src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecEw_VdwCSTab_GeomW4P1_sse2_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 sse2_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse2_double.h"
  48 #include "kernelutil_x86_sse2_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4P1_VF_sse2_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwCSTab_GeomW4P1_VF_sse2_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwioffset3;
  87     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  88     int              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  91     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  92     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  93     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  94     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  95     real             *charge;
  96     int              nvdwtype;
  97     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  98     int              *vdwtype;
  99     real             *vdwparam;
 100     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 101     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 102     __m128i          vfitab;
 103     __m128i          ifour       = _mm_set1_epi32(4);
 104     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 105     real             *vftab;
 106     __m128i          ewitab;
 107     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 108     real             *ewtab;
 109     __m128d          dummy_mask,cutoff_mask;
 110     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 111     __m128d          one     = _mm_set1_pd(1.0);
 112     __m128d          two     = _mm_set1_pd(2.0);
 113     x                = xx[0];
 114     f                = ff[0];
 115
 116     nri              = nlist->nri;
 117     iinr             = nlist->iinr;
 118     jindex           = nlist->jindex;
 119     jjnr             = nlist->jjnr;
 120     shiftidx         = nlist->shift;
 121     gid              = nlist->gid;
 122     shiftvec         = fr->shift_vec[0];
 123     fshift           = fr->fshift[0];
 124     facel            = _mm_set1_pd(fr->epsfac);
 125     charge           = mdatoms->chargeA;
 126     nvdwtype         = fr->ntype;
 127     vdwparam         = fr->nbfp;
 128     vdwtype          = mdatoms->typeA;
 129
 130     vftab            = kernel_data->table_vdw->data;
 131     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 132
 133     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 134     ewtab            = fr->ic->tabq_coul_FDV0;
 135     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 136     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 137
 138     /* Setup water-specific parameters */
 139     inr              = nlist->iinr[0];
 140     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 141     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 142     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 143     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 144
 145     /* Avoid stupid compiler warnings */
 146     jnrA = jnrB = 0;
 147     j_coord_offsetA = 0;
 148     j_coord_offsetB = 0;
 149
 150     outeriter        = 0;
 151     inneriter        = 0;
 152
 153     /* Start outer loop over neighborlists */
 154     for(iidx=0; iidx<nri; iidx++)
 155     {
 156         /* Load shift vector for this list */
 157         i_shift_offset   = DIM*shiftidx[iidx];
 158
 159         /* Load limits for loop over neighbors */
 160         j_index_start    = jindex[iidx];
 161         j_index_end      = jindex[iidx+1];
 162
 163         /* Get outer coordinate index */
 164         inr              = iinr[iidx];
 165         i_coord_offset   = DIM*inr;
 166
 167         /* Load i particle coords and add shift vector */
 168         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 169                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 170
 171         fix0             = _mm_setzero_pd();
 172         fiy0             = _mm_setzero_pd();
 173         fiz0             = _mm_setzero_pd();
 174         fix1             = _mm_setzero_pd();
 175         fiy1             = _mm_setzero_pd();
 176         fiz1             = _mm_setzero_pd();
 177         fix2             = _mm_setzero_pd();
 178         fiy2             = _mm_setzero_pd();
 179         fiz2             = _mm_setzero_pd();
 180         fix3             = _mm_setzero_pd();
 181         fiy3             = _mm_setzero_pd();
 182         fiz3             = _mm_setzero_pd();
 183
 184         /* Reset potential sums */
 185         velecsum         = _mm_setzero_pd();
 186         vvdwsum          = _mm_setzero_pd();
 187
 188         /* Start inner kernel loop */
 189         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 190         {
 191
 192             /* Get j neighbor index, and coordinate index */
 193             jnrA             = jjnr[jidx];
 194             jnrB             = jjnr[jidx+1];
 195             j_coord_offsetA  = DIM*jnrA;
 196             j_coord_offsetB  = DIM*jnrB;
 197
 198             /* load j atom coordinates */
 199             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 200                                               &jx0,&jy0,&jz0);
 201
 202             /* Calculate displacement vector */
 203             dx00             = _mm_sub_pd(ix0,jx0);
 204             dy00             = _mm_sub_pd(iy0,jy0);
 205             dz00             = _mm_sub_pd(iz0,jz0);
 206             dx10             = _mm_sub_pd(ix1,jx0);
 207             dy10             = _mm_sub_pd(iy1,jy0);
 208             dz10             = _mm_sub_pd(iz1,jz0);
 209             dx20             = _mm_sub_pd(ix2,jx0);
 210             dy20             = _mm_sub_pd(iy2,jy0);
 211             dz20             = _mm_sub_pd(iz2,jz0);
 212             dx30             = _mm_sub_pd(ix3,jx0);
 213             dy30             = _mm_sub_pd(iy3,jy0);
 214             dz30             = _mm_sub_pd(iz3,jz0);
 215
 216             /* Calculate squared distance and things based on it */
 217             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 218             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 219             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 220             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 221
 222             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 223             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 224             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 225             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 226
 227             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 228             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 229             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 230
 231             /* Load parameters for j particles */
 232             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 233             vdwjidx0A        = 2*vdwtype[jnrA+0];
 234             vdwjidx0B        = 2*vdwtype[jnrB+0];
 235
 236             fjx0             = _mm_setzero_pd();
 237             fjy0             = _mm_setzero_pd();
 238             fjz0             = _mm_setzero_pd();
 239
 240             /**************************
 241              * CALCULATE INTERACTIONS *
 242              **************************/
 243
 244             r00              = _mm_mul_pd(rsq00,rinv00);
 245
 246             /* Compute parameters for interactions between i and j atoms */
 247             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 248                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 249
 250             /* Calculate table index by multiplying r with table scale and truncate to integer */
 251             rt               = _mm_mul_pd(r00,vftabscale);
 252             vfitab           = _mm_cvttpd_epi32(rt);
 253             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 254             vfitab           = _mm_slli_epi32(vfitab,3);
