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 369             ewitab           = _mm_slli_epi32(ewitab,2);
 370             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 371             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 372             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 373             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 374             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 375             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 376             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 377             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 378             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 379             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 380
 381             /* Update potential sum for this i atom from the interaction with this j atom. */
 382             velecsum         = _mm_add_pd(velecsum,velec);
 383
 384             fscal            = felec;
 385
 386             /* Calculate temporary vectorial force */
 387             tx               = _mm_mul_pd(fscal,dx20);
 388             ty               = _mm_mul_pd(fscal,dy20);
 389             tz               = _mm_mul_pd(fscal,dz20);
 390
 391             /* Update vectorial force */
 392             fix2             = _mm_add_pd(fix2,tx);
 393             fiy2             = _mm_add_pd(fiy2,ty);
 394             fiz2             = _mm_add_pd(fiz2,tz);
 395
 396             fjx0             = _mm_add_pd(fjx0,tx);
 397             fjy0             = _mm_add_pd(fjy0,ty);
 398             fjz0             = _mm_add_pd(fjz0,tz);
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             r30              = _mm_mul_pd(rsq30,rinv30);
 405
 406             /* Compute parameters for interactions between i and j atoms */
 407             qq30             = _mm_mul_pd(iq3,jq0);
 408
 409             /* EWALD ELECTROSTATICS */
 410
 411             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 412             ewrt             = _mm_mul_pd(r30,ewtabscale);
 413             ewitab           = _mm_cvttpd_epi32(ewrt);
 414             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 415             ewitab           = _mm_slli_epi32(ewitab,2);
 416             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 417             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 418             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 419             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 420             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 421             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 422             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 423             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 424             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 425             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 426
 427             /* Update potential sum for this i atom from the interaction with this j atom. */
 428             velecsum         = _mm_add_pd(velecsum,velec);
 429
 430             fscal            = felec;
 431
 432             /* Calculate temporary vectorial force */
 433             tx               = _mm_mul_pd(fscal,dx30);
 434             ty               = _mm_mul_pd(fscal,dy30);
 435             tz               = _mm_mul_pd(fscal,dz30);
 436
 437             /* Update vectorial force */
 438             fix3             = _mm_add_pd(fix3,tx);
 439             fiy3             = _mm_add_pd(fiy3,ty);
 440             fiz3             = _mm_add_pd(fiz3,tz);
 441
 442             fjx0             = _mm_add_pd(fjx0,tx);
 443             fjy0             = _mm_add_pd(fjy0,ty);
 444             fjz0             = _mm_add_pd(fjz0,tz);
 445
 446             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 447
 448             /* Inner loop uses 182 flops */
 449         }
 450
 451         if(jidx<j_index_end)
 452         {
 453
 454             jnrA             = jjnr[jidx];
 455             j_coord_offsetA  = DIM*jnrA;
 456
 457             /* load j atom coordinates */
 458             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 459                                               &jx0,&jy0,&jz0);
 460
 461             /* Calculate displacement vector */
 462             dx00             = _mm_sub_pd(ix0,jx0);
 463             dy00             = _mm_sub_pd(iy0,jy0);
 464             dz00             = _mm_sub_pd(iz0,jz0);
 465             dx10             = _mm_sub_pd(ix1,jx0);
 466             dy10             = _mm_sub_pd(iy1,jy0);
 467             dz10             = _mm_sub_pd(iz1,jz0);
 468             dx20             = _mm_sub_pd(ix2,jx0);
 469             dy20             = _mm_sub_pd(iy2,jy0);
 470             dz20             = _mm_sub_pd(iz2,jz0);
 471             dx30             = _mm_sub_pd(ix3,jx0);
 472             dy30             = _mm_sub_pd(iy3,jy0);
 473             dz30             = _mm_sub_pd(iz3,jz0);
 474
 475             /* Calculate squared distance and things based on it */
 476             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 477             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 478             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 479             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 480
 481             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 482             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 483             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 484             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 485
 486             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 487             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 488             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 489
 490             /* Load parameters for j particles */
 491             jq0              = _mm_load_sd(charge+jnrA+0);
 492             vdwjidx0A        = 2*vdwtype[jnrA+0];
 493
 494             fjx0             = _mm_setzero_pd();
 495             fjy0             = _mm_setzero_pd();
 496             fjz0             = _mm_setzero_pd();
