src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEw_VdwLJ_GeomW4W4_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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_double.h"
  48 #include "kernelutil_x86_avx_256_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  82     real             scratch[4*DIM];
  83     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  84     real *           vdwioffsetptr0;
  85     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  86     real *           vdwioffsetptr1;
  87     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  88     real *           vdwioffsetptr2;
  89     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  90     real *           vdwioffsetptr3;
  91     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  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     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  99     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 100     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 104     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 107     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 108     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 109     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 110     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 111     real             *charge;
 112     int              nvdwtype;
 113     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 114     int              *vdwtype;
 115     real             *vdwparam;
 116     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 117     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 118     __m128i          ewitab;
 119     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 120     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 121     real             *ewtab;
 122     __m256d          dummy_mask,cutoff_mask;
 123     __m128           tmpmask0,tmpmask1;
 124     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 125     __m256d          one     = _mm256_set1_pd(1.0);
 126     __m256d          two     = _mm256_set1_pd(2.0);
 127     x                = xx[0];
 128     f                = ff[0];
 129
 130     nri              = nlist->nri;
 131     iinr             = nlist->iinr;
 132     jindex           = nlist->jindex;
 133     jjnr             = nlist->jjnr;
 134     shiftidx         = nlist->shift;
 135     gid              = nlist->gid;
 136     shiftvec         = fr->shift_vec[0];
 137     fshift           = fr->fshift[0];
 138     facel            = _mm256_set1_pd(fr->epsfac);
 139     charge           = mdatoms->chargeA;
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 145     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 146     beta2            = _mm256_mul_pd(beta,beta);
 147     beta3            = _mm256_mul_pd(beta,beta2);
 148
 149     ewtab            = fr->ic->tabq_coul_FDV0;
 150     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 151     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 152
 153     /* Setup water-specific parameters */
 154     inr              = nlist->iinr[0];
 155     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 156     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 157     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 158     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 159
 160     jq1              = _mm256_set1_pd(charge[inr+1]);
 161     jq2              = _mm256_set1_pd(charge[inr+2]);
 162     jq3              = _mm256_set1_pd(charge[inr+3]);
 163     vdwjidx0A        = 2*vdwtype[inr+0];
 164     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 165     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 166     qq11             = _mm256_mul_pd(iq1,jq1);
 167     qq12             = _mm256_mul_pd(iq1,jq2);
 168     qq13             = _mm256_mul_pd(iq1,jq3);
 169     qq21             = _mm256_mul_pd(iq2,jq1);
 170     qq22             = _mm256_mul_pd(iq2,jq2);
 171     qq23             = _mm256_mul_pd(iq2,jq3);
 172     qq31             = _mm256_mul_pd(iq3,jq1);
 173     qq32             = _mm256_mul_pd(iq3,jq2);
 174     qq33             = _mm256_mul_pd(iq3,jq3);
 175
 176     /* Avoid stupid compiler warnings */
 177     jnrA = jnrB = jnrC = jnrD = 0;
 178     j_coord_offsetA = 0;
 179     j_coord_offsetB = 0;
 180     j_coord_offsetC = 0;
 181     j_coord_offsetD = 0;
 182
 183     outeriter        = 0;
 184     inneriter        = 0;
 185
 186     for(iidx=0;iidx<4*DIM;iidx++)
 187     {
 188         scratch[iidx] = 0.0;
 189     }
 190
 191     /* Start outer loop over neighborlists */
 192     for(iidx=0; iidx<nri; iidx++)
 193     {
 194         /* Load shift vector for this list */
 195         i_shift_offset   = DIM*shiftidx[iidx];
 196
 197         /* Load limits for loop over neighbors */
 198         j_index_start    = jindex[iidx];
 199         j_index_end      = jindex[iidx+1];
 200
 201         /* Get outer coordinate index */
 202         inr              = iinr[iidx];
 203         i_coord_offset   = DIM*inr;
 204
 205         /* Load i particle coords and add shift vector */
 206         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 207                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 208
 209         fix0             = _mm256_setzero_pd();
 210         fiy0             = _mm256_setzero_pd();
 211         fiz0             = _mm256_setzero_pd();
 212         fix1             = _mm256_setzero_pd();
 213         fiy1             = _mm256_setzero_pd();
 214         fiz1             = _mm256_setzero_pd();
 215         fix2             = _mm256_setzero_pd();