 255
 256             /* CUBIC SPLINE TABLE DISPERSION */
 257             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 258             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 259             GMX_MM_TRANSPOSE2_PD(Y,F);
 260             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 261             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 262             GMX_MM_TRANSPOSE2_PD(G,H);
 263             Heps             = _mm_mul_pd(vfeps,H);
 264             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 265             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 266             vvdw6            = _mm_mul_pd(c6_00,VV);
 267             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 268             fvdw6            = _mm_mul_pd(c6_00,FF);
 269
 270             /* CUBIC SPLINE TABLE REPULSION */
 271             vfitab           = _mm_add_epi32(vfitab,ifour);
 272             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 273             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 274             GMX_MM_TRANSPOSE2_PD(Y,F);
 275             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 276             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 277             GMX_MM_TRANSPOSE2_PD(G,H);
 278             Heps             = _mm_mul_pd(vfeps,H);
 279             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 280             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 281             vvdw12           = _mm_mul_pd(c12_00,VV);
 282             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 283             fvdw12           = _mm_mul_pd(c12_00,FF);
 284             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 285             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 286
 287             /* Update potential sum for this i atom from the interaction with this j atom. */
 288             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 289
 290             fscal            = fvdw;
 291
 292             /* Calculate temporary vectorial force */
 293             tx               = _mm_mul_pd(fscal,dx00);
 294             ty               = _mm_mul_pd(fscal,dy00);
 295             tz               = _mm_mul_pd(fscal,dz00);
 296
 297             /* Update vectorial force */
 298             fix0             = _mm_add_pd(fix0,tx);
 299             fiy0             = _mm_add_pd(fiy0,ty);
 300             fiz0             = _mm_add_pd(fiz0,tz);
 301
 302             fjx0             = _mm_add_pd(fjx0,tx);
 303             fjy0             = _mm_add_pd(fjy0,ty);
 304             fjz0             = _mm_add_pd(fjz0,tz);
 305
 306             /**************************
 307              * CALCULATE INTERACTIONS *
 308              **************************/
 309
 310             r10              = _mm_mul_pd(rsq10,rinv10);
 311
 312             /* Compute parameters for interactions between i and j atoms */
 313             qq10             = _mm_mul_pd(iq1,jq0);
 314
 315             /* EWALD ELECTROSTATICS */
 316
 317             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 318             ewrt             = _mm_mul_pd(r10,ewtabscale);
 319             ewitab           = _mm_cvttpd_epi32(ewrt);
 320             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 321             ewitab           = _mm_slli_epi32(ewitab,2);
 322             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 323             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 324             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 325             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 326             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 327             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 328             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 329             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 330             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 331             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velecsum         = _mm_add_pd(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             /* Calculate temporary vectorial force */
 339             tx               = _mm_mul_pd(fscal,dx10);
 340             ty               = _mm_mul_pd(fscal,dy10);
 341             tz               = _mm_mul_pd(fscal,dz10);
 342
 343             /* Update vectorial force */
 344             fix1             = _mm_add_pd(fix1,tx);
 345             fiy1             = _mm_add_pd(fiy1,ty);
 346             fiz1             = _mm_add_pd(fiz1,tz);
 347
 348             fjx0             = _mm_add_pd(fjx0,tx);
 349             fjy0             = _mm_add_pd(fjy0,ty);
 350             fjz0             = _mm_add_pd(fjz0,tz);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 354              **************************/
 355
 356             r20              = _mm_mul_pd(rsq20,rinv20);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq20             = _mm_mul_pd(iq2,jq0);
 360
 361             /* EWALD ELECTROSTATICS */
 362
 363             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 364             ewrt             = _mm_mul_pd(r20,ewtabscale);
 365             ewitab           = _mm_cvttpd_epi32(ewrt);
 366             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 367             ewitab           = _mm_slli_epi32(ewitab,2);
 368             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 369             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 370             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 371             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 372             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 373             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 374             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 375             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 376             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 377             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             velecsum         = _mm_add_pd(velecsum,velec);
 381
 382             fscal            = felec;
 383
 384             /* Calculate temporary vectorial force */
 385             tx               = _mm_mul_pd(fscal,dx20);
 386             ty               = _mm_mul_pd(fscal,dy20);
 387             tz               = _mm_mul_pd(fscal,dz20);
 388
 389             /* Update vectorial force */
 390             fix2             = _mm_add_pd(fix2,tx);
 391             fiy2             = _mm_add_pd(fiy2,ty);
 392             fiz2             = _mm_add_pd(fiz2,tz);
 393
 394             fjx0             = _mm_add_pd(fjx0,tx);
 395             fjy0             = _mm_add_pd(fjy0,ty);
 396             fjz0             = _mm_add_pd(fjz0,tz);
 397
 398             /**************************
 399              * CALCULATE INTERACTIONS *
 400              **************************/
 401
 402             r30              = _mm_mul_pd(rsq30,rinv30);
 403
 404             /* Compute parameters for interactions between i and j atoms */
 405             qq30             = _mm_mul_pd(iq3,jq0);
 406
 407             /* EWALD ELECTROSTATICS */
 408
 409             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 410             ewrt             = _mm_mul_pd(r30,ewtabscale);
 411             ewitab           = _mm_cvttpd_epi32(ewrt);
 412             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 413             ewitab           = _mm_slli_epi32(ewitab,2);
 414             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 415             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 416             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 417             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 418             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 419             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 420             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 421             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 422             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 423             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 424