 497
 498             /**************************
 499              * CALCULATE INTERACTIONS *
 500              **************************/
 501
 502             r00              = _mm_mul_pd(rsq00,rinv00);
 503
 504             /* Compute parameters for interactions between i and j atoms */
 505             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 506
 507             /* Calculate table index by multiplying r with table scale and truncate to integer */
 508             rt               = _mm_mul_pd(r00,vftabscale);
 509             vfitab           = _mm_cvttpd_epi32(rt);
 510             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 511             vfitab           = _mm_slli_epi32(vfitab,3);
 512
 513             /* CUBIC SPLINE TABLE DISPERSION */
 514             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 515             F                = _mm_setzero_pd();
 516             GMX_MM_TRANSPOSE2_PD(Y,F);
 517             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 518             H                = _mm_setzero_pd();
 519             GMX_MM_TRANSPOSE2_PD(G,H);
 520             Heps             = _mm_mul_pd(vfeps,H);
 521             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 522             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 523             vvdw6            = _mm_mul_pd(c6_00,VV);
 524             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 525             fvdw6            = _mm_mul_pd(c6_00,FF);
 526
 527             /* CUBIC SPLINE TABLE REPULSION */
 528             vfitab           = _mm_add_epi32(vfitab,ifour);
 529             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 530             F                = _mm_setzero_pd();
 531             GMX_MM_TRANSPOSE2_PD(Y,F);
 532             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 533             H                = _mm_setzero_pd();
 534             GMX_MM_TRANSPOSE2_PD(G,H);
 535             Heps             = _mm_mul_pd(vfeps,H);
 536             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 537             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 538             vvdw12           = _mm_mul_pd(c12_00,VV);
 539             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 540             fvdw12           = _mm_mul_pd(c12_00,FF);
 541             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 542             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 546             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 547
 548             fscal            = fvdw;
 549
 550             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 551
 552             /* Calculate temporary vectorial force */
 553             tx               = _mm_mul_pd(fscal,dx00);
 554             ty               = _mm_mul_pd(fscal,dy00);
 555             tz               = _mm_mul_pd(fscal,dz00);
 556
 557             /* Update vectorial force */
 558             fix0             = _mm_add_pd(fix0,tx);
 559             fiy0             = _mm_add_pd(fiy0,ty);
 560             fiz0             = _mm_add_pd(fiz0,tz);
 561
 562             fjx0             = _mm_add_pd(fjx0,tx);
 563             fjy0             = _mm_add_pd(fjy0,ty);
 564             fjz0             = _mm_add_pd(fjz0,tz);
 565
 566             /**************************
 567              * CALCULATE INTERACTIONS *
 568              **************************/
 569
 570             r10              = _mm_mul_pd(rsq10,rinv10);
 571
 572             /* Compute parameters for interactions between i and j atoms */
 573             qq10             = _mm_mul_pd(iq1,jq0);
 574
 575             /* EWALD ELECTROSTATICS */
 576
 577             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 578             ewrt             = _mm_mul_pd(r10,ewtabscale);
 579             ewitab           = _mm_cvttpd_epi32(ewrt);
 580             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 581             ewitab           = _mm_slli_epi32(ewitab,2);
 582             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 583             ewtabD           = _mm_setzero_pd();
 584             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 585             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 586             ewtabFn          = _mm_setzero_pd();
 587             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 588             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 589             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 590             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 591             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 592
 593             /* Update potential sum for this i atom from the interaction with this j atom. */
 594             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 595             velecsum         = _mm_add_pd(velecsum,velec);
 596
 597             fscal            = felec;
 598
 599             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 600
 601             /* Calculate temporary vectorial force */
 602             tx               = _mm_mul_pd(fscal,dx10);
 603             ty               = _mm_mul_pd(fscal,dy10);
 604             tz               = _mm_mul_pd(fscal,dz10);
 605
 606             /* Update vectorial force */
 607             fix1             = _mm_add_pd(fix1,tx);
 608             fiy1             = _mm_add_pd(fiy1,ty);
 609             fiz1             = _mm_add_pd(fiz1,tz);
 610
 611             fjx0             = _mm_add_pd(fjx0,tx);
 612             fjy0             = _mm_add_pd(fjy0,ty);
 613             fjz0             = _mm_add_pd(fjz0,tz);
 614
 615             /**************************
 616              * CALCULATE INTERACTIONS *
 617              **************************/
 618
 619             r20              = _mm_mul_pd(rsq20,rinv20);
 620
 621             /* Compute parameters for interactions between i and j atoms */
 622             qq20             = _mm_mul_pd(iq2,jq0);
 623
 624             /* EWALD ELECTROSTATICS */
 625
 626             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 627             ewrt             = _mm_mul_pd(r20,ewtabscale);