 216         fiy2             = _mm256_setzero_pd();
 217         fiz2             = _mm256_setzero_pd();
 218         fix3             = _mm256_setzero_pd();
 219         fiy3             = _mm256_setzero_pd();
 220         fiz3             = _mm256_setzero_pd();
 221
 222         /* Reset potential sums */
 223         velecsum         = _mm256_setzero_pd();
 224         vvdwsum          = _mm256_setzero_pd();
 225
 226         /* Start inner kernel loop */
 227         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 228         {
 229
 230             /* Get j neighbor index, and coordinate index */
 231             jnrA             = jjnr[jidx];
 232             jnrB             = jjnr[jidx+1];
 233             jnrC             = jjnr[jidx+2];
 234             jnrD             = jjnr[jidx+3];
 235             j_coord_offsetA  = DIM*jnrA;
 236             j_coord_offsetB  = DIM*jnrB;
 237             j_coord_offsetC  = DIM*jnrC;
 238             j_coord_offsetD  = DIM*jnrD;
 239
 240             /* load j atom coordinates */
 241             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 242                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 243                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 244                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 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             dx11             = _mm256_sub_pd(ix1,jx1);
 251             dy11             = _mm256_sub_pd(iy1,jy1);
 252             dz11             = _mm256_sub_pd(iz1,jz1);
 253             dx12             = _mm256_sub_pd(ix1,jx2);
 254             dy12             = _mm256_sub_pd(iy1,jy2);
 255             dz12             = _mm256_sub_pd(iz1,jz2);
 256             dx13             = _mm256_sub_pd(ix1,jx3);
 257             dy13             = _mm256_sub_pd(iy1,jy3);
 258             dz13             = _mm256_sub_pd(iz1,jz3);
 259             dx21             = _mm256_sub_pd(ix2,jx1);
 260             dy21             = _mm256_sub_pd(iy2,jy1);
 261             dz21             = _mm256_sub_pd(iz2,jz1);
 262             dx22             = _mm256_sub_pd(ix2,jx2);
 263             dy22             = _mm256_sub_pd(iy2,jy2);
 264             dz22             = _mm256_sub_pd(iz2,jz2);
 265             dx23             = _mm256_sub_pd(ix2,jx3);
 266             dy23             = _mm256_sub_pd(iy2,jy3);
 267             dz23             = _mm256_sub_pd(iz2,jz3);
 268             dx31             = _mm256_sub_pd(ix3,jx1);
 269             dy31             = _mm256_sub_pd(iy3,jy1);
 270             dz31             = _mm256_sub_pd(iz3,jz1);
 271             dx32             = _mm256_sub_pd(ix3,jx2);
 272             dy32             = _mm256_sub_pd(iy3,jy2);
 273             dz32             = _mm256_sub_pd(iz3,jz2);
 274             dx33             = _mm256_sub_pd(ix3,jx3);
 275             dy33             = _mm256_sub_pd(iy3,jy3);
 276             dz33             = _mm256_sub_pd(iz3,jz3);
 277
 278             /* Calculate squared distance and things based on it */
 279             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 280             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 281             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 282             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 283             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 284             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 285             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 286             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 287             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 288             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 289
 290             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 291             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 292             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 293             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 294             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 295             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 296             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 297             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 298             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 299
 300             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 301             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 302             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 303             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 304             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 305             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 306             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 307             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 308             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 309             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 310
 311             fjx0             = _mm256_setzero_pd();
 312             fjy0             = _mm256_setzero_pd();
 313             fjz0             = _mm256_setzero_pd();
 314             fjx1             = _mm256_setzero_pd();
 315             fjy1             = _mm256_setzero_pd();
 316             fjz1             = _mm256_setzero_pd();
 317             fjx2             = _mm256_setzero_pd();
 318             fjy2             = _mm256_setzero_pd();
 319             fjz2             = _mm256_setzero_pd();
