src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEw_VdwCSTab_GeomW3W3_avx_256_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 avx_256_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_avx_256_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_avx_256_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_avx_256_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,C,D refer to j loop unrolling done with AVX, e.g. for the four 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,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 100     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 101     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 102     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 107     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 108     real             *charge;
 109     int              nvdwtype;
 110     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 111     int              *vdwtype;
 112     real             *vdwparam;
 113     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 114     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 115     __m128i          vfitab;
 116     __m128i          ifour       = _mm_set1_epi32(4);
 117     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 118     real             *vftab;
 119     __m128i          ewitab;
 120     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 121     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 122     real             *ewtab;
 123     __m256d          dummy_mask,cutoff_mask;
 124     __m128           tmpmask0,tmpmask1;
 125     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 126     __m256d          one     = _mm256_set1_pd(1.0);
 127     __m256d          two     = _mm256_set1_pd(2.0);
 128     x                = xx[0];
 129     f                = ff[0];
 130
 131     nri              = nlist->nri;
 132     iinr             = nlist->iinr;
 133     jindex           = nlist->jindex;
 134     jjnr             = nlist->jjnr;
 135     shiftidx         = nlist->shift;
 136     gid              = nlist->gid;
 137     shiftvec         = fr->shift_vec[0];
 138     fshift           = fr->fshift[0];
 139     facel            = _mm256_set1_pd(fr->epsfac);
 140     charge           = mdatoms->chargeA;
 141     nvdwtype         = fr->ntype;
 142     vdwparam         = fr->nbfp;
 143     vdwtype          = mdatoms->typeA;
 144
 145     vftab            = kernel_data->table_vdw->data;
 146     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 147
 148     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 149     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 150     beta2            = _mm256_mul_pd(beta,beta);
 151     beta3            = _mm256_mul_pd(beta,beta2);
 152
 153     ewtab            = fr->ic->tabq_coul_FDV0;
 154     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 155     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 156
 157     /* Setup water-specific parameters */
 158     inr              = nlist->iinr[0];
 159     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 160     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 161     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 162     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 163
 164     jq0              = _mm256_set1_pd(charge[inr+0]);
 165     jq1              = _mm256_set1_pd(charge[inr+1]);
 166     jq2              = _mm256_set1_pd(charge[inr+2]);
 167     vdwjidx0A        = 2*vdwtype[inr+0];
 168     qq00             = _mm256_mul_pd(iq0,jq0);
 169     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 170     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 171     qq01             = _mm256_mul_pd(iq0,jq1);
 172     qq02             = _mm256_mul_pd(iq0,jq2);
 173     qq10             = _mm256_mul_pd(iq1,jq0);
 174     qq11             = _mm256_mul_pd(iq1,jq1);
 175     qq12             = _mm256_mul_pd(iq1,jq2);
 176     qq20             = _mm256_mul_pd(iq2,jq0);
 177     qq21             = _mm256_mul_pd(iq2,jq1);
 178     qq22             = _mm256_mul_pd(iq2,jq2);
 179
 180     /* Avoid stupid compiler warnings */
 181     jnrA = jnrB = jnrC = jnrD = 0;
 182     j_coord_offsetA = 0;
 183     j_coord_offsetB = 0;
 184     j_coord_offsetC = 0;
 185     j_coord_offsetD = 0;
 186
 187     outeriter        = 0;
 188     inneriter        = 0;
 189
 190     for(iidx=0;iidx<4*DIM;iidx++)
 191     {
 192         scratch[iidx] = 0.0;
 193     }
 194
 195     /* Start outer loop over neighborlists */
 196     for(iidx=0; iidx<nri; iidx++)
 197     {
 198         /* Load shift vector for this list */
 199         i_shift_offset   = DIM*shiftidx[iidx];
 200
 201         /* Load limits for loop over neighbors */
 202         j_index_start    = jindex[iidx];
 203         j_index_end      = jindex[iidx+1];
 204
 205         /* Get outer coordinate index */
 206         inr              = iinr[iidx];
 207         i_coord_offset   = DIM*inr;
 208
 209         /* Load i particle coords and add shift vector */
 210         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 211                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 212
 213         fix0             = _mm256_setzero_pd();
 214         fiy0             = _mm256_setzero_pd();
 215         fiz0             = _mm256_setzero_pd();
 216         fix1             = _mm256_setzero_pd();
 217         fiy1             = _mm256_setzero_pd();
 218         fiz1             = _mm256_setzero_pd();
 219         fix2             = _mm256_setzero_pd();
 220         fiy2             = _mm256_setzero_pd();
 221         fiz2             = _mm256_setzero_pd();
 222
 223         /* Reset potential sums */
 224         velecsum         = _mm256_setzero_pd();
 225         vvdwsum          = _mm256_setzero_pd();
 226
 227         /* Start inner kernel loop */
 228         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 229         {
 230
 231             /* Get j neighbor index, and coordinate index */
 232             jnrA             = jjnr[jidx];
 233             jnrB             = jjnr[jidx+1];
 234             jnrC             = jjnr[jidx+2];
 235             jnrD             = jjnr[jidx+3];
 236             j_coord_offsetA  = DIM*jnrA;
 237             j_coord_offsetB  = DIM*jnrB;
 238             j_coord_offsetC  = DIM*jnrC;
 239             j_coord_offsetD  = DIM*jnrD;
 240
 241             /* load j atom coordinates */
 242             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 243                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 244                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 245
 246             /* Calculate displacement vector */
 247             dx00             = _mm256_sub_pd(ix0,jx0);
 248             dy00             = _mm256_sub_pd(iy0,jy0);
 249             dz00             = _mm256_sub_pd(iz0,jz0);
 250             dx01             = _mm256_sub_pd(ix0,jx1);
 251             dy01             = _mm256_sub_pd(iy0,jy1);
 252             dz01             = _mm256_sub_pd(iz0,jz1);
 253             dx02             = _mm256_sub_pd(ix0,jx2);
 254             dy02             = _mm256_sub_pd(iy0,jy2);
 255             dz02             = _mm256_sub_pd(iz0,jz2);
 256             dx10             = _mm256_sub_pd(ix1,jx0);
 257             dy10             = _mm256_sub_pd(iy1,jy0);
 258             dz10             = _mm256_sub_pd(iz1,jz0);
 259             dx11             = _mm256_sub_pd(ix1,jx1);
 260             dy11             = _mm256_sub_pd(iy1,jy1);
 261             dz11             = _mm256_sub_pd(iz1,jz1);
 262             dx12             = _mm256_sub_pd(ix1,jx2);
 263             dy12             = _mm256_sub_pd(iy1,jy2);
 264             dz12             = _mm256_sub_pd(iz1,jz2);
 265             dx20             = _mm256_sub_pd(ix2,jx0);
 266             dy20             = _mm256_sub_pd(iy2,jy0);
 267             dz20             = _mm256_sub_pd(iz2,jz0);
 268             dx21             = _mm256_sub_pd(ix2,jx1);
 269             dy21             = _mm256_sub_pd(iy2,jy1);
 270             dz21             = _mm256_sub_pd(iz2,jz1);
 271             dx22             = _mm256_sub_pd(ix2,jx2);
 272             dy22             = _mm256_sub_pd(iy2,jy2);
 273             dz22             = _mm256_sub_pd(iz2,jz2);
 274
 275             /* Calculate squared distance and things based on it */
 276             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 277             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 278             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 279             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 280             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 281             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 282             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 283             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 284             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 285
 286             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 287             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 288             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 289             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 290             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 291             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 292             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 293             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 294             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 295
 296             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 297             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 298             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 299             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 300             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 301             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 302             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 303             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 304             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 305
 306             fjx0             = _mm256_setzero_pd();
 307             fjy0             = _mm256_setzero_pd();
 308             fjz0             = _mm256_setzero_pd();
 309             fjx1             = _mm256_setzero_pd();
 310             fjy1             = _mm256_setzero_pd();
 311             fjz1             = _mm256_setzero_pd();
 312             fjx2             = _mm256_setzero_pd();
 313             fjy2             = _mm256_setzero_pd();
 314             fjz2             = _mm256_setzero_pd();
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             r00              = _mm256_mul_pd(rsq00,rinv00);
 321
 322             /* Calculate table index by multiplying r with table scale and truncate to integer */
 323             rt               = _mm256_mul_pd(r00,vftabscale);
 324             vfitab           = _mm256_cvttpd_epi32(rt);
 325             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 326             vfitab           = _mm_slli_epi32(vfitab,3);
 327
 328             /* EWALD ELECTROSTATICS */
 329
 330             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 331             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 332             ewitab           = _mm256_cvttpd_epi32(ewrt);
 333             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 334             ewitab           = _mm_slli_epi32(ewitab,2);
 335             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 336             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 337             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 338             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 339             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 340             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 341             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 342             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 343             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 344
 345             /* CUBIC SPLINE TABLE DISPERSION */
 346             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 347             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 348             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 349             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 350             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 351             Heps             = _mm256_mul_pd(vfeps,H);
 352             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 353             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 354             vvdw6            = _mm256_mul_pd(c6_00,VV);
 355             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 356             fvdw6            = _mm256_mul_pd(c6_00,FF);
 357
 358             /* CUBIC SPLINE TABLE REPULSION */
 359             vfitab           = _mm_add_epi32(vfitab,ifour);
 360             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 361             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 362             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 363             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 364             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 365             Heps             = _mm256_mul_pd(vfeps,H);
 366             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 367             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 368             vvdw12           = _mm256_mul_pd(c12_00,VV);
 369             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 370             fvdw12           = _mm256_mul_pd(c12_00,FF);
 371             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 372             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 373
 374             /* Update potential sum for this i atom from the interaction with this j atom. */
 375             velecsum         = _mm256_add_pd(velecsum,velec);
 376             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 377
 378             fscal            = _mm256_add_pd(felec,fvdw);
 379
 380             /* Calculate temporary vectorial force */
 381             tx               = _mm256_mul_pd(fscal,dx00);
 382             ty               = _mm256_mul_pd(fscal,dy00);
 383             tz               = _mm256_mul_pd(fscal,dz00);
 384
 385             /* Update vectorial force */
 386             fix0             = _mm256_add_pd(fix0,tx);
 387             fiy0             = _mm256_add_pd(fiy0,ty);
 388             fiz0             = _mm256_add_pd(fiz0,tz);
 389
 390             fjx0             = _mm256_add_pd(fjx0,tx);
 391             fjy0             = _mm256_add_pd(fjy0,ty);
 392             fjz0             = _mm256_add_pd(fjz0,tz);
 393
 394             /**************************
 395              * CALCULATE INTERACTIONS *
 396              **************************/
 397
 398             r01              = _mm256_mul_pd(rsq01,rinv01);
 399
 400             /* EWALD ELECTROSTATICS */
 401
 402             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 403             ewrt             = _mm256_mul_pd(r01,ewtabscale);
 404             ewitab           = _mm256_cvttpd_epi32(ewrt);
 405             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 406             ewitab           = _mm_slli_epi32(ewitab,2);
 407             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 408             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 409             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 410             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 411             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 412             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 413             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 414             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(rinv01,velec));