 425             /* Update potential sum for this i atom from the interaction with this j atom. */
 426             velecsum         = _mm_add_pd(velecsum,velec);
 427
 428             fscal            = felec;
 429
 430             /* Calculate temporary vectorial force */
 431             tx               = _mm_mul_pd(fscal,dx30);
 432             ty               = _mm_mul_pd(fscal,dy30);
 433             tz               = _mm_mul_pd(fscal,dz30);
 434
 435             /* Update vectorial force */
 436             fix3             = _mm_add_pd(fix3,tx);
 437             fiy3             = _mm_add_pd(fiy3,ty);
 438             fiz3             = _mm_add_pd(fiz3,tz);
 439
 440             fjx0             = _mm_add_pd(fjx0,tx);
 441             fjy0             = _mm_add_pd(fjy0,ty);
 442             fjz0             = _mm_add_pd(fjz0,tz);
 443
 444             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 445
 446             /* Inner loop uses 182 flops */
 447         }
 448
 449         if(jidx<j_index_end)
 450         {
 451
 452             jnrA             = jjnr[jidx];
 453             j_coord_offsetA  = DIM*jnrA;
 454
 455             /* load j atom coordinates */
 456             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 457                                               &jx0,&jy0,&jz0);
 458
 459             /* Calculate displacement vector */
 460             dx00             = _mm_sub_pd(ix0,jx0);
 461             dy00             = _mm_sub_pd(iy0,jy0);
 462             dz00             = _mm_sub_pd(iz0,jz0);
 463             dx10             = _mm_sub_pd(ix1,jx0);
 464             dy10             = _mm_sub_pd(iy1,jy0);
 465             dz10             = _mm_sub_pd(iz1,jz0);
 466             dx20             = _mm_sub_pd(ix2,jx0);
 467             dy20             = _mm_sub_pd(iy2,jy0);
 468             dz20             = _mm_sub_pd(iz2,jz0);
 469             dx30             = _mm_sub_pd(ix3,jx0);
 470             dy30             = _mm_sub_pd(iy3,jy0);
 471             dz30             = _mm_sub_pd(iz3,jz0);
 472
 473             /* Calculate squared distance and things based on it */
 474             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 475             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 476             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 477             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 478
 479             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 480             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 481             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 482             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 483
 484             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 485             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 486             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 487
 488             /* Load parameters for j particles */
 489             jq0              = _mm_load_sd(charge+jnrA+0);
 490             vdwjidx0A        = 2*vdwtype[jnrA+0];
 491
 492             fjx0             = _mm_setzero_pd();
 493             fjy0             = _mm_setzero_pd();
 494             fjz0             = _mm_setzero_pd();
 495
 496             /**************************
 497              * CALCULATE INTERACTIONS *
 498              **************************/
 499
 500             r00              = _mm_mul_pd(rsq00,rinv00);
 501
 502             /* Compute parameters for interactions between i and j atoms */
 503             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 504
 505             /* Calculate table index by multiplying r with table scale and truncate to integer */
 506             rt               = _mm_mul_pd(r00,vftabscale);
 507             vfitab           = _mm_cvttpd_epi32(rt);
 508             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 509             vfitab           = _mm_slli_epi32(vfitab,3);
 510
 511             /* CUBIC SPLINE TABLE DISPERSION */
 512             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 513             F                = _mm_setzero_pd();
 514             GMX_MM_TRANSPOSE2_PD(Y,F);
 515             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 516             H                = _mm_setzero_pd();
 517             GMX_MM_TRANSPOSE2_PD(G,H);
 518             Heps             = _mm_mul_pd(vfeps,H);
 519             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 520             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 521             vvdw6            = _mm_mul_pd(c6_00,VV);
 522             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 523             fvdw6            = _mm_mul_pd(c6_00,FF);
 524
 525             /* CUBIC SPLINE TABLE REPULSION */
 526             vfitab           = _mm_add_epi32(vfitab,ifour);
 527             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 528             F                = _mm_setzero_pd();
 529             GMX_MM_TRANSPOSE2_PD(Y,F);
 530             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 531             H                = _mm_setzero_pd();
 532             GMX_MM_TRANSPOSE2_PD(G,H);
 533             Heps             = _mm_mul_pd(vfeps,H);
 534             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 535             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 536             vvdw12           = _mm_mul_pd(c12_00,VV);
 537             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 538             fvdw12           = _mm_mul_pd(c12_00,FF);
 539             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 540             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 541
 542             /* Update potential sum for this i atom from the interaction with this j atom. */
 543             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 544             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 545
 546             fscal            = fvdw;
 547
 548             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 549
 550             /* Calculate temporary vectorial force */
 551             tx               = _mm_mul_pd(fscal,dx00);
 552             ty               = _mm_mul_pd(fscal,dy00);
 553             tz               = _mm_mul_pd(fscal,dz00);
 554
 555             /* Update vectorial force */
 556             fix0             = _mm_add_pd(fix0,tx);
 557             fiy0             = _mm_add_pd(fiy0,ty);
 558             fiz0             = _mm_add_pd(fiz0,tz);
 559
 560             fjx0             = _mm_add_pd(fjx0,tx);
 561             fjy0             = _mm_add_pd(fjy0,ty);
 562             fjz0             = _mm_add_pd(fjz0,tz);
 563
 564             /**************************
 565              * CALCULATE INTERACTIONS *
 566              **************************/
 567
 568             r10              = _mm_mul_pd(rsq10,rinv10);
 569
 570             /* Compute parameters for interactions between i and j atoms */
 571             qq10             = _mm_mul_pd(iq1,jq0);
 572
 573             /* EWALD ELECTROSTATICS */
 574
 575             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 576             ewrt             = _mm_mul_pd(r10,ewtabscale);
 577             ewitab           = _mm_cvttpd_epi32(ewrt);
 578             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 579             ewitab           = _mm_slli_epi32(ewitab,2);
 580             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 581             ewtabD           = _mm_setzero_pd();
 582             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 583             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 584             ewtabFn          = _mm_setzero_pd();