 628             ewitab           = _mm_cvttpd_epi32(ewrt);
 629             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 630             ewitab           = _mm_slli_epi32(ewitab,2);
 631             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 632             ewtabD           = _mm_setzero_pd();
 633             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 634             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 635             ewtabFn          = _mm_setzero_pd();
 636             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 637             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 638             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 639             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 640             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 641
 642             /* Update potential sum for this i atom from the interaction with this j atom. */
 643             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 644             velecsum         = _mm_add_pd(velecsum,velec);
 645
 646             fscal            = felec;
 647
 648             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 649
 650             /* Calculate temporary vectorial force */
 651             tx               = _mm_mul_pd(fscal,dx20);
 652             ty               = _mm_mul_pd(fscal,dy20);
 653             tz               = _mm_mul_pd(fscal,dz20);
 654
 655             /* Update vectorial force */
 656             fix2             = _mm_add_pd(fix2,tx);
 657             fiy2             = _mm_add_pd(fiy2,ty);
 658             fiz2             = _mm_add_pd(fiz2,tz);
 659
 660             fjx0             = _mm_add_pd(fjx0,tx);
 661             fjy0             = _mm_add_pd(fjy0,ty);
 662             fjz0             = _mm_add_pd(fjz0,tz);
 663
 664             /**************************
 665              * CALCULATE INTERACTIONS *
 666              **************************/
 667
 668             r30              = _mm_mul_pd(rsq30,rinv30);
 669
 670             /* Compute parameters for interactions between i and j atoms */
 671             qq30             = _mm_mul_pd(iq3,jq0);
 672
 673             /* EWALD ELECTROSTATICS */
 674
 675             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 676             ewrt             = _mm_mul_pd(r30,ewtabscale);
 677             ewitab           = _mm_cvttpd_epi32(ewrt);
 678             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 679             ewitab           = _mm_slli_epi32(ewitab,2);
 680             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 681             ewtabD           = _mm_setzero_pd();
 682             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 683             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 684             ewtabFn          = _mm_setzero_pd();
 685             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 686             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 687             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 688             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 689             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 690
 691             /* Update potential sum for this i atom from the interaction with this j atom. */
 692             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 693             velecsum         = _mm_add_pd(velecsum,velec);
 694
 695             fscal            = felec;
 696
 697             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 698
 699             /* Calculate temporary vectorial force */
 700             tx               = _mm_mul_pd(fscal,dx30);
 701             ty               = _mm_mul_pd(fscal,dy30);
 702             tz               = _mm_mul_pd(fscal,dz30);
 703
 704             /* Update vectorial force */
 705             fix3             = _mm_add_pd(fix3,tx);
 706             fiy3             = _mm_add_pd(fiy3,ty);
 707             fiz3             = _mm_add_pd(fiz3,tz);
 708
 709             fjx0             = _mm_add_pd(fjx0,tx);
 710             fjy0             = _mm_add_pd(fjy0,ty);
 711             fjz0             = _mm_add_pd(fjz0,tz);
 712
 713             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 714
 715             /* Inner loop uses 182 flops */
 716         }
 717
 718         /* End of innermost loop */
 719
 720         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 721                                               f+i_coord_offset,fshift+i_shift_offset);
 722
 723         ggid                        = gid[iidx];
 724         /* Update potential energies */
 725         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 726         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 727
 728         /* Increment number of inner iterations */
 729         inneriter                  += j_index_end - j_index_start;
 730
 731         /* Outer loop uses 26 flops */
 732     }
 733
 734     /* Increment number of outer iterations */
 735     outeriter        += nri;
 736
 737     /* Update outer/inner flops */
 738
 739     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*182);
 740 }
 741 /*
 742  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4P1_F_sse4_1_double
 743  * Electrostatics interaction: Ewald
 744  * VdW interaction:            CubicSplineTable
 745  * Geometry:                   Water4-Particle
 746  * Calculate force/pot:        Force
 747  */
 748 void
 749 nb_kernel_ElecEw_VdwCSTab_GeomW4P1_F_sse4_1_double
 750                     (t_nblist                    * gmx_restrict       nlist,
 751                      rvec                        * gmx_restrict          xx,
 752                      rvec                        * gmx_restrict          ff,
 753                      t_forcerec                  * gmx_restrict          fr,
 754                      t_mdatoms                   * gmx_restrict     mdatoms,
 755                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 756                      t_nrnb                      * gmx_restrict        nrnb)
 757 {
 758     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 759      * just 0 for non-waters.