 415             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velecsum         = _mm256_add_pd(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             /* Calculate temporary vectorial force */
 423             tx               = _mm256_mul_pd(fscal,dx01);
 424             ty               = _mm256_mul_pd(fscal,dy01);
 425             tz               = _mm256_mul_pd(fscal,dz01);
 426
 427             /* Update vectorial force */
 428             fix0             = _mm256_add_pd(fix0,tx);
 429             fiy0             = _mm256_add_pd(fiy0,ty);
 430             fiz0             = _mm256_add_pd(fiz0,tz);
 431
 432             fjx1             = _mm256_add_pd(fjx1,tx);
 433             fjy1             = _mm256_add_pd(fjy1,ty);
 434             fjz1             = _mm256_add_pd(fjz1,tz);
 435
 436             /**************************
 437              * CALCULATE INTERACTIONS *
 438              **************************/
 439
 440             r02              = _mm256_mul_pd(rsq02,rinv02);
 441
 442             /* EWALD ELECTROSTATICS */
 443
 444             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 445             ewrt             = _mm256_mul_pd(r02,ewtabscale);
 446             ewitab           = _mm256_cvttpd_epi32(ewrt);
 447             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 448             ewitab           = _mm_slli_epi32(ewitab,2);
 449             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 450             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 451             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 452             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 453             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 454             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 455             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 456             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(rinv02,velec));
 457             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
 458
 459             /* Update potential sum for this i atom from the interaction with this j atom. */
 460             velecsum         = _mm256_add_pd(velecsum,velec);
 461
 462             fscal            = felec;
 463
 464             /* Calculate temporary vectorial force */
 465             tx               = _mm256_mul_pd(fscal,dx02);
 466             ty               = _mm256_mul_pd(fscal,dy02);
 467             tz               = _mm256_mul_pd(fscal,dz02);
 468
 469             /* Update vectorial force */
 470             fix0             = _mm256_add_pd(fix0,tx);
 471             fiy0             = _mm256_add_pd(fiy0,ty);
 472             fiz0             = _mm256_add_pd(fiz0,tz);
 473
 474             fjx2             = _mm256_add_pd(fjx2,tx);
 475             fjy2             = _mm256_add_pd(fjy2,ty);
 476             fjz2             = _mm256_add_pd(fjz2,tz);
 477
 478             /**************************
 479              * CALCULATE INTERACTIONS *
 480              **************************/
 481
 482             r10              = _mm256_mul_pd(rsq10,rinv10);
 483
 484             /* EWALD ELECTROSTATICS */
 485
 486             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 487             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 488             ewitab           = _mm256_cvttpd_epi32(ewrt);
 489             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 490             ewitab           = _mm_slli_epi32(ewitab,2);
 491             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 492             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 493             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 494             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 495             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 496             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 497             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 498             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
 499             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 500
 501             /* Update potential sum for this i atom from the interaction with this j atom. */
 502             velecsum         = _mm256_add_pd(velecsum,velec);
 503
 504             fscal            = felec;
 505
 506             /* Calculate temporary vectorial force */
 507             tx               = _mm256_mul_pd(fscal,dx10);
 508             ty               = _mm256_mul_pd(fscal,dy10);
 509             tz               = _mm256_mul_pd(fscal,dz10);
 510
 511             /* Update vectorial force */
 512             fix1             = _mm256_add_pd(fix1,tx);
 513             fiy1             = _mm256_add_pd(fiy1,ty);
 514             fiz1             = _mm256_add_pd(fiz1,tz);
 515
 516             fjx0             = _mm256_add_pd(fjx0,tx);
 517             fjy0             = _mm256_add_pd(fjy0,ty);
 518             fjz0             = _mm256_add_pd(fjz0,tz);
 519
 520             /**************************
 521              * CALCULATE INTERACTIONS *
 522              **************************/
 523
 524             r11              = _mm256_mul_pd(rsq11,rinv11);
 525
 526             /* EWALD ELECTROSTATICS */
 527
 528             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 529             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 530             ewitab           = _mm256_cvttpd_epi32(ewrt);
 531             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 532             ewitab           = _mm_slli_epi32(ewitab,2);
 533             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 534             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 535             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 536             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 537             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 538             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 539             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 540             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
 541             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 542
 543             /* Update potential sum for this i atom from the interaction with this j atom. */
 544             velecsum         = _mm256_add_pd(velecsum,velec);
 545
 546             fscal            = felec;
 547
 548             /* Calculate temporary vectorial force */
 549             tx               = _mm256_mul_pd(fscal,dx11);
 550             ty               = _mm256_mul_pd(fscal,dy11);
 551             tz               = _mm256_mul_pd(fscal,dz11);
 552
 553             /* Update vectorial force */
 554             fix1             = _mm256_add_pd(fix1,tx);
 555             fiy1             = _mm256_add_pd(fiy1,ty);
 556             fiz1             = _mm256_add_pd(fiz1,tz);
 557
 558             fjx1             = _mm256_add_pd(fjx1,tx);
 559             fjy1             = _mm256_add_pd(fjy1,ty);
 560             fjz1             = _mm256_add_pd(fjz1,tz);
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             r12              = _mm256_mul_pd(rsq12,rinv12);
 567
 568             /* EWALD ELECTROSTATICS */
 569
 570             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 571             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 572             ewitab           = _mm256_cvttpd_epi32(ewrt);
 573             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 574             ewitab           = _mm_slli_epi32(ewitab,2);
 575             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 576             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 577             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 578             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 579             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 580             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 581             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 582             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
 583             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 584
 585             /* Update potential sum for this i atom from the interaction with this j atom. */
 586             velecsum         = _mm256_add_pd(velecsum,velec);
 587
 588             fscal            = felec;
 589
 590             /* Calculate temporary vectorial force */
 591             tx               = _mm256_mul_pd(fscal,dx12);
 592             ty               = _mm256_mul_pd(fscal,dy12);
 593             tz               = _mm256_mul_pd(fscal,dz12);
 594
 595             /* Update vectorial force */
 596             fix1             = _mm256_add_pd(fix1,tx);
 597             fiy1             = _mm256_add_pd(fiy1,ty);
 598             fiz1             = _mm256_add_pd(fiz1,tz);
 599
 600             fjx2             = _mm256_add_pd(fjx2,tx);
 601             fjy2             = _mm256_add_pd(fjy2,ty);
 602             fjz2             = _mm256_add_pd(fjz2,tz);
 603
 604             /**************************
 605              * CALCULATE INTERACTIONS *
 606              **************************/
 607
 608             r20              = _mm256_mul_pd(rsq20,rinv20);
 609
 610             /* EWALD ELECTROSTATICS */
 611
 612             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 613             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 614             ewitab           = _mm256_cvttpd_epi32(ewrt);
 615             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 616             ewitab           = _mm_slli_epi32(ewitab,2);
 617             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 618             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 619             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 620             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 621             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 622             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 623             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 624             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
 625             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 626
 627             /* Update potential sum for this i atom from the interaction with this j atom. */
 628             velecsum         = _mm256_add_pd(velecsum,velec);
 629
 630             fscal            = felec;
 631
 632             /* Calculate temporary vectorial force */
 633             tx               = _mm256_mul_pd(fscal,dx20);
 634             ty               = _mm256_mul_pd(fscal,dy20);
 635             tz               = _mm256_mul_pd(fscal,dz20);
 636
 637             /* Update vectorial force */
 638             fix2             = _mm256_add_pd(fix2,tx);
 639             fiy2             = _mm256_add_pd(fiy2,ty);
 640             fiz2             = _mm256_add_pd(fiz2,tz);
 641
 642             fjx0             = _mm256_add_pd(fjx0,tx);
 643             fjy0             = _mm256_add_pd(fjy0,ty);
 644             fjz0             = _mm256_add_pd(fjz0,tz);
 645
 646             /**************************
 647              * CALCULATE INTERACTIONS *
 648              **************************/
 649
 650             r21              = _mm256_mul_pd(rsq21,rinv21);
 651
 652             /* EWALD ELECTROSTATICS */
 653
 654             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 655             ewrt             = _mm256_mul_pd(r21,ewtabscale);
 656             ewitab           = _mm256_cvttpd_epi32(ewrt);
 657             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 658             ewitab           = _mm_slli_epi32(ewitab,2);
 659             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 660             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 661             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 662             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 663             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 664             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 665             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 666             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
 667             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
 668
 669             /* Update potential sum for this i atom from the interaction with this j atom. */
 670             velecsum         = _mm256_add_pd(velecsum,velec);
 671
 672             fscal            = felec;
 673
 674             /* Calculate temporary vectorial force */
 675             tx               = _mm256_mul_pd(fscal,dx21);
 676             ty               = _mm256_mul_pd(fscal,dy21);
 677             tz               = _mm256_mul_pd(fscal,dz21);
 678
 679             /* Update vectorial force */
 680             fix2             = _mm256_add_pd(fix2,tx);
 681             fiy2             = _mm256_add_pd(fiy2,ty);
 682             fiz2             = _mm256_add_pd(fiz2,tz);
 683
 684             fjx1             = _mm256_add_pd(fjx1,tx);
 685             fjy1             = _mm256_add_pd(fjy1,ty);
 686             fjz1             = _mm256_add_pd(fjz1,tz);
 687
 688             /**************************
 689              * CALCULATE INTERACTIONS *
 690              **************************/
 691
 692             r22              = _mm256_mul_pd(rsq22,rinv22);
 693
 694             /* EWALD ELECTROSTATICS */
 695
 696             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 697             ewrt             = _mm256_mul_pd(r22,ewtabscale);
 698             ewitab           = _mm256_cvttpd_epi32(ewrt);
 699             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 700             ewitab           = _mm_slli_epi32(ewitab,2);
 701             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 702             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 703             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 704             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 705             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 706             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 707             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 708             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
 709             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
 710
 711             /* Update potential sum for this i atom from the interaction with this j atom. */
 712             velecsum         = _mm256_add_pd(velecsum,velec);
 713
 714             fscal            = felec;
 715
 716             /* Calculate temporary vectorial force */
 717             tx               = _mm256_mul_pd(fscal,dx22);
 718             ty               = _mm256_mul_pd(fscal,dy22);
 719             tz               = _mm256_mul_pd(fscal,dz22);
 720
 721             /* Update vectorial force */
 722             fix2             = _mm256_add_pd(fix2,tx);
 723             fiy2             = _mm256_add_pd(fiy2,ty);
 724             fiz2             = _mm256_add_pd(fiz2,tz);
 725
 726             fjx2             = _mm256_add_pd(fjx2,tx);
 727             fjy2             = _mm256_add_pd(fjy2,ty);
 728             fjz2             = _mm256_add_pd(fjz2,tz);
 729
 730             fjptrA             = f+j_coord_offsetA;
 731             fjptrB             = f+j_coord_offsetB;
 732             fjptrC             = f+j_coord_offsetC;
 733             fjptrD             = f+j_coord_offsetD;
 734
 735             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 736                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 737
 738             /* Inner loop uses 403 flops */
 739         }
 740
 741         if(jidx<j_index_end)
 742         {
 743
 744             /* Get j neighbor index, and coordinate index */
 745             jnrlistA         = jjnr[jidx];
 746             jnrlistB         = jjnr[jidx+1];
 747             jnrlistC         = jjnr[jidx+2];
 748             jnrlistD         = jjnr[jidx+3];
 749             /* Sign of each element will be negative for non-real atoms.