 585             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 586             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 587             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 588             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 589             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 590
 591             /* Update potential sum for this i atom from the interaction with this j atom. */
 592             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 593             velecsum         = _mm_add_pd(velecsum,velec);
 594
 595             fscal            = felec;
 596
 597             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 598
 599             /* Calculate temporary vectorial force */
 600             tx               = _mm_mul_pd(fscal,dx10);
 601             ty               = _mm_mul_pd(fscal,dy10);
 602             tz               = _mm_mul_pd(fscal,dz10);
 603
 604             /* Update vectorial force */
 605             fix1             = _mm_add_pd(fix1,tx);
 606             fiy1             = _mm_add_pd(fiy1,ty);
 607             fiz1             = _mm_add_pd(fiz1,tz);
 608
 609             fjx0             = _mm_add_pd(fjx0,tx);
 610             fjy0             = _mm_add_pd(fjy0,ty);
 611             fjz0             = _mm_add_pd(fjz0,tz);
 612
 613             /**************************
 614              * CALCULATE INTERACTIONS *
 615              **************************/
 616
 617             r20              = _mm_mul_pd(rsq20,rinv20);
 618
 619             /* Compute parameters for interactions between i and j atoms */
 620             qq20             = _mm_mul_pd(iq2,jq0);
 621
 622             /* EWALD ELECTROSTATICS */
 623
 624             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 625             ewrt             = _mm_mul_pd(r20,ewtabscale);
 626             ewitab           = _mm_cvttpd_epi32(ewrt);
 627             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 628             ewitab           = _mm_slli_epi32(ewitab,2);
 629             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 630             ewtabD           = _mm_setzero_pd();
 631             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 632             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 633             ewtabFn          = _mm_setzero_pd();
 634             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 635             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 636             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 637             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 638             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 639
 640             /* Update potential sum for this i atom from the interaction with this j atom. */
 641             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 642             velecsum         = _mm_add_pd(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 647
 648             /* Calculate temporary vectorial force */
 649             tx               = _mm_mul_pd(fscal,dx20);
 650             ty               = _mm_mul_pd(fscal,dy20);
 651             tz               = _mm_mul_pd(fscal,dz20);
 652
 653             /* Update vectorial force */
 654             fix2             = _mm_add_pd(fix2,tx);
 655             fiy2             = _mm_add_pd(fiy2,ty);
 656             fiz2             = _mm_add_pd(fiz2,tz);
 657
 658             fjx0             = _mm_add_pd(fjx0,tx);
 659             fjy0             = _mm_add_pd(fjy0,ty);
 660             fjz0             = _mm_add_pd(fjz0,tz);
 661
 662             /**************************
 663              * CALCULATE INTERACTIONS *
 664              **************************/
 665
 666             r30              = _mm_mul_pd(rsq30,rinv30);
 667
 668             /* Compute parameters for interactions between i and j atoms */
 669             qq30             = _mm_mul_pd(iq3,jq0);
 670
 671             /* EWALD ELECTROSTATICS */
 672
 673             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 674             ewrt             = _mm_mul_pd(r30,ewtabscale);
 675             ewitab           = _mm_cvttpd_epi32(ewrt);
 676             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 677             ewitab           = _mm_slli_epi32(ewitab,2);
 678             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 679             ewtabD           = _mm_setzero_pd();
 680             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 681             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 682             ewtabFn          = _mm_setzero_pd();
 683             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 684             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 685             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 686             velec            = _mm_mul_pd(qq30,_mm_sub_pd(rinv30,velec));
 687             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
 688
 689             /* Update potential sum for this i atom from the interaction with this j atom. */
 690             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 691             velecsum         = _mm_add_pd(velecsum,velec);
 692
 693             fscal            = felec;
 694
 695             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 696
 697             /* Calculate temporary vectorial force */
 698             tx               = _mm_mul_pd(fscal,dx30);
 699             ty               = _mm_mul_pd(fscal,dy30);
 700             tz               = _mm_mul_pd(fscal,dz30);
 701
 702             /* Update vectorial force */
 703             fix3             = _mm_add_pd(fix3,tx);
 704             fiy3             = _mm_add_pd(fiy3,ty);
 705             fiz3             = _mm_add_pd(fiz3,tz);
 706
 707             fjx0             = _mm_add_pd(fjx0,tx);
 708             fjy0             = _mm_add_pd(fjy0,ty);
 709             fjz0             = _mm_add_pd(fjz0,tz);
 710
 711             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 712
 713             /* Inner loop uses 182 flops */
 714         }
 715
 716         /* End of innermost loop */
 717
 718         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 719                                               f+i_coord_offset,fshift+i_shift_offset);
 720
 721         ggid                        = gid[iidx];
 722         /* Update potential energies */
 723         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 724         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 725
 726         /* Increment number of inner iterations */
 727         inneriter                  += j_index_end - j_index_start;
 728
 729         /* Outer loop uses 26 flops */
 730     }
 731
 732     /* Increment number of outer iterations */
 733     outeriter        += nri;
 734
 735     /* Update outer/inner flops */
 736
 737     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*182);
 738 }
 739 /*
 740  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4P1_F_sse2_double
 741  * Electrostatics interaction: Ewald
 742  * VdW interaction:            CubicSplineTable
 743  * Geometry:                   Water4-Particle
 744  * Calculate force/pot:        Force
 745  */
 746 void
 747 nb_kernel_ElecEw_VdwCSTab_GeomW4P1_F_sse2_double
 748                     (t_nblist                    * gmx_restrict       nlist,
 749                      rvec                        * gmx_restrict          xx,
 750                      rvec                        * gmx_restrict          ff,
 751                      t_forcerec                  * gmx_restrict          fr,
 752                      t_mdatoms                   * gmx_restrict     mdatoms,
 753                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 754                      t_nrnb                      * gmx_restrict        nrnb)
 755 {
 756     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 757      * just 0 for non-waters.