 320             fjx3             = _mm256_setzero_pd();
 321             fjy3             = _mm256_setzero_pd();
 322             fjz3             = _mm256_setzero_pd();
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             /* LENNARD-JONES DISPERSION/REPULSION */
 329
 330             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 331             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 332             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 333             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 334             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 338
 339             fscal            = fvdw;
 340
 341             /* Calculate temporary vectorial force */
 342             tx               = _mm256_mul_pd(fscal,dx00);
 343             ty               = _mm256_mul_pd(fscal,dy00);
 344             tz               = _mm256_mul_pd(fscal,dz00);
 345
 346             /* Update vectorial force */
 347             fix0             = _mm256_add_pd(fix0,tx);
 348             fiy0             = _mm256_add_pd(fiy0,ty);
 349             fiz0             = _mm256_add_pd(fiz0,tz);
 350
 351             fjx0             = _mm256_add_pd(fjx0,tx);
 352             fjy0             = _mm256_add_pd(fjy0,ty);
 353             fjz0             = _mm256_add_pd(fjz0,tz);
 354
 355             /**************************
 356              * CALCULATE INTERACTIONS *
 357              **************************/
 358
 359             r11              = _mm256_mul_pd(rsq11,rinv11);
 360
 361             /* EWALD ELECTROSTATICS */
 362
 363             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 364             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 365             ewitab           = _mm256_cvttpd_epi32(ewrt);
 366             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 367             ewitab           = _mm_slli_epi32(ewitab,2);
 368             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 369             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 370             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 371             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 372             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 373             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 374             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 375             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
 376             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 377
 378             /* Update potential sum for this i atom from the interaction with this j atom. */
 379             velecsum         = _mm256_add_pd(velecsum,velec);
 380
 381             fscal            = felec;
 382
 383             /* Calculate temporary vectorial force */
 384             tx               = _mm256_mul_pd(fscal,dx11);
 385             ty               = _mm256_mul_pd(fscal,dy11);
 386             tz               = _mm256_mul_pd(fscal,dz11);
 387
 388             /* Update vectorial force */
 389             fix1             = _mm256_add_pd(fix1,tx);
 390             fiy1             = _mm256_add_pd(fiy1,ty);
 391             fiz1             = _mm256_add_pd(fiz1,tz);
 392
 393             fjx1             = _mm256_add_pd(fjx1,tx);
 394             fjy1             = _mm256_add_pd(fjy1,ty);
 395             fjz1             = _mm256_add_pd(fjz1,tz);
 396
 397             /**************************
 398              * CALCULATE INTERACTIONS *
 399              **************************/
 400
 401             r12              = _mm256_mul_pd(rsq12,rinv12);
 402
 403             /* EWALD ELECTROSTATICS */
 404
 405             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 406             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 407             ewitab           = _mm256_cvttpd_epi32(ewrt);
 408             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 409             ewitab           = _mm_slli_epi32(ewitab,2);
 410             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 411             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 412             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 413             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 414             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 415             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 416             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 417             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
 418             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 419
 420             /* Update potential sum for this i atom from the interaction with this j atom. */
 421             velecsum         = _mm256_add_pd(velecsum,velec);
 422
 423             fscal            = felec;
 424
 425             /* Calculate temporary vectorial force */
 426             tx               = _mm256_mul_pd(fscal,dx12);
 427             ty               = _mm256_mul_pd(fscal,dy12);
 428             tz               = _mm256_mul_pd(fscal,dz12);
 429
 430             /* Update vectorial force */
 431             fix1             = _mm256_add_pd(fix1,tx);
 432             fiy1             = _mm256_add_pd(fiy1,ty);
 433             fiz1             = _mm256_add_pd(fiz1,tz);
 434
 435             fjx2             = _mm256_add_pd(fjx2,tx);
 436             fjy2             = _mm256_add_pd(fjy2,ty);
 437             fjz2             = _mm256_add_pd(fjz2,tz);
 438
 439             /**************************
 440              * CALCULATE INTERACTIONS *