 750              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 751              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 752              */
 753             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 754
 755             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 756             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 757             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 758
 759             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 760             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 761             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 762             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 763             j_coord_offsetA  = DIM*jnrA;
 764             j_coord_offsetB  = DIM*jnrB;
 765             j_coord_offsetC  = DIM*jnrC;
 766             j_coord_offsetD  = DIM*jnrD;
 767
 768             /* load j atom coordinates */
 769             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 770                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 771                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 772
 773             /* Calculate displacement vector */
 774             dx00             = _mm256_sub_pd(ix0,jx0);
 775             dy00             = _mm256_sub_pd(iy0,jy0);
 776             dz00             = _mm256_sub_pd(iz0,jz0);
 777             dx01             = _mm256_sub_pd(ix0,jx1);
 778             dy01             = _mm256_sub_pd(iy0,jy1);
 779             dz01             = _mm256_sub_pd(iz0,jz1);
 780             dx02             = _mm256_sub_pd(ix0,jx2);
 781             dy02             = _mm256_sub_pd(iy0,jy2);
 782             dz02             = _mm256_sub_pd(iz0,jz2);
 783             dx10             = _mm256_sub_pd(ix1,jx0);
 784             dy10             = _mm256_sub_pd(iy1,jy0);
 785             dz10             = _mm256_sub_pd(iz1,jz0);
 786             dx11             = _mm256_sub_pd(ix1,jx1);
 787             dy11             = _mm256_sub_pd(iy1,jy1);
 788             dz11             = _mm256_sub_pd(iz1,jz1);
 789             dx12             = _mm256_sub_pd(ix1,jx2);
 790             dy12             = _mm256_sub_pd(iy1,jy2);
 791             dz12             = _mm256_sub_pd(iz1,jz2);
 792             dx20             = _mm256_sub_pd(ix2,jx0);
 793             dy20             = _mm256_sub_pd(iy2,jy0);
 794             dz20             = _mm256_sub_pd(iz2,jz0);
 795             dx21             = _mm256_sub_pd(ix2,jx1);
 796             dy21             = _mm256_sub_pd(iy2,jy1);
 797             dz21             = _mm256_sub_pd(iz2,jz1);
 798             dx22             = _mm256_sub_pd(ix2,jx2);
 799             dy22             = _mm256_sub_pd(iy2,jy2);
 800             dz22             = _mm256_sub_pd(iz2,jz2);
 801
 802             /* Calculate squared distance and things based on it */
 803             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 804             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 805             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 806             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 807             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 808             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 809             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 810             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 811             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 812
 813             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 814             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 815             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 816             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 817             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 818             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 819             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 820             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 821             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 822
 823             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 824             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 825             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 826             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 827             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 828             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 829             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 830             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 831             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 832
 833             fjx0             = _mm256_setzero_pd();
 834             fjy0             = _mm256_setzero_pd();
 835             fjz0             = _mm256_setzero_pd();
 836             fjx1             = _mm256_setzero_pd();
 837             fjy1             = _mm256_setzero_pd();
 838             fjz1             = _mm256_setzero_pd();
 839             fjx2             = _mm256_setzero_pd();
 840             fjy2             = _mm256_setzero_pd();
 841             fjz2             = _mm256_setzero_pd();
 842
 843             /**************************
 844              * CALCULATE INTERACTIONS *
 845              **************************/
 846
 847             r00              = _mm256_mul_pd(rsq00,rinv00);
 848             r00              = _mm256_andnot_pd(dummy_mask,r00);
 849
 850             /* Calculate table index by multiplying r with table scale and truncate to integer */
 851             rt               = _mm256_mul_pd(r00,vftabscale);
 852             vfitab           = _mm256_cvttpd_epi32(rt);
 853             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 854             vfitab           = _mm_slli_epi32(vfitab,3);
 855
 856             /* EWALD ELECTROSTATICS */
 857
 858             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 859             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 860             ewitab           = _mm256_cvttpd_epi32(ewrt);
 861             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 862             ewitab           = _mm_slli_epi32(ewitab,2);
 863             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 864             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 865             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 866             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 867             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 868             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 869             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 870             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 871             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 872
 873             /* CUBIC SPLINE TABLE DISPERSION */
 874             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 875             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 876             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 877             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 878             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 879             Heps             = _mm256_mul_pd(vfeps,H);
 880             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 881             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 882             vvdw6            = _mm256_mul_pd(c6_00,VV);
 883             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 884             fvdw6            = _mm256_mul_pd(c6_00,FF);
 885
 886             /* CUBIC SPLINE TABLE REPULSION */
 887             vfitab           = _mm_add_epi32(vfitab,ifour);
 888             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 889             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 890             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 891             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 892             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 893             Heps             = _mm256_mul_pd(vfeps,H);
 894             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 895             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 896             vvdw12           = _mm256_mul_pd(c12_00,VV);
 897             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 898             fvdw12           = _mm256_mul_pd(c12_00,FF);
 899             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 900             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 901
 902             /* Update potential sum for this i atom from the interaction with this j atom. */
 903             velec            = _mm256_andnot_pd(dummy_mask,velec);
 904             velecsum         = _mm256_add_pd(velecsum,velec);
 905             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 906             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 907
 908             fscal            = _mm256_add_pd(felec,fvdw);
 909
 910             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 911
 912             /* Calculate temporary vectorial force */
 913             tx               = _mm256_mul_pd(fscal,dx00);
 914             ty               = _mm256_mul_pd(fscal,dy00);
 915             tz               = _mm256_mul_pd(fscal,dz00);
 916
 917             /* Update vectorial force */
 918             fix0             = _mm256_add_pd(fix0,tx);
 919             fiy0             = _mm256_add_pd(fiy0,ty);
 920             fiz0             = _mm256_add_pd(fiz0,tz);
 921
 922             fjx0             = _mm256_add_pd(fjx0,tx);
 923             fjy0             = _mm256_add_pd(fjy0,ty);
 924             fjz0             = _mm256_add_pd(fjz0,tz);
 925
 926             /**************************
 927              * CALCULATE INTERACTIONS *
 928              **************************/
 929
 930             r01              = _mm256_mul_pd(rsq01,rinv01);
 931             r01              = _mm256_andnot_pd(dummy_mask,r01);
 932
 933             /* EWALD ELECTROSTATICS */
 934
 935             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 936             ewrt             = _mm256_mul_pd(r01,ewtabscale);
 937             ewitab           = _mm256_cvttpd_epi32(ewrt);
 938             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 939             ewitab           = _mm_slli_epi32(ewitab,2);
 940             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 941             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 942             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 943             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 944             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 945             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 946             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 947             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(rinv01,velec));
 948             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
 949
 950             /* Update potential sum for this i atom from the interaction with this j atom. */
 951             velec            = _mm256_andnot_pd(dummy_mask,velec);
 952             velecsum         = _mm256_add_pd(velecsum,velec);
 953
 954             fscal            = felec;
 955
 956             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 957
 958             /* Calculate temporary vectorial force */
 959             tx               = _mm256_mul_pd(fscal,dx01);
 960             ty               = _mm256_mul_pd(fscal,dy01);
 961             tz               = _mm256_mul_pd(fscal,dz01);
 962
 963             /* Update vectorial force */
 964             fix0             = _mm256_add_pd(fix0,tx);
 965             fiy0             = _mm256_add_pd(fiy0,ty);
 966             fiz0             = _mm256_add_pd(fiz0,tz);
 967
 968             fjx1             = _mm256_add_pd(fjx1,tx);
 969             fjy1             = _mm256_add_pd(fjy1,ty);
 970             fjz1             = _mm256_add_pd(fjz1,tz);
 971
 972             /**************************
 973              * CALCULATE INTERACTIONS *
 974              **************************/
 975
 976             r02              = _mm256_mul_pd(rsq02,rinv02);
 977             r02              = _mm256_andnot_pd(dummy_mask,r02);
 978
 979             /* EWALD ELECTROSTATICS */
 980
 981             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 982             ewrt             = _mm256_mul_pd(r02,ewtabscale);
 983             ewitab           = _mm256_cvttpd_epi32(ewrt);
 984             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 985             ewitab           = _mm_slli_epi32(ewitab,2);
 986             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 987             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 988             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 989             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 990             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 991             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 992             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 993             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(rinv02,velec));
 994             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
 995
 996             /* Update potential sum for this i atom from the interaction with this j atom. */
 997             velec            = _mm256_andnot_pd(dummy_mask,velec);
 998             velecsum         = _mm256_add_pd(velecsum,velec);
 999
1000             fscal            = felec;
1001
1002             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1003
1004             /* Calculate temporary vectorial force */
1005             tx               = _mm256_mul_pd(fscal,dx02);
1006             ty               = _mm256_mul_pd(fscal,dy02);
1007             tz               = _mm256_mul_pd(fscal,dz02);
1008
1009             /* Update vectorial force */
1010             fix0             = _mm256_add_pd(fix0,tx);
1011             fiy0             = _mm256_add_pd(fiy0,ty);
1012             fiz0             = _mm256_add_pd(fiz0,tz);
1013
1014             fjx2             = _mm256_add_pd(fjx2,tx);
1015             fjy2             = _mm256_add_pd(fjy2,ty);
1016             fjz2             = _mm256_add_pd(fjz2,tz);
1017
1018             /**************************
1019              * CALCULATE INTERACTIONS *
1020              **************************/
1021
1022             r10              = _mm256_mul_pd(rsq10,rinv10);
1023             r10              = _mm256_andnot_pd(dummy_mask,r10);
1024
1025             /* EWALD ELECTROSTATICS */
1026
1027             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1029             ewitab           = _mm256_cvttpd_epi32(ewrt);
1030             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1031             ewitab           = _mm_slli_epi32(ewitab,2);