 760      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 761      * jnr indices corresponding to data put in the four positions in the SIMD register.
 762      */
 763     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 764     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 765     int              jnrA,jnrB;
 766     int              j_coord_offsetA,j_coord_offsetB;
 767     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 768     real             rcutoff_scalar;
 769     real             *shiftvec,*fshift,*x,*f;
 770     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 771     int              vdwioffset0;
 772     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 773     int              vdwioffset1;
 774     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 775     int              vdwioffset2;
 776     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 777     int              vdwioffset3;
 778     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 779     int              vdwjidx0A,vdwjidx0B;
 780     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 781     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 782     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 783     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 784     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 785     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 786     real             *charge;
 787     int              nvdwtype;
 788     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 789     int              *vdwtype;
 790     real             *vdwparam;
 791     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 792     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 793     __m128i          vfitab;
 794     __m128i          ifour       = _mm_set1_epi32(4);
 795     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 796     real             *vftab;
 797     __m128i          ewitab;
 798     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 799     real             *ewtab;
 800     __m128d          dummy_mask,cutoff_mask;
 801     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 802     __m128d          one     = _mm_set1_pd(1.0);
 803     __m128d          two     = _mm_set1_pd(2.0);
 804     x                = xx[0];
 805     f                = ff[0];
 806
 807     nri              = nlist->nri;
 808     iinr             = nlist->iinr;
 809     jindex           = nlist->jindex;
 810     jjnr             = nlist->jjnr;
 811     shiftidx         = nlist->shift;
 812     gid              = nlist->gid;
 813     shiftvec         = fr->shift_vec[0];
 814     fshift           = fr->fshift[0];
 815     facel            = _mm_set1_pd(fr->epsfac);
 816     charge           = mdatoms->chargeA;
 817     nvdwtype         = fr->ntype;
 818     vdwparam         = fr->nbfp;
 819     vdwtype          = mdatoms->typeA;
 820
 821     vftab            = kernel_data->table_vdw->data;
 822     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 823
 824     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 825     ewtab            = fr->ic->tabq_coul_F;
 826     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 827     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 828
 829     /* Setup water-specific parameters */
 830     inr              = nlist->iinr[0];
 831     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 832     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 833     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 834     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 835
 836     /* Avoid stupid compiler warnings */
 837     jnrA = jnrB = 0;
 838     j_coord_offsetA = 0;
 839     j_coord_offsetB = 0;
 840
 841     outeriter        = 0;
 842     inneriter        = 0;
 843
 844     /* Start outer loop over neighborlists */
 845     for(iidx=0; iidx<nri; iidx++)
 846     {
 847         /* Load shift vector for this list */
 848         i_shift_offset   = DIM*shiftidx[iidx];
 849
 850         /* Load limits for loop over neighbors */
 851         j_index_start    = jindex[iidx];
 852         j_index_end      = jindex[iidx+1];
 853
 854         /* Get outer coordinate index */
 855         inr              = iinr[iidx];
 856         i_coord_offset   = DIM*inr;
 857
 858         /* Load i particle coords and add shift vector */
 859         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 860                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 861
 862         fix0             = _mm_setzero_pd();
 863         fiy0             = _mm_setzero_pd();
 864         fiz0             = _mm_setzero_pd();
 865         fix1             = _mm_setzero_pd();
 866         fiy1             = _mm_setzero_pd();
 867         fiz1             = _mm_setzero_pd();
 868         fix2             = _mm_setzero_pd();
 869         fiy2             = _mm_setzero_pd();
 870         fiz2             = _mm_setzero_pd();
 871         fix3             = _mm_setzero_pd();
 872         fiy3             = _mm_setzero_pd();
 873         fiz3             = _mm_setzero_pd();
 874
 875         /* Start inner kernel loop */
 876         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 877         {
 878