 758      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 759      * jnr indices corresponding to data put in the four positions in the SIMD register.
 760      */
 761     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 762     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 763     int              jnrA,jnrB;
 764     int              j_coord_offsetA,j_coord_offsetB;
 765     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 766     real             rcutoff_scalar;
 767     real             *shiftvec,*fshift,*x,*f;
 768     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 769     int              vdwioffset0;
 770     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 771     int              vdwioffset1;
 772     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 773     int              vdwioffset2;
 774     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 775     int              vdwioffset3;
 776     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 777     int              vdwjidx0A,vdwjidx0B;
 778     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 779     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 780     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 781     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 782     __m128d          dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 783     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 784     real             *charge;
 785     int              nvdwtype;
 786     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 787     int              *vdwtype;
 788     real             *vdwparam;
 789     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 790     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 791     __m128i          vfitab;
 792     __m128i          ifour       = _mm_set1_epi32(4);
 793     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 794     real             *vftab;
 795     __m128i          ewitab;
 796     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 797     real             *ewtab;
 798     __m128d          dummy_mask,cutoff_mask;
 799     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 800     __m128d          one     = _mm_set1_pd(1.0);
 801     __m128d          two     = _mm_set1_pd(2.0);
 802     x                = xx[0];
 803     f                = ff[0];
 804
 805     nri              = nlist->nri;
 806     iinr             = nlist->iinr;
 807     jindex           = nlist->jindex;
 808     jjnr             = nlist->jjnr;
 809     shiftidx         = nlist->shift;
 810     gid              = nlist->gid;
 811     shiftvec         = fr->shift_vec[0];
 812     fshift           = fr->fshift[0];
 813     facel            = _mm_set1_pd(fr->epsfac);
 814     charge           = mdatoms->chargeA;
 815     nvdwtype         = fr->ntype;
 816     vdwparam         = fr->nbfp;
 817     vdwtype          = mdatoms->typeA;
 818
 819     vftab            = kernel_data->table_vdw->data;
 820     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 821
 822     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 823     ewtab            = fr->ic->tabq_coul_F;
 824     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 825     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 826
 827     /* Setup water-specific parameters */
 828     inr              = nlist->iinr[0];
 829     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 830     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 831     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 832     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 833
 834     /* Avoid stupid compiler warnings */
 835     jnrA = jnrB = 0;
 836     j_coord_offsetA = 0;
 837     j_coord_offsetB = 0;
 838
 839     outeriter        = 0;
 840     inneriter        = 0;
 841
 842     /* Start outer loop over neighborlists */
 843     for(iidx=0; iidx<nri; iidx++)
 844     {
 845         /* Load shift vector for this list */
 846         i_shift_offset   = DIM*shiftidx[iidx];
 847
 848         /* Load limits for loop over neighbors */
 849         j_index_start    = jindex[iidx];
 850         j_index_end      = jindex[iidx+1];
 851
 852         /* Get outer coordinate index */
 853         inr              = iinr[iidx];
 854         i_coord_offset   = DIM*inr;
 855
 856         /* Load i particle coords and add shift vector */
 857         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 858                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 859
 860         fix0             = _mm_setzero_pd();
 861         fiy0             = _mm_setzero_pd();
 862         fiz0             = _mm_setzero_pd();
 863         fix1             = _mm_setzero_pd();
 864         fiy1             = _mm_setzero_pd();
 865         fiz1             = _mm_setzero_pd();
 866         fix2             = _mm_setzero_pd();
 867         fiy2             = _mm_setzero_pd();
 868         fiz2             = _mm_setzero_pd();
 869         fix3             = _mm_setzero_pd();
 870         fiy3             = _mm_setzero_pd();
 871         fiz3             = _mm_setzero_pd();
 872
 873         /* Start inner kernel loop */
 874         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 875         {
 876
 877             /* Get j neighbor index, and coordinate index */
 878             jnrA             = jjnr[jidx];
 879             jnrB             = jjnr[jidx+1];
 880             j_coord_offsetA  = DIM*jnrA;
 881             j_coord_offsetB  = DIM*jnrB;
 882
 883             /* load j atom coordinates */
 884             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 885                                               &jx0,&jy0,&jz0);
 886
 887             /* Calculate displacement vector */
 888             dx00             = _mm_sub_pd(ix0,jx0);
 889             dy00             = _mm_sub_pd(iy0,jy0);
 890             dz00             = _mm_sub_pd(iz0,jz0);
 891             dx10             = _mm_sub_pd(ix1,jx0);
 892             dy10             = _mm_sub_pd(iy1,jy0);
 893             dz10             = _mm_sub_pd(iz1,jz0);
 894             dx20             = _mm_sub_pd(ix2,jx0);
 895             dy20             = _mm_sub_pd(iy2,jy0);
 896             dz20             = _mm_sub_pd(iz2,jz0);
 897             dx30             = _mm_sub_pd(ix3,jx0);
 898             dy30             = _mm_sub_pd(iy3,jy0);
 899             dz30             = _mm_sub_pd(iz3,jz0);
 900
 901             /* Calculate squared distance and things based on it */