1032             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1033             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1034             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1035             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1036             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1037             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1038             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1039             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1040             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1041
1042             /* Update potential sum for this i atom from the interaction with this j atom. */
1043             velec            = _mm256_andnot_pd(dummy_mask,velec);
1044             velecsum         = _mm256_add_pd(velecsum,velec);
1045
1046             fscal            = felec;
1047
1048             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1049
1050             /* Calculate temporary vectorial force */
1051             tx               = _mm256_mul_pd(fscal,dx10);
1052             ty               = _mm256_mul_pd(fscal,dy10);
1053             tz               = _mm256_mul_pd(fscal,dz10);
1054
1055             /* Update vectorial force */
1056             fix1             = _mm256_add_pd(fix1,tx);
1057             fiy1             = _mm256_add_pd(fiy1,ty);
1058             fiz1             = _mm256_add_pd(fiz1,tz);
1059
1060             fjx0             = _mm256_add_pd(fjx0,tx);
1061             fjy0             = _mm256_add_pd(fjy0,ty);
1062             fjz0             = _mm256_add_pd(fjz0,tz);
1063
1064             /**************************
1065              * CALCULATE INTERACTIONS *
1066              **************************/
1067
1068             r11              = _mm256_mul_pd(rsq11,rinv11);
1069             r11              = _mm256_andnot_pd(dummy_mask,r11);
1070
1071             /* EWALD ELECTROSTATICS */
1072
1073             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1074             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1075             ewitab           = _mm256_cvttpd_epi32(ewrt);
1076             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1077             ewitab           = _mm_slli_epi32(ewitab,2);
1078             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1079             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1080             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1081             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1082             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1083             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1084             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1085             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
1086             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1087
1088             /* Update potential sum for this i atom from the interaction with this j atom. */
1089             velec            = _mm256_andnot_pd(dummy_mask,velec);
1090             velecsum         = _mm256_add_pd(velecsum,velec);
1091
1092             fscal            = felec;
1093
1094             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1095
1096             /* Calculate temporary vectorial force */
1097             tx               = _mm256_mul_pd(fscal,dx11);
1098             ty               = _mm256_mul_pd(fscal,dy11);
1099             tz               = _mm256_mul_pd(fscal,dz11);
1100
1101             /* Update vectorial force */
1102             fix1             = _mm256_add_pd(fix1,tx);
1103             fiy1             = _mm256_add_pd(fiy1,ty);
1104             fiz1             = _mm256_add_pd(fiz1,tz);
1105
1106             fjx1             = _mm256_add_pd(fjx1,tx);
1107             fjy1             = _mm256_add_pd(fjy1,ty);
1108             fjz1             = _mm256_add_pd(fjz1,tz);
1109
1110             /**************************
1111              * CALCULATE INTERACTIONS *
1112              **************************/
1113
1114             r12              = _mm256_mul_pd(rsq12,rinv12);
1115             r12              = _mm256_andnot_pd(dummy_mask,r12);
1116
1117             /* EWALD ELECTROSTATICS */
1118
1119             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1120             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1121             ewitab           = _mm256_cvttpd_epi32(ewrt);
1122             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1123             ewitab           = _mm_slli_epi32(ewitab,2);
1124             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1125             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1126             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1127             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1128             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1129             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1130             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1131             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
1132             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1133
1134             /* Update potential sum for this i atom from the interaction with this j atom. */
1135             velec            = _mm256_andnot_pd(dummy_mask,velec);
1136             velecsum         = _mm256_add_pd(velecsum,velec);
1137
1138             fscal            = felec;
1139
1140             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1141
1142             /* Calculate temporary vectorial force */
1143             tx               = _mm256_mul_pd(fscal,dx12);
1144             ty               = _mm256_mul_pd(fscal,dy12);
1145             tz               = _mm256_mul_pd(fscal,dz12);
1146
1147             /* Update vectorial force */
1148             fix1             = _mm256_add_pd(fix1,tx);
1149             fiy1             = _mm256_add_pd(fiy1,ty);
1150             fiz1             = _mm256_add_pd(fiz1,tz);
1151
1152             fjx2             = _mm256_add_pd(fjx2,tx);
1153             fjy2             = _mm256_add_pd(fjy2,ty);
1154             fjz2             = _mm256_add_pd(fjz2,tz);
1155
1156             /**************************
1157              * CALCULATE INTERACTIONS *
1158              **************************/
1159
1160             r20              = _mm256_mul_pd(rsq20,rinv20);
1161             r20              = _mm256_andnot_pd(dummy_mask,r20);
1162
1163             /* EWALD ELECTROSTATICS */
1164
1165             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1166             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1167             ewitab           = _mm256_cvttpd_epi32(ewrt);
1168             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1169             ewitab           = _mm_slli_epi32(ewitab,2);
1170             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1171             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1172             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1173             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1174             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1175             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1176             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1177             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1178             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1179
1180             /* Update potential sum for this i atom from the interaction with this j atom. */
1181             velec            = _mm256_andnot_pd(dummy_mask,velec);
1182             velecsum         = _mm256_add_pd(velecsum,velec);
1183
1184             fscal            = felec;
1185
1186             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1187
1188             /* Calculate temporary vectorial force */
1189             tx               = _mm256_mul_pd(fscal,dx20);
1190             ty               = _mm256_mul_pd(fscal,dy20);
1191             tz               = _mm256_mul_pd(fscal,dz20);
1192
1193             /* Update vectorial force */
1194             fix2             = _mm256_add_pd(fix2,tx);
1195             fiy2             = _mm256_add_pd(fiy2,ty);
1196             fiz2             = _mm256_add_pd(fiz2,tz);
1197
1198             fjx0             = _mm256_add_pd(fjx0,tx);
1199             fjy0             = _mm256_add_pd(fjy0,ty);
1200             fjz0             = _mm256_add_pd(fjz0,tz);
1201
1202             /**************************
1203              * CALCULATE INTERACTIONS *
1204              **************************/
1205
1206             r21              = _mm256_mul_pd(rsq21,rinv21);
1207             r21              = _mm256_andnot_pd(dummy_mask,r21);
1208
1209             /* EWALD ELECTROSTATICS */
1210
1211             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1212             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1213             ewitab           = _mm256_cvttpd_epi32(ewrt);
1214             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1215             ewitab           = _mm_slli_epi32(ewitab,2);
1216             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1217             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1218             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1219             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1220             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1221             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1222             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1223             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
1224             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1225
1226             /* Update potential sum for this i atom from the interaction with this j atom. */
1227             velec            = _mm256_andnot_pd(dummy_mask,velec);
1228             velecsum         = _mm256_add_pd(velecsum,velec);
1229
1230             fscal            = felec;
1231
1232             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1233
1234             /* Calculate temporary vectorial force */
1235             tx               = _mm256_mul_pd(fscal,dx21);
1236             ty               = _mm256_mul_pd(fscal,dy21);
1237             tz               = _mm256_mul_pd(fscal,dz21);
1238
1239             /* Update vectorial force */
1240             fix2             = _mm256_add_pd(fix2,tx);
1241             fiy2             = _mm256_add_pd(fiy2,ty);
1242             fiz2             = _mm256_add_pd(fiz2,tz);
1243
1244             fjx1             = _mm256_add_pd(fjx1,tx);
1245             fjy1             = _mm256_add_pd(fjy1,ty);
1246             fjz1             = _mm256_add_pd(fjz1,tz);
1247
1248             /**************************
1249              * CALCULATE INTERACTIONS *
1250              **************************/
1251
1252             r22              = _mm256_mul_pd(rsq22,rinv22);
1253             r22              = _mm256_andnot_pd(dummy_mask,r22);
1254
1255             /* EWALD ELECTROSTATICS */
1256
1257             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1258             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1259             ewitab           = _mm256_cvttpd_epi32(ewrt);
1260             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1261             ewitab           = _mm_slli_epi32(ewitab,2);
1262             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1263             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1264             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1265             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1266             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1267             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1268             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1269             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
1270             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1271
1272             /* Update potential sum for this i atom from the interaction with this j atom. */
1273             velec            = _mm256_andnot_pd(dummy_mask,velec);
1274             velecsum         = _mm256_add_pd(velecsum,velec);
1275
1276             fscal            = felec;
1277
1278             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1279
1280             /* Calculate temporary vectorial force */
1281             tx               = _mm256_mul_pd(fscal,dx22);
1282             ty               = _mm256_mul_pd(fscal,dy22);
1283             tz               = _mm256_mul_pd(fscal,dz22);
1284
1285             /* Update vectorial force */
1286             fix2             = _mm256_add_pd(fix2,tx);
1287             fiy2             = _mm256_add_pd(fiy2,ty);
1288             fiz2             = _mm256_add_pd(fiz2,tz);
1289
1290             fjx2             = _mm256_add_pd(fjx2,tx);
1291             fjy2             = _mm256_add_pd(fjy2,ty);
1292             fjz2             = _mm256_add_pd(fjz2,tz);
1293
1294             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1295             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1296             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1297             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1298
1299             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1300                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1301
1302             /* Inner loop uses 412 flops */
1303         }
1304
1305         /* End of innermost loop */
1306
1307         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1308                                                  f+i_coord_offset,fshift+i_shift_offset);
1309
1310         ggid                        = gid[iidx];
1311         /* Update potential energies */
1312         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1313         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1314
1315         /* Increment number of inner iterations */
1316         inneriter                  += j_index_end - j_index_start;
1317
1318         /* Outer loop uses 20 flops */
1319     }
1320
1321     /* Increment number of outer iterations */
1322     outeriter        += nri;
1323
1324     /* Update outer/inner flops */
1325
1326     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*412);
1327 }
1328 /*
1329  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_avx_256_double
1330  * Electrostatics interaction: Ewald
1331  * VdW interaction:            CubicSplineTable
1332  * Geometry:                   Water3-Water3
1333  * Calculate force/pot:        Force
1334  */
1335 void
1336 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_avx_256_double
1337                     (t_nblist                    * gmx_restrict       nlist,
1338                      rvec                        * gmx_restrict          xx,
1339                      rvec                        * gmx_restrict          ff,
1340                      t_forcerec                  * gmx_restrict          fr,
1341                      t_mdatoms                   * gmx_restrict     mdatoms,
1342                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1343                      t_nrnb                      * gmx_restrict        nrnb)
1344 {
1345     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1346      * just 0 for non-waters.