 879             /* Get j neighbor index, and coordinate index */
 880             jnrA             = jjnr[jidx];
 881             jnrB             = jjnr[jidx+1];
 882             j_coord_offsetA  = DIM*jnrA;
 883             j_coord_offsetB  = DIM*jnrB;
 884
 885             /* load j atom coordinates */
 886             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 887                                               &jx0,&jy0,&jz0);
 888
 889             /* Calculate displacement vector */
 890             dx00             = _mm_sub_pd(ix0,jx0);
 891             dy00             = _mm_sub_pd(iy0,jy0);
 892             dz00             = _mm_sub_pd(iz0,jz0);
 893             dx10             = _mm_sub_pd(ix1,jx0);
 894             dy10             = _mm_sub_pd(iy1,jy0);
 895             dz10             = _mm_sub_pd(iz1,jz0);
 896             dx20             = _mm_sub_pd(ix2,jx0);
 897             dy20             = _mm_sub_pd(iy2,jy0);
 898             dz20             = _mm_sub_pd(iz2,jz0);
 899             dx30             = _mm_sub_pd(ix3,jx0);
 900             dy30             = _mm_sub_pd(iy3,jy0);
 901             dz30             = _mm_sub_pd(iz3,jz0);
 902
 903             /* Calculate squared distance and things based on it */
 904             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 905             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 906             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 907             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 908
 909             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 910             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 911             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 912             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 913
 914             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 915             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 916             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 917
 918             /* Load parameters for j particles */
 919             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 920             vdwjidx0A        = 2*vdwtype[jnrA+0];
 921             vdwjidx0B        = 2*vdwtype[jnrB+0];
 922
 923             fjx0             = _mm_setzero_pd();
 924             fjy0             = _mm_setzero_pd();
 925             fjz0             = _mm_setzero_pd();
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             r00              = _mm_mul_pd(rsq00,rinv00);
 932
 933             /* Compute parameters for interactions between i and j atoms */
 934             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 935                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 936
 937             /* Calculate table index by multiplying r with table scale and truncate to integer */
 938             rt               = _mm_mul_pd(r00,vftabscale);
 939             vfitab           = _mm_cvttpd_epi32(rt);
 940             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 941             vfitab           = _mm_slli_epi32(vfitab,3);
 942
 943             /* CUBIC SPLINE TABLE DISPERSION */
 944             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 945             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 946             GMX_MM_TRANSPOSE2_PD(Y,F);
 947             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 948             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 949             GMX_MM_TRANSPOSE2_PD(G,H);
 950             Heps             = _mm_mul_pd(vfeps,H);
 951             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 952             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 953             fvdw6            = _mm_mul_pd(c6_00,FF);
 954
 955             /* CUBIC SPLINE TABLE REPULSION */
 956             vfitab           = _mm_add_epi32(vfitab,ifour);
 957             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 958             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 959             GMX_MM_TRANSPOSE2_PD(Y,F);
 960             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 961             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 962             GMX_MM_TRANSPOSE2_PD(G,H);
 963             Heps             = _mm_mul_pd(vfeps,H);
 964             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 965             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 966             fvdw12           = _mm_mul_pd(c12_00,FF);
 967             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 968
 969             fscal            = fvdw;
 970
 971             /* Calculate temporary vectorial force */
 972             tx               = _mm_mul_pd(fscal,dx00);
 973             ty               = _mm_mul_pd(fscal,dy00);
 974             tz               = _mm_mul_pd(fscal,dz00);
 975
 976             /* Update vectorial force */
 977             fix0             = _mm_add_pd(fix0,tx);
 978             fiy0             = _mm_add_pd(fiy0,ty);
 979             fiz0             = _mm_add_pd(fiz0,tz);
 980
 981             fjx0             = _mm_add_pd(fjx0,tx);
 982             fjy0             = _mm_add_pd(fjy0,ty);
 983             fjz0             = _mm_add_pd(fjz0,tz);