 902             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 903             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 904             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 905             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
 906
 907             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 908             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 909             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 910             rinv30           = gmx_mm_invsqrt_pd(rsq30);
 911
 912             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 913             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 914             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
 915
 916             /* Load parameters for j particles */
 917             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 918             vdwjidx0A        = 2*vdwtype[jnrA+0];
 919             vdwjidx0B        = 2*vdwtype[jnrB+0];
 920
 921             fjx0             = _mm_setzero_pd();
 922             fjy0             = _mm_setzero_pd();
 923             fjz0             = _mm_setzero_pd();
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             r00              = _mm_mul_pd(rsq00,rinv00);
 930
 931             /* Compute parameters for interactions between i and j atoms */
 932             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 933                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 934
 935             /* Calculate table index by multiplying r with table scale and truncate to integer */
 936             rt               = _mm_mul_pd(r00,vftabscale);
 937             vfitab           = _mm_cvttpd_epi32(rt);
 938             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 939             vfitab           = _mm_slli_epi32(vfitab,3);
 940
 941             /* CUBIC SPLINE TABLE DISPERSION */
 942             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 943             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 944             GMX_MM_TRANSPOSE2_PD(Y,F);
 945             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 946             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 947             GMX_MM_TRANSPOSE2_PD(G,H);
 948             Heps             = _mm_mul_pd(vfeps,H);
 949             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 950             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 951             fvdw6            = _mm_mul_pd(c6_00,FF);
 952
 953             /* CUBIC SPLINE TABLE REPULSION */
 954             vfitab           = _mm_add_epi32(vfitab,ifour);
 955             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 956             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 957             GMX_MM_TRANSPOSE2_PD(Y,F);
 958             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 959             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 960             GMX_MM_TRANSPOSE2_PD(G,H);
 961             Heps             = _mm_mul_pd(vfeps,H);
 962             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 963             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 964             fvdw12           = _mm_mul_pd(c12_00,FF);
 965             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 966
 967             fscal            = fvdw;
 968
 969             /* Calculate temporary vectorial force */
 970             tx               = _mm_mul_pd(fscal,dx00);
 971             ty               = _mm_mul_pd(fscal,dy00);
 972             tz               = _mm_mul_pd(fscal,dz00);
 973
 974             /* Update vectorial force */
 975             fix0             = _mm_add_pd(fix0,tx);
 976             fiy0             = _mm_add_pd(fiy0,ty);
 977             fiz0             = _mm_add_pd(fiz0,tz);
 978
 979             fjx0             = _mm_add_pd(fjx0,tx);
 980             fjy0             = _mm_add_pd(fjy0,ty);
 981             fjz0             = _mm_add_pd(fjz0,tz);
 982
 983             /**************************
 984              * CALCULATE INTERACTIONS *
 985              **************************/
 986
 987             r10              = _mm_mul_pd(rsq10,rinv10);
 988
 989             /* Compute parameters for interactions between i and j atoms */
 990             qq10             = _mm_mul_pd(iq1,jq0);
 991
 992             /* EWALD ELECTROSTATICS */
 993
 994             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 995             ewrt             = _mm_mul_pd(r10,ewtabscale);
 996             ewitab           = _mm_cvttpd_epi32(ewrt);
 997             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 998             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
 999                                          &ewtabF,&ewtabFn);
1000             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1001             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1002
1003             fscal            = felec;
1004
1005             /* Calculate temporary vectorial force */
1006             tx               = _mm_mul_pd(fscal,dx10);
1007             ty               = _mm_mul_pd(fscal,dy10);
1008             tz               = _mm_mul_pd(fscal,dz10);
1009
1010             /* Update vectorial force */
1011             fix1             = _mm_add_pd(fix1,tx);
1012             fiy1             = _mm_add_pd(fiy1,ty);
1013             fiz1             = _mm_add_pd(fiz1,tz);
1014
1015             fjx0             = _mm_add_pd(fjx0,tx);
1016             fjy0             = _mm_add_pd(fjy0,ty);
1017             fjz0             = _mm_add_pd(fjz0,tz);
1018
1019             /**************************
1020              * CALCULATE INTERACTIONS *
1021              **************************/
1022
1023             r20              = _mm_mul_pd(rsq20,rinv20);
1024
1025             /* Compute parameters for interactions between i and j atoms */
1026             qq20             = _mm_mul_pd(iq2,jq0);
1027
1028             /* EWALD ELECTROSTATICS */
1029
1030             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1031             ewrt             = _mm_mul_pd(r20,ewtabscale);
1032             ewitab           = _mm_cvttpd_epi32(ewrt);
1033             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1034             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1035                                          &ewtabF,&ewtabFn);
1036             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1037             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1038
1039             fscal            = felec;
1040
1041             /* Calculate temporary vectorial force */