1347      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1348      * jnr indices corresponding to data put in the four positions in the SIMD register.
1349      */
1350     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1351     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1352     int              jnrA,jnrB,jnrC,jnrD;
1353     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1354     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1355     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1356     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1357     real             rcutoff_scalar;
1358     real             *shiftvec,*fshift,*x,*f;
1359     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1360     real             scratch[4*DIM];
1361     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1362     real *           vdwioffsetptr0;
1363     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1364     real *           vdwioffsetptr1;
1365     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1366     real *           vdwioffsetptr2;
1367     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1368     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1369     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1370     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1371     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1372     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1373     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1374     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1375     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1376     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1377     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1378     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1379     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1380     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1381     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1382     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1383     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1384     real             *charge;
1385     int              nvdwtype;
1386     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1387     int              *vdwtype;
1388     real             *vdwparam;
1389     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1390     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1391     __m128i          vfitab;
1392     __m128i          ifour       = _mm_set1_epi32(4);
1393     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1394     real             *vftab;
1395     __m128i          ewitab;
1396     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1397     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1398     real             *ewtab;
1399     __m256d          dummy_mask,cutoff_mask;
1400     __m128           tmpmask0,tmpmask1;
1401     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1402     __m256d          one     = _mm256_set1_pd(1.0);
1403     __m256d          two     = _mm256_set1_pd(2.0);
1404     x                = xx[0];
1405     f                = ff[0];
1406
1407     nri              = nlist->nri;
1408     iinr             = nlist->iinr;
1409     jindex           = nlist->jindex;
1410     jjnr             = nlist->jjnr;
1411     shiftidx         = nlist->shift;
1412     gid              = nlist->gid;
1413     shiftvec         = fr->shift_vec[0];
1414     fshift           = fr->fshift[0];
1415     facel            = _mm256_set1_pd(fr->epsfac);
1416     charge           = mdatoms->chargeA;
1417     nvdwtype         = fr->ntype;
1418     vdwparam         = fr->nbfp;
1419     vdwtype          = mdatoms->typeA;
1420
1421     vftab            = kernel_data->table_vdw->data;
1422     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1423
1424     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
1425     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1426     beta2            = _mm256_mul_pd(beta,beta);
1427     beta3            = _mm256_mul_pd(beta,beta2);
1428
1429     ewtab            = fr->ic->tabq_coul_F;
1430     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
1431     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1432
1433     /* Setup water-specific parameters */
1434     inr              = nlist->iinr[0];
1435     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1436     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1437     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1438     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1439
1440     jq0              = _mm256_set1_pd(charge[inr+0]);
1441     jq1              = _mm256_set1_pd(charge[inr+1]);
1442     jq2              = _mm256_set1_pd(charge[inr+2]);
1443     vdwjidx0A        = 2*vdwtype[inr+0];
1444     qq00             = _mm256_mul_pd(iq0,jq0);
1445     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1446     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1447     qq01             = _mm256_mul_pd(iq0,jq1);
1448     qq02             = _mm256_mul_pd(iq0,jq2);
1449     qq10             = _mm256_mul_pd(iq1,jq0);
1450     qq11             = _mm256_mul_pd(iq1,jq1);
1451     qq12             = _mm256_mul_pd(iq1,jq2);
1452     qq20             = _mm256_mul_pd(iq2,jq0);
1453     qq21             = _mm256_mul_pd(iq2,jq1);
1454     qq22             = _mm256_mul_pd(iq2,jq2);
1455
1456     /* Avoid stupid compiler warnings */
1457     jnrA = jnrB = jnrC = jnrD = 0;
1458     j_coord_offsetA = 0;
1459     j_coord_offsetB = 0;
1460     j_coord_offsetC = 0;
1461     j_coord_offsetD = 0;
1462
1463     outeriter        = 0;
1464     inneriter        = 0;
1465
1466     for(iidx=0;iidx<4*DIM;iidx++)
1467     {
1468         scratch[iidx] = 0.0;
1469     }
1470
1471     /* Start outer loop over neighborlists */
1472     for(iidx=0; iidx<nri; iidx++)
1473     {
1474         /* Load shift vector for this list */
1475         i_shift_offset   = DIM*shiftidx[iidx];
1476
1477         /* Load limits for loop over neighbors */
1478         j_index_start    = jindex[iidx];
1479         j_index_end      = jindex[iidx+1];
1480
1481         /* Get outer coordinate index */
1482         inr              = iinr[iidx];
1483         i_coord_offset   = DIM*inr;
1484
1485         /* Load i particle coords and add shift vector */
1486         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1487                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1488
1489         fix0             = _mm256_setzero_pd();
1490         fiy0             = _mm256_setzero_pd();
1491         fiz0             = _mm256_setzero_pd();
1492         fix1             = _mm256_setzero_pd();
1493         fiy1             = _mm256_setzero_pd();
1494         fiz1             = _mm256_setzero_pd();
1495         fix2             = _mm256_setzero_pd();
1496         fiy2             = _mm256_setzero_pd();
1497         fiz2             = _mm256_setzero_pd();
1498
1499         /* Start inner kernel loop */
1500         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1501         {
1502
1503             /* Get j neighbor index, and coordinate index */
1504             jnrA             = jjnr[jidx];
1505             jnrB             = jjnr[jidx+1];
1506             jnrC             = jjnr[jidx+2];
1507             jnrD             = jjnr[jidx+3];
1508             j_coord_offsetA  = DIM*jnrA;
1509             j_coord_offsetB  = DIM*jnrB;
1510             j_coord_offsetC  = DIM*jnrC;
1511             j_coord_offsetD  = DIM*jnrD;
1512
1513             /* load j atom coordinates */
1514             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1515                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1516                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1517
1518             /* Calculate displacement vector */
1519             dx00             = _mm256_sub_pd(ix0,jx0);
1520             dy00             = _mm256_sub_pd(iy0,jy0);
1521             dz00             = _mm256_sub_pd(iz0,jz0);
1522             dx01             = _mm256_sub_pd(ix0,jx1);
1523             dy01             = _mm256_sub_pd(iy0,jy1);
1524             dz01             = _mm256_sub_pd(iz0,jz1);
1525             dx02             = _mm256_sub_pd(ix0,jx2);
1526             dy02             = _mm256_sub_pd(iy0,jy2);
1527             dz02             = _mm256_sub_pd(iz0,jz2);
1528             dx10             = _mm256_sub_pd(ix1,jx0);
1529             dy10             = _mm256_sub_pd(iy1,jy0);
1530             dz10             = _mm256_sub_pd(iz1,jz0);
1531             dx11             = _mm256_sub_pd(ix1,jx1);
1532             dy11             = _mm256_sub_pd(iy1,jy1);
1533             dz11             = _mm256_sub_pd(iz1,jz1);
1534             dx12             = _mm256_sub_pd(ix1,jx2);
1535             dy12             = _mm256_sub_pd(iy1,jy2);
1536             dz12             = _mm256_sub_pd(iz1,jz2);
1537             dx20             = _mm256_sub_pd(ix2,jx0);
1538             dy20             = _mm256_sub_pd(iy2,jy0);
1539             dz20             = _mm256_sub_pd(iz2,jz0);
1540             dx21             = _mm256_sub_pd(ix2,jx1);
1541             dy21             = _mm256_sub_pd(iy2,jy1);
1542             dz21             = _mm256_sub_pd(iz2,jz1);
1543             dx22             = _mm256_sub_pd(ix2,jx2);
1544             dy22             = _mm256_sub_pd(iy2,jy2);
1545             dz22             = _mm256_sub_pd(iz2,jz2);
1546
1547             /* Calculate squared distance and things based on it */
1548             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1549             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1550             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1551             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1552             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1553             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1554             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1555             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1556             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1557
1558             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1559             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1560             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1561             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1562             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1563             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1564             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1565             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1566             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1567
1568             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1569             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1570             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1571             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1572             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1573             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1574             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1575             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1576             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1577
1578             fjx0             = _mm256_setzero_pd();
1579             fjy0             = _mm256_setzero_pd();
1580             fjz0             = _mm256_setzero_pd();
1581             fjx1             = _mm256_setzero_pd();
1582             fjy1             = _mm256_setzero_pd();
1583             fjz1             = _mm256_setzero_pd();
1584             fjx2             = _mm256_setzero_pd();
1585             fjy2             = _mm256_setzero_pd();
1586             fjz2             = _mm256_setzero_pd();
1587
1588             /**************************
1589              * CALCULATE INTERACTIONS *
1590              **************************/
1591
1592             r00              = _mm256_mul_pd(rsq00,rinv00);
1593
1594             /* Calculate table index by multiplying r with table scale and truncate to integer */
1595             rt               = _mm256_mul_pd(r00,vftabscale);
1596             vfitab           = _mm256_cvttpd_epi32(rt);
1597             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1598             vfitab           = _mm_slli_epi32(vfitab,3);
1599
1600             /* EWALD ELECTROSTATICS */
1601
1602             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1603             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1604             ewitab           = _mm256_cvttpd_epi32(ewrt);
1605             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1606             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1607                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1608                                             &ewtabF,&ewtabFn);
1609             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1610             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1611