 441              **************************/
 442
 443             r13              = _mm256_mul_pd(rsq13,rinv13);
 444
 445             /* EWALD ELECTROSTATICS */
 446
 447             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 448             ewrt             = _mm256_mul_pd(r13,ewtabscale);
 449             ewitab           = _mm256_cvttpd_epi32(ewrt);
 450             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 451             ewitab           = _mm_slli_epi32(ewitab,2);
 452             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 453             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 454             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 455             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 456             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 457             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 458             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 459             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(rinv13,velec));
 460             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
 461
 462             /* Update potential sum for this i atom from the interaction with this j atom. */
 463             velecsum         = _mm256_add_pd(velecsum,velec);
 464
 465             fscal            = felec;
 466
 467             /* Calculate temporary vectorial force */
 468             tx               = _mm256_mul_pd(fscal,dx13);
 469             ty               = _mm256_mul_pd(fscal,dy13);
 470             tz               = _mm256_mul_pd(fscal,dz13);
 471
 472             /* Update vectorial force */
 473             fix1             = _mm256_add_pd(fix1,tx);
 474             fiy1             = _mm256_add_pd(fiy1,ty);
 475             fiz1             = _mm256_add_pd(fiz1,tz);
 476
 477             fjx3             = _mm256_add_pd(fjx3,tx);
 478             fjy3             = _mm256_add_pd(fjy3,ty);
 479             fjz3             = _mm256_add_pd(fjz3,tz);
 480
 481             /**************************
 482              * CALCULATE INTERACTIONS *
 483              **************************/
 484
 485             r21              = _mm256_mul_pd(rsq21,rinv21);
 486
 487             /* EWALD ELECTROSTATICS */
 488
 489             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 490             ewrt             = _mm256_mul_pd(r21,ewtabscale);
 491             ewitab           = _mm256_cvttpd_epi32(ewrt);
 492             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 493             ewitab           = _mm_slli_epi32(ewitab,2);
 494             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 495             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 496             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 497             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 498             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 499             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 500             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 501             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
 502             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
 503
 504             /* Update potential sum for this i atom from the interaction with this j atom. */
 505             velecsum         = _mm256_add_pd(velecsum,velec);
 506
 507             fscal            = felec;
 508
 509             /* Calculate temporary vectorial force */
 510             tx               = _mm256_mul_pd(fscal,dx21);
 511             ty               = _mm256_mul_pd(fscal,dy21);
 512             tz               = _mm256_mul_pd(fscal,dz21);
 513
 514             /* Update vectorial force */
 515             fix2             = _mm256_add_pd(fix2,tx);
 516             fiy2             = _mm256_add_pd(fiy2,ty);
 517             fiz2             = _mm256_add_pd(fiz2,tz);
 518
 519             fjx1             = _mm256_add_pd(fjx1,tx);
 520             fjy1             = _mm256_add_pd(fjy1,ty);
 521             fjz1             = _mm256_add_pd(fjz1,tz);
 522
 523             /**************************
 524              * CALCULATE INTERACTIONS *
 525              **************************/
 526
 527             r22              = _mm256_mul_pd(rsq22,rinv22);
 528
 529             /* EWALD ELECTROSTATICS */
 530
 531             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 532             ewrt             = _mm256_mul_pd(r22,ewtabscale);
 533             ewitab           = _mm256_cvttpd_epi32(ewrt);
 534             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 535             ewitab           = _mm_slli_epi32(ewitab,2);
 536             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 537             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 538             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 539             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 540             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 541             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 542             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 543             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
 544             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             velecsum         = _mm256_add_pd(velecsum,velec);
 548
 549             fscal            = felec;
 550
 551             /* Calculate temporary vectorial force */
 552             tx               = _mm256_mul_pd(fscal,dx22);