 984
 985             /**************************
 986              * CALCULATE INTERACTIONS *
 987              **************************/
 988
 989             r10              = _mm_mul_pd(rsq10,rinv10);
 990
 991             /* Compute parameters for interactions between i and j atoms */
 992             qq10             = _mm_mul_pd(iq1,jq0);
 993
 994             /* EWALD ELECTROSTATICS */
 995
 996             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 997             ewrt             = _mm_mul_pd(r10,ewtabscale);
 998             ewitab           = _mm_cvttpd_epi32(ewrt);
 999             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1000             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1001                                          &ewtabF,&ewtabFn);
1002             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1003             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1004
1005             fscal            = felec;
1006
1007             /* Calculate temporary vectorial force */
1008             tx               = _mm_mul_pd(fscal,dx10);
1009             ty               = _mm_mul_pd(fscal,dy10);
1010             tz               = _mm_mul_pd(fscal,dz10);
1011
1012             /* Update vectorial force */
1013             fix1             = _mm_add_pd(fix1,tx);
1014             fiy1             = _mm_add_pd(fiy1,ty);
1015             fiz1             = _mm_add_pd(fiz1,tz);
1016
1017             fjx0             = _mm_add_pd(fjx0,tx);
1018             fjy0             = _mm_add_pd(fjy0,ty);
1019             fjz0             = _mm_add_pd(fjz0,tz);
1020
1021             /**************************
1022              * CALCULATE INTERACTIONS *
1023              **************************/
1024
1025             r20              = _mm_mul_pd(rsq20,rinv20);
1026
1027             /* Compute parameters for interactions between i and j atoms */
1028             qq20             = _mm_mul_pd(iq2,jq0);
1029
1030             /* EWALD ELECTROSTATICS */
1031
1032             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1033             ewrt             = _mm_mul_pd(r20,ewtabscale);
1034             ewitab           = _mm_cvttpd_epi32(ewrt);
1035             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1036             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1037                                          &ewtabF,&ewtabFn);
1038             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1039             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1040
1041             fscal            = felec;
1042
1043             /* Calculate temporary vectorial force */
1044             tx               = _mm_mul_pd(fscal,dx20);
1045             ty               = _mm_mul_pd(fscal,dy20);
1046             tz               = _mm_mul_pd(fscal,dz20);
1047
1048             /* Update vectorial force */
1049             fix2             = _mm_add_pd(fix2,tx);
1050             fiy2             = _mm_add_pd(fiy2,ty);
1051             fiz2             = _mm_add_pd(fiz2,tz);
1052
1053             fjx0             = _mm_add_pd(fjx0,tx);
1054             fjy0             = _mm_add_pd(fjy0,ty);
1055             fjz0             = _mm_add_pd(fjz0,tz);
1056
1057             /**************************
1058              * CALCULATE INTERACTIONS *
1059              **************************/
1060
1061             r30              = _mm_mul_pd(rsq30,rinv30);
1062
1063             /* Compute parameters for interactions between i and j atoms */
1064             qq30             = _mm_mul_pd(iq3,jq0);
1065
1066             /* EWALD ELECTROSTATICS */
1067
1068             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1069             ewrt             = _mm_mul_pd(r30,ewtabscale);
1070             ewitab           = _mm_cvttpd_epi32(ewrt);
1071             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1072             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1073                                          &ewtabF,&ewtabFn);
1074             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1075             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1076
1077             fscal            = felec;
1078
1079             /* Calculate temporary vectorial force */
1080             tx               = _mm_mul_pd(fscal,dx30);
1081             ty               = _mm_mul_pd(fscal,dy30);
1082             tz               = _mm_mul_pd(fscal,dz30);
1083
1084             /* Update vectorial force */
1085             fix3             = _mm_add_pd(fix3,tx);
1086             fiy3             = _mm_add_pd(fiy3,ty);
1087             fiz3             = _mm_add_pd(fiz3,tz);
1088
1089             fjx0             = _mm_add_pd(fjx0,tx);
1090             fjy0             = _mm_add_pd(fjy0,ty);
1091             fjz0             = _mm_add_pd(fjz0,tz);
1092
1093             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1094
1095             /* Inner loop uses 159 flops */
1096         }
1097
1098         if(jidx<j_index_end)
1099         {
1100
1101             jnrA             = jjnr[jidx];
1102             j_coord_offsetA  = DIM*jnrA;
1103
1104             /* load j atom coordinates */