1042             tx               = _mm_mul_pd(fscal,dx20);
1043             ty               = _mm_mul_pd(fscal,dy20);
1044             tz               = _mm_mul_pd(fscal,dz20);
1045
1046             /* Update vectorial force */
1047             fix2             = _mm_add_pd(fix2,tx);
1048             fiy2             = _mm_add_pd(fiy2,ty);
1049             fiz2             = _mm_add_pd(fiz2,tz);
1050
1051             fjx0             = _mm_add_pd(fjx0,tx);
1052             fjy0             = _mm_add_pd(fjy0,ty);
1053             fjz0             = _mm_add_pd(fjz0,tz);
1054
1055             /**************************
1056              * CALCULATE INTERACTIONS *
1057              **************************/
1058
1059             r30              = _mm_mul_pd(rsq30,rinv30);
1060
1061             /* Compute parameters for interactions between i and j atoms */
1062             qq30             = _mm_mul_pd(iq3,jq0);
1063
1064             /* EWALD ELECTROSTATICS */
1065
1066             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1067             ewrt             = _mm_mul_pd(r30,ewtabscale);
1068             ewitab           = _mm_cvttpd_epi32(ewrt);
1069             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1070             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1071                                          &ewtabF,&ewtabFn);
1072             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1073             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1074
1075             fscal            = felec;
1076
1077             /* Calculate temporary vectorial force */
1078             tx               = _mm_mul_pd(fscal,dx30);
1079             ty               = _mm_mul_pd(fscal,dy30);
1080             tz               = _mm_mul_pd(fscal,dz30);
1081
1082             /* Update vectorial force */
1083             fix3             = _mm_add_pd(fix3,tx);
1084             fiy3             = _mm_add_pd(fiy3,ty);
1085             fiz3             = _mm_add_pd(fiz3,tz);
1086
1087             fjx0             = _mm_add_pd(fjx0,tx);
1088             fjy0             = _mm_add_pd(fjy0,ty);
1089             fjz0             = _mm_add_pd(fjz0,tz);
1090
1091             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1092
1093             /* Inner loop uses 159 flops */
1094         }
1095
1096         if(jidx<j_index_end)
1097         {
1098
1099             jnrA             = jjnr[jidx];
1100             j_coord_offsetA  = DIM*jnrA;
1101
1102             /* load j atom coordinates */
1103             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1104                                               &jx0,&jy0,&jz0);
1105
1106             /* Calculate displacement vector */
1107             dx00             = _mm_sub_pd(ix0,jx0);
1108             dy00             = _mm_sub_pd(iy0,jy0);
1109             dz00             = _mm_sub_pd(iz0,jz0);
1110             dx10             = _mm_sub_pd(ix1,jx0);
1111             dy10             = _mm_sub_pd(iy1,jy0);
1112             dz10             = _mm_sub_pd(iz1,jz0);
1113             dx20             = _mm_sub_pd(ix2,jx0);
1114             dy20             = _mm_sub_pd(iy2,jy0);
1115             dz20             = _mm_sub_pd(iz2,jz0);
1116             dx30             = _mm_sub_pd(ix3,jx0);
1117             dy30             = _mm_sub_pd(iy3,jy0);
1118             dz30             = _mm_sub_pd(iz3,jz0);
1119
1120             /* Calculate squared distance and things based on it */
1121             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1122             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1123             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1124             rsq30            = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
1125
1126             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1127             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1128             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1129             rinv30           = gmx_mm_invsqrt_pd(rsq30);
1130
1131             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1132             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1133             rinvsq30         = _mm_mul_pd(rinv30,rinv30);
1134
1135             /* Load parameters for j particles */
1136             jq0              = _mm_load_sd(charge+jnrA+0);
1137             vdwjidx0A        = 2*vdwtype[jnrA+0];
1138
1139             fjx0             = _mm_setzero_pd();
1140             fjy0             = _mm_setzero_pd();
1141             fjz0             = _mm_setzero_pd();
1142
1143             /**************************
1144              * CALCULATE INTERACTIONS *
1145              **************************/
1146
1147             r00              = _mm_mul_pd(rsq00,rinv00);
1148
1149             /* Compute parameters for interactions between i and j atoms */
1150             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1151
1152             /* Calculate table index by multiplying r with table scale and truncate to integer */
1153             rt               = _mm_mul_pd(r00,vftabscale);
1154             vfitab           = _mm_cvttpd_epi32(rt);
1155             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1156             vfitab           = _mm_slli_epi32(vfitab,3);
1157
1158             /* CUBIC SPLINE TABLE DISPERSION */
1159             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1160             F                = _mm_setzero_pd();
1161             GMX_MM_TRANSPOSE2_PD(Y,F);
1162             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1163             H                = _mm_setzero_pd();
1164             GMX_MM_TRANSPOSE2_PD(G,H);
1165             Heps             = _mm_mul_pd(vfeps,H);
1166             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1167             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1168             fvdw6            = _mm_mul_pd(c6_00,FF);
1169
1170             /* CUBIC SPLINE TABLE REPULSION */
1171             vfitab           = _mm_add_epi32(vfitab,ifour);
1172             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1173             F                = _mm_setzero_pd();
1174             GMX_MM_TRANSPOSE2_PD(Y,F);
1175             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1176             H                = _mm_setzero_pd();
1177             GMX_MM_TRANSPOSE2_PD(G,H);
1178             Heps             = _mm_mul_pd(vfeps,H);