 553             ty               = _mm256_mul_pd(fscal,dy22);
 554             tz               = _mm256_mul_pd(fscal,dz22);
 555
 556             /* Update vectorial force */
 557             fix2             = _mm256_add_pd(fix2,tx);
 558             fiy2             = _mm256_add_pd(fiy2,ty);
 559             fiz2             = _mm256_add_pd(fiz2,tz);
 560
 561             fjx2             = _mm256_add_pd(fjx2,tx);
 562             fjy2             = _mm256_add_pd(fjy2,ty);
 563             fjz2             = _mm256_add_pd(fjz2,tz);
 564
 565             /**************************
 566              * CALCULATE INTERACTIONS *
 567              **************************/
 568
 569             r23              = _mm256_mul_pd(rsq23,rinv23);
 570
 571             /* EWALD ELECTROSTATICS */
 572
 573             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 574             ewrt             = _mm256_mul_pd(r23,ewtabscale);
 575             ewitab           = _mm256_cvttpd_epi32(ewrt);
 576             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 577             ewitab           = _mm_slli_epi32(ewitab,2);
 578             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 579             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 580             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 581             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 582             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 583             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 584             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 585             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(rinv23,velec));
 586             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
 587
 588             /* Update potential sum for this i atom from the interaction with this j atom. */
 589             velecsum         = _mm256_add_pd(velecsum,velec);
 590
 591             fscal            = felec;
 592
 593             /* Calculate temporary vectorial force */
 594             tx               = _mm256_mul_pd(fscal,dx23);
 595             ty               = _mm256_mul_pd(fscal,dy23);
 596             tz               = _mm256_mul_pd(fscal,dz23);
 597
 598             /* Update vectorial force */
 599             fix2             = _mm256_add_pd(fix2,tx);
 600             fiy2             = _mm256_add_pd(fiy2,ty);
 601             fiz2             = _mm256_add_pd(fiz2,tz);
 602
 603             fjx3             = _mm256_add_pd(fjx3,tx);
 604             fjy3             = _mm256_add_pd(fjy3,ty);
 605             fjz3             = _mm256_add_pd(fjz3,tz);
 606
 607             /**************************
 608              * CALCULATE INTERACTIONS *
 609              **************************/
 610
 611             r31              = _mm256_mul_pd(rsq31,rinv31);
 612
 613             /* EWALD ELECTROSTATICS */
 614
 615             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 616             ewrt             = _mm256_mul_pd(r31,ewtabscale);
 617             ewitab           = _mm256_cvttpd_epi32(ewrt);
 618             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 619             ewitab           = _mm_slli_epi32(ewitab,2);
 620             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 621             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 622             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 623             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 624             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 625             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 626             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 627             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(rinv31,velec));
 628             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
 629
 630             /* Update potential sum for this i atom from the interaction with this j atom. */
 631             velecsum         = _mm256_add_pd(velecsum,velec);
 632
 633             fscal            = felec;
 634
 635             /* Calculate temporary vectorial force */
 636             tx               = _mm256_mul_pd(fscal,dx31);
 637             ty               = _mm256_mul_pd(fscal,dy31);
 638             tz               = _mm256_mul_pd(fscal,dz31);
 639
 640             /* Update vectorial force */
 641             fix3             = _mm256_add_pd(fix3,tx);
 642             fiy3             = _mm256_add_pd(fiy3,ty);
 643             fiz3             = _mm256_add_pd(fiz3,tz);
 644
 645             fjx1             = _mm256_add_pd(fjx1,tx);
 646             fjy1             = _mm256_add_pd(fjy1,ty);
 647             fjz1             = _mm256_add_pd(fjz1,tz);
 648
 649             /**************************
 650              * CALCULATE INTERACTIONS *
 651              **************************/
 652
 653             r32              = _mm256_mul_pd(rsq32,rinv32);
 654
 655             /* EWALD ELECTROSTATICS */
 656
 657             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 658             ewrt             = _mm256_mul_pd(r32,ewtabscale);
 659             ewitab           = _mm256_cvttpd_epi32(ewrt);
 660             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 661             ewitab           = _mm_slli_epi32(ewitab,2);
 662             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 663             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 664             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 665             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 666             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 667             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 668             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 669             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(rinv32,velec));