1105             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1106                                               &jx0,&jy0,&jz0);
1107
1108             /* Calculate displacement vector */
1109             dx00             = _mm_sub_pd(ix0,jx0);
1110             dy00             = _mm_sub_pd(iy0,jy0);
1111             dz00             = _mm_sub_pd(iz0,jz0);
1112             dx10             = _mm_sub_pd(ix1,jx0);
1113             dy10             = _mm_sub_pd(iy1,jy0);
1114             dz10             = _mm_sub_pd(iz1,jz0);
1115             dx20             = _mm_sub_pd(ix2,jx0);
1116             dy20             = _mm_sub_pd(iy2,jy0);
1117             dz20             = _mm_sub_pd(iz2,jz0);
1118             dx30             = _mm_sub_pd(ix3,jx0);
1119             dy30             = _mm_sub_pd(iy3,jy0);
1120             dz30             = _mm_sub_pd(iz3,jz0);
1121
1122             /* Calculate squared distance and things based on it */
1123             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1124             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1125             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1126             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
1127
1128             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1129             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1130             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1131             rinv30           = gmx_mm_invsqrt_pd(rsq30);
1132
1133             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1134             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1135             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
1136
1137             /* Load parameters for j particles */
1138             jq0              = _mm_load_sd(charge+jnrA+0);
1139             vdwjidx0A        = 2*vdwtype[jnrA+0];
1140
1141             fjx0             = _mm_setzero_pd();
1142             fjy0             = _mm_setzero_pd();
1143             fjz0             = _mm_setzero_pd();
1144
1145             /**************************
1146              * CALCULATE INTERACTIONS *
1147              **************************/
1148
1149             r00              = _mm_mul_pd(rsq00,rinv00);
1150
1151             /* Compute parameters for interactions between i and j atoms */
1152             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1153
1154             /* Calculate table index by multiplying r with table scale and truncate to integer */
1155             rt               = _mm_mul_pd(r00,vftabscale);
1156             vfitab           = _mm_cvttpd_epi32(rt);
1157             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1158             vfitab           = _mm_slli_epi32(vfitab,3);
1159
1160             /* CUBIC SPLINE TABLE DISPERSION */
1161             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1162             F                = _mm_setzero_pd();
1163             GMX_MM_TRANSPOSE2_PD(Y,F);
1164             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1165             H                = _mm_setzero_pd();
1166             GMX_MM_TRANSPOSE2_PD(G,H);
1167             Heps             = _mm_mul_pd(vfeps,H);
1168             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1169             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1170             fvdw6            = _mm_mul_pd(c6_00,FF);
1171
1172             /* CUBIC SPLINE TABLE REPULSION */
1173             vfitab           = _mm_add_epi32(vfitab,ifour);
1174             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1175             F                = _mm_setzero_pd();
1176             GMX_MM_TRANSPOSE2_PD(Y,F);
1177             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1178             H                = _mm_setzero_pd();
1179             GMX_MM_TRANSPOSE2_PD(G,H);
1180             Heps             = _mm_mul_pd(vfeps,H);
1181             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1182             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1183             fvdw12           = _mm_mul_pd(c12_00,FF);
1184             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1185
1186             fscal            = fvdw;
1187
1188             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1189
1190             /* Calculate temporary vectorial force */
1191             tx               = _mm_mul_pd(fscal,dx00);
1192             ty               = _mm_mul_pd(fscal,dy00);
1193             tz               = _mm_mul_pd(fscal,dz00);
1194
1195             /* Update vectorial force */
1196             fix0             = _mm_add_pd(fix0,tx);
1197             fiy0             = _mm_add_pd(fiy0,ty);
1198             fiz0             = _mm_add_pd(fiz0,tz);
1199
1200             fjx0             = _mm_add_pd(fjx0,tx);
1201             fjy0             = _mm_add_pd(fjy0,ty);
1202             fjz0             = _mm_add_pd(fjz0,tz);
1203
1204             /**************************
1205              * CALCULATE INTERACTIONS *
1206              **************************/
1207
1208             r10              = _mm_mul_pd(rsq10,rinv10);
1209
1210             /* Compute parameters for interactions between i and j atoms */
1211             qq10             = _mm_mul_pd(iq1,jq0);
1212