1612             /* CUBIC SPLINE TABLE DISPERSION */
1613             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1614             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1615             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1616             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1617             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1618             Heps             = _mm256_mul_pd(vfeps,H);
1619             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1620             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1621             fvdw6            = _mm256_mul_pd(c6_00,FF);
1622
1623             /* CUBIC SPLINE TABLE REPULSION */
1624             vfitab           = _mm_add_epi32(vfitab,ifour);
1625             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1626             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1627             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1628             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1629             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1630             Heps             = _mm256_mul_pd(vfeps,H);
1631             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1632             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1633             fvdw12           = _mm256_mul_pd(c12_00,FF);
1634             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1635
1636             fscal            = _mm256_add_pd(felec,fvdw);
1637
1638             /* Calculate temporary vectorial force */
1639             tx               = _mm256_mul_pd(fscal,dx00);
1640             ty               = _mm256_mul_pd(fscal,dy00);
1641             tz               = _mm256_mul_pd(fscal,dz00);
1642
1643             /* Update vectorial force */
1644             fix0             = _mm256_add_pd(fix0,tx);
1645             fiy0             = _mm256_add_pd(fiy0,ty);
1646             fiz0             = _mm256_add_pd(fiz0,tz);
1647
1648             fjx0             = _mm256_add_pd(fjx0,tx);
1649             fjy0             = _mm256_add_pd(fjy0,ty);
1650             fjz0             = _mm256_add_pd(fjz0,tz);
1651
1652             /**************************
1653              * CALCULATE INTERACTIONS *
1654              **************************/
1655
1656             r01              = _mm256_mul_pd(rsq01,rinv01);
1657
1658             /* EWALD ELECTROSTATICS */
1659
1660             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1661             ewrt             = _mm256_mul_pd(r01,ewtabscale);
1662             ewitab           = _mm256_cvttpd_epi32(ewrt);
1663             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1664             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1665                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1666                                             &ewtabF,&ewtabFn);
1667             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1668             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1669
1670             fscal            = felec;
1671
1672             /* Calculate temporary vectorial force */
1673             tx               = _mm256_mul_pd(fscal,dx01);
1674             ty               = _mm256_mul_pd(fscal,dy01);
1675             tz               = _mm256_mul_pd(fscal,dz01);
1676
1677             /* Update vectorial force */
1678             fix0             = _mm256_add_pd(fix0,tx);
1679             fiy0             = _mm256_add_pd(fiy0,ty);
1680             fiz0             = _mm256_add_pd(fiz0,tz);
1681
1682             fjx1             = _mm256_add_pd(fjx1,tx);
1683             fjy1             = _mm256_add_pd(fjy1,ty);
1684             fjz1             = _mm256_add_pd(fjz1,tz);
1685
1686             /**************************
1687              * CALCULATE INTERACTIONS *
1688              **************************/
1689
1690             r02              = _mm256_mul_pd(rsq02,rinv02);
1691
1692             /* EWALD ELECTROSTATICS */
1693
1694             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1695             ewrt             = _mm256_mul_pd(r02,ewtabscale);
1696             ewitab           = _mm256_cvttpd_epi32(ewrt);
1697             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1698             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1699                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1700                                             &ewtabF,&ewtabFn);
1701             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1702             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1703
1704             fscal            = felec;
1705
1706             /* Calculate temporary vectorial force */
1707             tx               = _mm256_mul_pd(fscal,dx02);
1708             ty               = _mm256_mul_pd(fscal,dy02);
1709             tz               = _mm256_mul_pd(fscal,dz02);
1710
1711             /* Update vectorial force */
1712             fix0             = _mm256_add_pd(fix0,tx);
1713             fiy0             = _mm256_add_pd(fiy0,ty);
1714             fiz0             = _mm256_add_pd(fiz0,tz);
1715
1716             fjx2             = _mm256_add_pd(fjx2,tx);
1717             fjy2             = _mm256_add_pd(fjy2,ty);
1718             fjz2             = _mm256_add_pd(fjz2,tz);
1719
1720             /**************************
1721              * CALCULATE INTERACTIONS *
1722              **************************/
1723
1724             r10              = _mm256_mul_pd(rsq10,rinv10);
1725
1726             /* EWALD ELECTROSTATICS */
1727
1728             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1729             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1730             ewitab           = _mm256_cvttpd_epi32(ewrt);
1731             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1732             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1733                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1734                                             &ewtabF,&ewtabFn);
1735             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1736             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1737
1738             fscal            = felec;
1739
1740             /* Calculate temporary vectorial force */
1741             tx               = _mm256_mul_pd(fscal,dx10);
1742             ty               = _mm256_mul_pd(fscal,dy10);
1743             tz               = _mm256_mul_pd(fscal,dz10);
1744
1745             /* Update vectorial force */
1746             fix1             = _mm256_add_pd(fix1,tx);
1747             fiy1             = _mm256_add_pd(fiy1,ty);
1748             fiz1             = _mm256_add_pd(fiz1,tz);
1749
1750             fjx0             = _mm256_add_pd(fjx0,tx);
1751             fjy0             = _mm256_add_pd(fjy0,ty);
1752             fjz0             = _mm256_add_pd(fjz0,tz);
1753
1754             /**************************
1755              * CALCULATE INTERACTIONS *
1756              **************************/
1757
1758             r11              = _mm256_mul_pd(rsq11,rinv11);
1759
1760             /* EWALD ELECTROSTATICS */
1761
1762             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1763             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1764             ewitab           = _mm256_cvttpd_epi32(ewrt);
1765             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1766             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1767                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1768                                             &ewtabF,&ewtabFn);
1769             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1770             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1771
1772             fscal            = felec;
1773
1774             /* Calculate temporary vectorial force */
1775             tx               = _mm256_mul_pd(fscal,dx11);
1776             ty               = _mm256_mul_pd(fscal,dy11);
1777             tz               = _mm256_mul_pd(fscal,dz11);
1778
1779             /* Update vectorial force */
1780             fix1             = _mm256_add_pd(fix1,tx);
1781             fiy1             = _mm256_add_pd(fiy1,ty);
1782             fiz1             = _mm256_add_pd(fiz1,tz);
1783
1784             fjx1             = _mm256_add_pd(fjx1,tx);
1785             fjy1             = _mm256_add_pd(fjy1,ty);
1786             fjz1             = _mm256_add_pd(fjz1,tz);
1787
1788             /**************************
1789              * CALCULATE INTERACTIONS *
1790              **************************/
1791
1792             r12              = _mm256_mul_pd(rsq12,rinv12);
1793
1794             /* EWALD ELECTROSTATICS */
1795
1796             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1797             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1798             ewitab           = _mm256_cvttpd_epi32(ewrt);
1799             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1800             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1801                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1802                                             &ewtabF,&ewtabFn);
1803             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1804             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1805
1806             fscal            = felec;
1807
1808             /* Calculate temporary vectorial force */
1809             tx               = _mm256_mul_pd(fscal,dx12);
1810             ty               = _mm256_mul_pd(fscal,dy12);
1811             tz               = _mm256_mul_pd(fscal,dz12);
1812
1813             /* Update vectorial force */
1814             fix1             = _mm256_add_pd(fix1,tx);
1815             fiy1             = _mm256_add_pd(fiy1,ty);
1816             fiz1             = _mm256_add_pd(fiz1,tz);
1817
1818             fjx2             = _mm256_add_pd(fjx2,tx);
1819             fjy2             = _mm256_add_pd(fjy2,ty);
1820             fjz2             = _mm256_add_pd(fjz2,tz);
1821
1822             /**************************
1823              * CALCULATE INTERACTIONS *
1824              **************************/
1825
1826             r20              = _mm256_mul_pd(rsq20,rinv20);
1827
1828             /* EWALD ELECTROSTATICS */
1829
1830             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1831             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1832             ewitab           = _mm256_cvttpd_epi32(ewrt);
1833             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1834             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1835                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1836                                             &ewtabF,&ewtabFn);
1837             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1838             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1839
1840             fscal            = felec;
1841
1842             /* Calculate temporary vectorial force */
1843             tx               = _mm256_mul_pd(fscal,dx20);
1844             ty               = _mm256_mul_pd(fscal,dy20);
1845             tz               = _mm256_mul_pd(fscal,dz20);
1846
1847             /* Update vectorial force */
1848             fix2             = _mm256_add_pd(fix2,tx);
1849             fiy2             = _mm256_add_pd(fiy2,ty);
1850             fiz2             = _mm256_add_pd(fiz2,tz);
1851
1852             fjx0             = _mm256_add_pd(fjx0,tx);
1853             fjy0             = _mm256_add_pd(fjy0,ty);
1854             fjz0             = _mm256_add_pd(fjz0,tz);
1855
1856             /**************************
1857              * CALCULATE INTERACTIONS *
1858              **************************/
1859
1860             r21              = _mm256_mul_pd(rsq21,rinv21);
1861
1862             /* EWALD ELECTROSTATICS */
1863
1864             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1865             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1866             ewitab           = _mm256_cvttpd_epi32(ewrt);
1867             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1868             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1869                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1870                                             &ewtabF,&ewtabFn);
1871             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1872             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1873
1874             fscal            = felec;
1875
1876             /* Calculate temporary vectorial force */
1877             tx               = _mm256_mul_pd(fscal,dx21);
1878             ty               = _mm256_mul_pd(fscal,dy21);
1879             tz               = _mm256_mul_pd(fscal,dz21);
1880
1881             /* Update vectorial force */
1882             fix2             = _mm256_add_pd(fix2,tx);
1883             fiy2             = _mm256_add_pd(fiy2,ty);
1884             fiz2             = _mm256_add_pd(fiz2,tz);
1885
1886             fjx1             = _mm256_add_pd(fjx1,tx);