1179             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1180             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1181             fvdw12           = _mm_mul_pd(c12_00,FF);
1182             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1183
1184             fscal            = fvdw;
1185
1186             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1187
1188             /* Calculate temporary vectorial force */
1189             tx               = _mm_mul_pd(fscal,dx00);
1190             ty               = _mm_mul_pd(fscal,dy00);
1191             tz               = _mm_mul_pd(fscal,dz00);
1192
1193             /* Update vectorial force */
1194             fix0             = _mm_add_pd(fix0,tx);
1195             fiy0             = _mm_add_pd(fiy0,ty);
1196             fiz0             = _mm_add_pd(fiz0,tz);
1197
1198             fjx0             = _mm_add_pd(fjx0,tx);
1199             fjy0             = _mm_add_pd(fjy0,ty);
1200             fjz0             = _mm_add_pd(fjz0,tz);
1201
1202             /**************************
1203              * CALCULATE INTERACTIONS *
1204              **************************/
1205
1206             r10              = _mm_mul_pd(rsq10,rinv10);
1207
1208             /* Compute parameters for interactions between i and j atoms */
1209             qq10             = _mm_mul_pd(iq1,jq0);
1210
1211             /* EWALD ELECTROSTATICS */
1212
1213             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1214             ewrt             = _mm_mul_pd(r10,ewtabscale);
1215             ewitab           = _mm_cvttpd_epi32(ewrt);
1216             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1217             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1218             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1219             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1220
1221             fscal            = felec;
1222
1223             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1224
1225             /* Calculate temporary vectorial force */
1226             tx               = _mm_mul_pd(fscal,dx10);
1227             ty               = _mm_mul_pd(fscal,dy10);
1228             tz               = _mm_mul_pd(fscal,dz10);
1229
1230             /* Update vectorial force */
1231             fix1             = _mm_add_pd(fix1,tx);
1232             fiy1             = _mm_add_pd(fiy1,ty);
1233             fiz1             = _mm_add_pd(fiz1,tz);
1234
1235             fjx0             = _mm_add_pd(fjx0,tx);
1236             fjy0             = _mm_add_pd(fjy0,ty);
1237             fjz0             = _mm_add_pd(fjz0,tz);
1238
1239             /**************************
1240              * CALCULATE INTERACTIONS *
1241              **************************/
1242
1243             r20              = _mm_mul_pd(rsq20,rinv20);
1244
1245             /* Compute parameters for interactions between i and j atoms */
1246             qq20             = _mm_mul_pd(iq2,jq0);
1247
1248             /* EWALD ELECTROSTATICS */
1249
1250             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1251             ewrt             = _mm_mul_pd(r20,ewtabscale);
1252             ewitab           = _mm_cvttpd_epi32(ewrt);
1253             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1254             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1255             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1256             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1257
1258             fscal            = felec;
1259
1260             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1261
1262             /* Calculate temporary vectorial force */
1263             tx               = _mm_mul_pd(fscal,dx20);
1264             ty               = _mm_mul_pd(fscal,dy20);
1265             tz               = _mm_mul_pd(fscal,dz20);
1266
1267             /* Update vectorial force */
1268             fix2             = _mm_add_pd(fix2,tx);
1269             fiy2             = _mm_add_pd(fiy2,ty);
1270             fiz2             = _mm_add_pd(fiz2,tz);
1271
1272             fjx0             = _mm_add_pd(fjx0,tx);
1273             fjy0             = _mm_add_pd(fjy0,ty);
1274             fjz0             = _mm_add_pd(fjz0,tz);
1275
1276             /**************************
1277              * CALCULATE INTERACTIONS *
1278              **************************/
1279
1280             r30              = _mm_mul_pd(rsq30,rinv30);
1281
1282             /* Compute parameters for interactions between i and j atoms */
1283             qq30             = _mm_mul_pd(iq3,jq0);
1284
1285             /* EWALD ELECTROSTATICS */
1286
1287             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1288             ewrt             = _mm_mul_pd(r30,ewtabscale);
1289             ewitab           = _mm_cvttpd_epi32(ewrt);
1290             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1291             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1292             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1293             felec            = _mm_mul_pd(_mm_mul_pd(qq30,rinv30),_mm_sub_pd(rinvsq30,felec));
1294
1295             fscal            = felec;
1296
1297             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1298
1299             /* Calculate temporary vectorial force */
1300             tx               = _mm_mul_pd(fscal,dx30);
1301             ty               = _mm_mul_pd(fscal,dy30);
1302             tz               = _mm_mul_pd(fscal,dz30);
1303
1304             /* Update vectorial force */
1305             fix3             = _mm_add_pd(fix3,tx);
1306             fiy3             = _mm_add_pd(fiy3,ty);
1307             fiz3             = _mm_add_pd(fiz3,tz);
1308
1309             fjx0             = _mm_add_pd(fjx0,tx);
1310             fjy0             = _mm_add_pd(fjy0,ty);
1311             fjz0             = _mm_add_pd(fjz0,tz);
1312
1313             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1314
1315             /* Inner loop uses 159 flops */
1316         }
1317
1318         /* End of innermost loop */
1319
1320         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1321                                               f+i_coord_offset,fshift+i_shift_offset);
1322
1323         /* Increment number of inner iterations */
1324         inneriter                  += j_index_end - j_index_start;
1325
1326         /* Outer loop uses 24 flops */
1327     }
1328
1329     /* Increment number of outer iterations */
1330     outeriter        += nri;
1331
1332     /* Update outer/inner flops */
1333
1334     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);
1335 }