1213             /* EWALD ELECTROSTATICS */
1214
1215             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1216             ewrt             = _mm_mul_pd(r10,ewtabscale);
1217             ewitab           = _mm_cvttpd_epi32(ewrt);
1218             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1219             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1220             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1221             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1222
1223             fscal            = felec;
1224
1225             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1226
1227             /* Calculate temporary vectorial force */
1228             tx               = _mm_mul_pd(fscal,dx10);
1229             ty               = _mm_mul_pd(fscal,dy10);
1230             tz               = _mm_mul_pd(fscal,dz10);
1231
1232             /* Update vectorial force */
1233             fix1             = _mm_add_pd(fix1,tx);
1234             fiy1             = _mm_add_pd(fiy1,ty);
1235             fiz1             = _mm_add_pd(fiz1,tz);
1236
1237             fjx0             = _mm_add_pd(fjx0,tx);
1238             fjy0             = _mm_add_pd(fjy0,ty);
1239             fjz0             = _mm_add_pd(fjz0,tz);
1240
1241             /**************************
1242              * CALCULATE INTERACTIONS *
1243              **************************/
1244
1245             r20              = _mm_mul_pd(rsq20,rinv20);
1246
1247             /* Compute parameters for interactions between i and j atoms */
1248             qq20             = _mm_mul_pd(iq2,jq0);
1249
1250             /* EWALD ELECTROSTATICS */
1251
1252             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1253             ewrt             = _mm_mul_pd(r20,ewtabscale);
1254             ewitab           = _mm_cvttpd_epi32(ewrt);
1255             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1256             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1257             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1258             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1259
1260             fscal            = felec;
1261
1262             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1263
1264             /* Calculate temporary vectorial force */
1265             tx               = _mm_mul_pd(fscal,dx20);
1266             ty               = _mm_mul_pd(fscal,dy20);
1267             tz               = _mm_mul_pd(fscal,dz20);
1268
1269             /* Update vectorial force */
1270             fix2             = _mm_add_pd(fix2,tx);
1271             fiy2             = _mm_add_pd(fiy2,ty);
1272             fiz2             = _mm_add_pd(fiz2,tz);
1273
1274             fjx0             = _mm_add_pd(fjx0,tx);
1275             fjy0             = _mm_add_pd(fjy0,ty);
1276             fjz0             = _mm_add_pd(fjz0,tz);
1277
1278             /**************************
1279              * CALCULATE INTERACTIONS *
1280              **************************/
1281
1282             r30              = _mm_mul_pd(rsq30,rinv30);
1283
1284             /* Compute parameters for interactions between i and j atoms */
1285             qq30             = _mm_mul_pd(iq3,jq0);
1286
1287             /* EWALD ELECTROSTATICS */
1288
1289             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1290             ewrt             = _mm_mul_pd(r30,ewtabscale);
1291             ewitab           = _mm_cvttpd_epi32(ewrt);
1292             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1293             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1294             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1295             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1296
1297             fscal            = felec;
1298
1299             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1300
1301             /* Calculate temporary vectorial force */
1302             tx               = _mm_mul_pd(fscal,dx30);
1303             ty               = _mm_mul_pd(fscal,dy30);
1304             tz               = _mm_mul_pd(fscal,dz30);
1305
1306             /* Update vectorial force */
1307             fix3             = _mm_add_pd(fix3,tx);
1308             fiy3             = _mm_add_pd(fiy3,ty);
1309             fiz3             = _mm_add_pd(fiz3,tz);
1310
1311             fjx0             = _mm_add_pd(fjx0,tx);
1312             fjy0             = _mm_add_pd(fjy0,ty);
1313             fjz0             = _mm_add_pd(fjz0,tz);
1314
1315             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1316
1317             /* Inner loop uses 159 flops */
1318         }
1319
1320         /* End of innermost loop */
1321
1322         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1323                                               f+i_coord_offset,fshift+i_shift_offset);
1324
1325         /* Increment number of inner iterations */
1326         inneriter                  += j_index_end - j_index_start;
1327
1328         /* Outer loop uses 24 flops */
1329     }
1330
1331     /* Increment number of outer iterations */
1332     outeriter        += nri;
1333
1334     /* Update outer/inner flops */
1335
1336     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);
1337 }