1887             fjy1             = _mm256_add_pd(fjy1,ty);
1888             fjz1             = _mm256_add_pd(fjz1,tz);
1889
1890             /**************************
1891              * CALCULATE INTERACTIONS *
1892              **************************/
1893
1894             r22              = _mm256_mul_pd(rsq22,rinv22);
1895
1896             /* EWALD ELECTROSTATICS */
1897
1898             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1899             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1900             ewitab           = _mm256_cvttpd_epi32(ewrt);
1901             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1902             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1903                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1904                                             &ewtabF,&ewtabFn);
1905             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1906             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1907
1908             fscal            = felec;
1909
1910             /* Calculate temporary vectorial force */
1911             tx               = _mm256_mul_pd(fscal,dx22);
1912             ty               = _mm256_mul_pd(fscal,dy22);
1913             tz               = _mm256_mul_pd(fscal,dz22);
1914
1915             /* Update vectorial force */
1916             fix2             = _mm256_add_pd(fix2,tx);
1917             fiy2             = _mm256_add_pd(fiy2,ty);
1918             fiz2             = _mm256_add_pd(fiz2,tz);
1919
1920             fjx2             = _mm256_add_pd(fjx2,tx);
1921             fjy2             = _mm256_add_pd(fjy2,ty);
1922             fjz2             = _mm256_add_pd(fjz2,tz);
1923
1924             fjptrA             = f+j_coord_offsetA;
1925             fjptrB             = f+j_coord_offsetB;
1926             fjptrC             = f+j_coord_offsetC;
1927             fjptrD             = f+j_coord_offsetD;
1928
1929             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1930                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1931
1932             /* Inner loop uses 350 flops */
1933         }
1934
1935         if(jidx<j_index_end)
1936         {
1937
1938             /* Get j neighbor index, and coordinate index */
1939             jnrlistA         = jjnr[jidx];
1940             jnrlistB         = jjnr[jidx+1];
1941             jnrlistC         = jjnr[jidx+2];
1942             jnrlistD         = jjnr[jidx+3];
1943             /* Sign of each element will be negative for non-real atoms.
1944              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1945              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1946              */
1947             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1948
1949             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1950             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1951             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1952
1953             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1954             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1955             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1956             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1957             j_coord_offsetA  = DIM*jnrA;
1958             j_coord_offsetB  = DIM*jnrB;
1959             j_coord_offsetC  = DIM*jnrC;
1960             j_coord_offsetD  = DIM*jnrD;
1961
1962             /* load j atom coordinates */
1963             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1964                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1965                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1966
1967             /* Calculate displacement vector */
1968             dx00             = _mm256_sub_pd(ix0,jx0);
1969             dy00             = _mm256_sub_pd(iy0,jy0);
1970             dz00             = _mm256_sub_pd(iz0,jz0);
1971             dx01             = _mm256_sub_pd(ix0,jx1);
1972             dy01             = _mm256_sub_pd(iy0,jy1);
1973             dz01             = _mm256_sub_pd(iz0,jz1);
1974             dx02             = _mm256_sub_pd(ix0,jx2);
1975             dy02             = _mm256_sub_pd(iy0,jy2);
1976             dz02             = _mm256_sub_pd(iz0,jz2);
1977             dx10             = _mm256_sub_pd(ix1,jx0);
1978             dy10             = _mm256_sub_pd(iy1,jy0);
1979             dz10             = _mm256_sub_pd(iz1,jz0);
1980             dx11             = _mm256_sub_pd(ix1,jx1);
1981             dy11             = _mm256_sub_pd(iy1,jy1);
1982             dz11             = _mm256_sub_pd(iz1,jz1);
1983             dx12             = _mm256_sub_pd(ix1,jx2);
1984             dy12             = _mm256_sub_pd(iy1,jy2);
1985             dz12             = _mm256_sub_pd(iz1,jz2);
1986             dx20             = _mm256_sub_pd(ix2,jx0);
1987             dy20             = _mm256_sub_pd(iy2,jy0);
1988             dz20             = _mm256_sub_pd(iz2,jz0);
1989             dx21             = _mm256_sub_pd(ix2,jx1);
1990             dy21             = _mm256_sub_pd(iy2,jy1);
1991             dz21             = _mm256_sub_pd(iz2,jz1);
1992             dx22             = _mm256_sub_pd(ix2,jx2);
1993             dy22             = _mm256_sub_pd(iy2,jy2);
1994             dz22             = _mm256_sub_pd(iz2,jz2);
1995
1996             /* Calculate squared distance and things based on it */
1997             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1998             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1999             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
2000             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
2001             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2002             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2003             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
2004             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2005             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2006
2007             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
2008             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
2009             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
2010             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
2011             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
2012             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
2013             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
2014             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
2015             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
2016
2017             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
2018             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
2019             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
2020             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
2021             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
2022             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
2023             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
2024             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
2025             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
2026
2027             fjx0             = _mm256_setzero_pd();
2028             fjy0             = _mm256_setzero_pd();
2029             fjz0             = _mm256_setzero_pd();
2030             fjx1             = _mm256_setzero_pd();
2031             fjy1             = _mm256_setzero_pd();
2032             fjz1             = _mm256_setzero_pd();
2033             fjx2             = _mm256_setzero_pd();
2034             fjy2             = _mm256_setzero_pd();
2035             fjz2             = _mm256_setzero_pd();
2036
2037             /**************************
2038              * CALCULATE INTERACTIONS *
2039              **************************/
2040
2041             r00              = _mm256_mul_pd(rsq00,rinv00);
2042             r00              = _mm256_andnot_pd(dummy_mask,r00);
2043
2044             /* Calculate table index by multiplying r with table scale and truncate to integer */
2045             rt               = _mm256_mul_pd(r00,vftabscale);
2046             vfitab           = _mm256_cvttpd_epi32(rt);
2047             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2048             vfitab           = _mm_slli_epi32(vfitab,3);
2049
2050             /* EWALD ELECTROSTATICS */
2051
2052             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2053             ewrt             = _mm256_mul_pd(r00,ewtabscale);
2054             ewitab           = _mm256_cvttpd_epi32(ewrt);
2055             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2056             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2057                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2058                                             &ewtabF,&ewtabFn);
2059             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2060             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
2061
2062             /* CUBIC SPLINE TABLE DISPERSION */
2063             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2064             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2065             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2066             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2067             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2068             Heps             = _mm256_mul_pd(vfeps,H);
2069             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2070             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2071             fvdw6            = _mm256_mul_pd(c6_00,FF);
2072
2073             /* CUBIC SPLINE TABLE REPULSION */
2074             vfitab           = _mm_add_epi32(vfitab,ifour);
2075             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2076             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2077             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2078             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2079             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2080             Heps             = _mm256_mul_pd(vfeps,H);
2081             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2082             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2083             fvdw12           = _mm256_mul_pd(c12_00,FF);
2084             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
2085
2086             fscal            = _mm256_add_pd(felec,fvdw);
2087
2088             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2089
2090             /* Calculate temporary vectorial force */
2091             tx               = _mm256_mul_pd(fscal,dx00);
2092             ty               = _mm256_mul_pd(fscal,dy00);
2093             tz               = _mm256_mul_pd(fscal,dz00);
2094
2095             /* Update vectorial force */
2096             fix0             = _mm256_add_pd(fix0,tx);
2097             fiy0             = _mm256_add_pd(fiy0,ty);
2098             fiz0             = _mm256_add_pd(fiz0,tz);
2099
2100             fjx0             = _mm256_add_pd(fjx0,tx);
2101             fjy0             = _mm256_add_pd(fjy0,ty);
2102             fjz0             = _mm256_add_pd(fjz0,tz);
2103
2104             /**************************
2105              * CALCULATE INTERACTIONS *
2106              **************************/
2107
2108             r01              = _mm256_mul_pd(rsq01,rinv01);
2109             r01              = _mm256_andnot_pd(dummy_mask,r01);
2110
2111             /* EWALD ELECTROSTATICS */
2112
2113             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2114             ewrt             = _mm256_mul_pd(r01,ewtabscale);
2115             ewitab           = _mm256_cvttpd_epi32(ewrt);
2116             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2117             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2118                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2119                                             &ewtabF,&ewtabFn);
2120             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2121             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
2122
2123             fscal            = felec;
2124
2125             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2126
2127             /* Calculate temporary vectorial force */
2128             tx               = _mm256_mul_pd(fscal,dx01);
2129             ty               = _mm256_mul_pd(fscal,dy01);
2130             tz               = _mm256_mul_pd(fscal,dz01);
2131
2132             /* Update vectorial force */
2133             fix0             = _mm256_add_pd(fix0,tx);
2134             fiy0             = _mm256_add_pd(fiy0,ty);
2135             fiz0             = _mm256_add_pd(fiz0,tz);
2136
2137             fjx1             = _mm256_add_pd(fjx1,tx);
2138             fjy1             = _mm256_add_pd(fjy1,ty);
2139             fjz1             = _mm256_add_pd(fjz1,tz);
2140
2141             /**************************
2142              * CALCULATE INTERACTIONS *
2143              **************************/
2144
2145             r02              = _mm256_mul_pd(rsq02,rinv02);
2146             r02              = _mm256_andnot_pd(dummy_mask,r02);
2147
2148             /* EWALD ELECTROSTATICS */
2149
2150             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2151             ewrt             = _mm256_mul_pd(r02,ewtabscale);
2152             ewitab           = _mm256_cvttpd_epi32(ewrt);
2153             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2154             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2155                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2156                                             &ewtabF,&ewtabFn);
2157             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2158             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
2159
2160             fscal            = felec;
2161
2162             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2163
2164             /* Calculate temporary vectorial force */
2165             tx               = _mm256_mul_pd(fscal,dx02);
2166             ty               = _mm256_mul_pd(fscal,dy02);
2167             tz               = _mm256_mul_pd(fscal,dz02);
2168
2169             /* Update vectorial force */
2170             fix0             = _mm256_add_pd(fix0,tx);
2171             fiy0             = _mm256_add_pd(fiy0,ty);
2172             fiz0             = _mm256_add_pd(fiz0,tz);
2173
2174             fjx2             = _mm256_add_pd(fjx2,tx);
2175             fjy2             = _mm256_add_pd(fjy2,ty);
2176             fjz2             = _mm256_add_pd(fjz2,tz);
2177
2178             /**************************
2179              * CALCULATE INTERACTIONS *
2180              **************************/
2181
2182             r10              = _mm256_mul_pd(rsq10,rinv10);
2183             r10              = _mm256_andnot_pd(dummy_mask,r10);
2184
2185             /* EWALD ELECTROSTATICS */
2186
2187             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2188             ewrt             = _mm256_mul_pd(r10,ewtabscale);
2189             ewitab           = _mm256_cvttpd_epi32(ewrt);
2190             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2191             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2192                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2193                                             &ewtabF,&ewtabFn);
2194             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2195             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
2196
2197             fscal            = felec;
2198
2199             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2200
2201             /* Calculate temporary vectorial force */
2202             tx               = _mm256_mul_pd(fscal,dx10);
2203             ty               = _mm256_mul_pd(fscal,dy10);
2204             tz               = _mm256_mul_pd(fscal,dz10);
2205
2206             /* Update vectorial force */
2207             fix1             = _mm256_add_pd(fix1,tx);
2208             fiy1             = _mm256_add_pd(fiy1,ty);
2209             fiz1             = _mm256_add_pd(fiz1,tz);
2210
2211             fjx0             = _mm256_add_pd(fjx0,tx);
2212             fjy0             = _mm256_add_pd(fjy0,ty);
2213             fjz0             = _mm256_add_pd(fjz0,tz);
2214
2215             /**************************
2216              * CALCULATE INTERACTIONS *
2217              **************************/
2218
2219             r11              = _mm256_mul_pd(rsq11,rinv11);
2220             r11              = _mm256_andnot_pd(dummy_mask,r11);
2221
2222             /* EWALD ELECTROSTATICS */
2223
2224             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2225             ewrt             = _mm256_mul_pd(r11,ewtabscale);
2226             ewitab           = _mm256_cvttpd_epi32(ewrt);
2227             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2228             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2229                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2230                                             &ewtabF,&ewtabFn);
2231             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2232             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2233
2234             fscal            = felec;
2235
2236             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2237
2238             /* Calculate temporary vectorial force */
2239             tx               = _mm256_mul_pd(fscal,dx11);
2240             ty               = _mm256_mul_pd(fscal,dy11);
2241             tz               = _mm256_mul_pd(fscal,dz11);
2242
2243             /* Update vectorial force */
2244             fix1             = _mm256_add_pd(fix1,tx);
2245             fiy1             = _mm256_add_pd(fiy1,ty);
2246             fiz1             = _mm256_add_pd(fiz1,tz);
2247
2248             fjx1             = _mm256_add_pd(fjx1,tx);
2249             fjy1             = _mm256_add_pd(fjy1,ty);
2250             fjz1             = _mm256_add_pd(fjz1,tz);
2251
2252             /**************************
2253              * CALCULATE INTERACTIONS *
2254              **************************/
2255
2256             r12              = _mm256_mul_pd(rsq12,rinv12);
2257             r12              = _mm256_andnot_pd(dummy_mask,r12);
2258
2259             /* EWALD ELECTROSTATICS */
2260
2261             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2262             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2263             ewitab           = _mm256_cvttpd_epi32(ewrt);
2264             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2265             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2266                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2267                                             &ewtabF,&ewtabFn);
2268             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2269             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2270
2271             fscal            = felec;
2272
2273             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2274
2275             /* Calculate temporary vectorial force */
2276             tx               = _mm256_mul_pd(fscal,dx12);
2277             ty               = _mm256_mul_pd(fscal,dy12);
2278             tz               = _mm256_mul_pd(fscal,dz12);
2279
2280             /* Update vectorial force */
2281             fix1             = _mm256_add_pd(fix1,tx);
2282             fiy1             = _mm256_add_pd(fiy1,ty);
2283             fiz1             = _mm256_add_pd(fiz1,tz);
2284
2285             fjx2             = _mm256_add_pd(fjx2,tx);
2286             fjy2             = _mm256_add_pd(fjy2,ty);
2287             fjz2             = _mm256_add_pd(fjz2,tz);
2288
2289             /**************************
2290              * CALCULATE INTERACTIONS *
2291              **************************/
2292
2293             r20              = _mm256_mul_pd(rsq20,rinv20);
2294             r20              = _mm256_andnot_pd(dummy_mask,r20);
2295
2296             /* EWALD ELECTROSTATICS */
2297
2298             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2299             ewrt             = _mm256_mul_pd(r20,ewtabscale);
2300             ewitab           = _mm256_cvttpd_epi32(ewrt);
2301             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2302             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2303                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2304                                             &ewtabF,&ewtabFn);
2305             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2306             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2307
2308             fscal            = felec;
2309
2310             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2311
2312             /* Calculate temporary vectorial force */
2313             tx               = _mm256_mul_pd(fscal,dx20);
2314             ty               = _mm256_mul_pd(fscal,dy20);
2315             tz               = _mm256_mul_pd(fscal,dz20);
2316
2317             /* Update vectorial force */
2318             fix2             = _mm256_add_pd(fix2,tx);
2319             fiy2             = _mm256_add_pd(fiy2,ty);
2320             fiz2             = _mm256_add_pd(fiz2,tz);
2321
2322             fjx0             = _mm256_add_pd(fjx0,tx);
2323             fjy0             = _mm256_add_pd(fjy0,ty);
2324             fjz0             = _mm256_add_pd(fjz0,tz);
2325
2326             /**************************
2327              * CALCULATE INTERACTIONS *
2328              **************************/
2329
2330             r21              = _mm256_mul_pd(rsq21,rinv21);
2331             r21              = _mm256_andnot_pd(dummy_mask,r21);
2332
2333             /* EWALD ELECTROSTATICS */
2334
2335             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2336             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2337             ewitab           = _mm256_cvttpd_epi32(ewrt);
2338             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2339             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2340                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2341                                             &ewtabF,&ewtabFn);
2342             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2343             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2344
2345             fscal            = felec;
2346
2347             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2348
2349             /* Calculate temporary vectorial force */
2350             tx               = _mm256_mul_pd(fscal,dx21);
2351             ty               = _mm256_mul_pd(fscal,dy21);
2352             tz               = _mm256_mul_pd(fscal,dz21);
2353
2354             /* Update vectorial force */
2355             fix2             = _mm256_add_pd(fix2,tx);
2356             fiy2             = _mm256_add_pd(fiy2,ty);
2357             fiz2             = _mm256_add_pd(fiz2,tz);
2358
2359             fjx1             = _mm256_add_pd(fjx1,tx);
2360             fjy1             = _mm256_add_pd(fjy1,ty);
2361             fjz1             = _mm256_add_pd(fjz1,tz);
2362
2363             /**************************
2364              * CALCULATE INTERACTIONS *
2365              **************************/
2366
2367             r22              = _mm256_mul_pd(rsq22,rinv22);
2368             r22              = _mm256_andnot_pd(dummy_mask,r22);
2369
2370             /* EWALD ELECTROSTATICS */
2371
2372             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2373             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2374             ewitab           = _mm256_cvttpd_epi32(ewrt);
2375             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2376             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2377                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2378                                             &ewtabF,&ewtabFn);
2379             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2380             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2381
2382             fscal            = felec;
2383
2384             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2385
2386             /* Calculate temporary vectorial force */
2387             tx               = _mm256_mul_pd(fscal,dx22);
2388             ty               = _mm256_mul_pd(fscal,dy22);
2389             tz               = _mm256_mul_pd(fscal,dz22);
2390
2391             /* Update vectorial force */
2392             fix2             = _mm256_add_pd(fix2,tx);
2393             fiy2             = _mm256_add_pd(fiy2,ty);
2394             fiz2             = _mm256_add_pd(fiz2,tz);
2395
2396             fjx2             = _mm256_add_pd(fjx2,tx);
2397             fjy2             = _mm256_add_pd(fjy2,ty);
2398             fjz2             = _mm256_add_pd(fjz2,tz);
2399
2400             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2401             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2402             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2403             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2404
2405             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2406                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2407
2408             /* Inner loop uses 359 flops */
2409         }
2410
2411         /* End of innermost loop */
2412
2413         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2414                                                  f+i_coord_offset,fshift+i_shift_offset);
2415
2416         /* Increment number of inner iterations */
2417         inneriter                  += j_index_end - j_index_start;
2418
2419         /* Outer loop uses 18 flops */
2420     }
2421
2422     /* Increment number of outer iterations */
2423     outeriter        += nri;
2424
2425     /* Update outer/inner flops */
2426
2427     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*359);
2428 }