 670             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
 671
 672             /* Update potential sum for this i atom from the interaction with this j atom. */
 673             velecsum         = _mm256_add_pd(velecsum,velec);
 674
 675             fscal            = felec;
 676
 677             /* Calculate temporary vectorial force */
 678             tx               = _mm256_mul_pd(fscal,dx32);
 679             ty               = _mm256_mul_pd(fscal,dy32);
 680             tz               = _mm256_mul_pd(fscal,dz32);
 681
 682             /* Update vectorial force */
 683             fix3             = _mm256_add_pd(fix3,tx);
 684             fiy3             = _mm256_add_pd(fiy3,ty);
 685             fiz3             = _mm256_add_pd(fiz3,tz);
 686
 687             fjx2             = _mm256_add_pd(fjx2,tx);
 688             fjy2             = _mm256_add_pd(fjy2,ty);
 689             fjz2             = _mm256_add_pd(fjz2,tz);
 690
 691             /**************************
 692              * CALCULATE INTERACTIONS *
 693              **************************/
 694
 695             r33              = _mm256_mul_pd(rsq33,rinv33);
 696
 697             /* EWALD ELECTROSTATICS */
 698
 699             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 700             ewrt             = _mm256_mul_pd(r33,ewtabscale);
 701             ewitab           = _mm256_cvttpd_epi32(ewrt);
 702             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 703             ewitab           = _mm_slli_epi32(ewitab,2);
 704             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 705             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 706             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 707             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 708             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 709             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 710             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 711             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(rinv33,velec));
 712             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
 713
 714             /* Update potential sum for this i atom from the interaction with this j atom. */
 715             velecsum         = _mm256_add_pd(velecsum,velec);
 716
 717             fscal            = felec;
 718
 719             /* Calculate temporary vectorial force */
 720             tx               = _mm256_mul_pd(fscal,dx33);
 721             ty               = _mm256_mul_pd(fscal,dy33);
 722             tz               = _mm256_mul_pd(fscal,dz33);
 723
 724             /* Update vectorial force */
 725             fix3             = _mm256_add_pd(fix3,tx);
 726             fiy3             = _mm256_add_pd(fiy3,ty);
 727             fiz3             = _mm256_add_pd(fiz3,tz);
 728
 729             fjx3             = _mm256_add_pd(fjx3,tx);
 730             fjy3             = _mm256_add_pd(fjy3,ty);
 731             fjz3             = _mm256_add_pd(fjz3,tz);
 732
 733             fjptrA             = f+j_coord_offsetA;
 734             fjptrB             = f+j_coord_offsetB;
 735             fjptrC             = f+j_coord_offsetC;
 736             fjptrD             = f+j_coord_offsetD;
 737
 738             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 739                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 740                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 741
 742             /* Inner loop uses 404 flops */
 743         }
 744
 745         if(jidx<j_index_end)
 746         {
 747
 748             /* Get j neighbor index, and coordinate index */
 749             jnrlistA         = jjnr[jidx];
 750             jnrlistB         = jjnr[jidx+1];
 751             jnrlistC         = jjnr[jidx+2];
 752             jnrlistD         = jjnr[jidx+3];
 753             /* Sign of each element will be negative for non-real atoms.
 754              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 755              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 756              */
 757             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 758
 759             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 760             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 761             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 762
 763             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 764             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 765             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 766             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 767             j_coord_offsetA  = DIM*jnrA;
 768             j_coord_offsetB  = DIM*jnrB;
 769             j_coord_offsetC  = DIM*jnrC;
 770             j_coord_offsetD  = DIM*jnrD;
 771
 772             /* load j atom coordinates */
 773             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 774                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 775                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 776                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 777
 778             /* Calculate displacement vector */
 779             dx00             = _mm256_sub_pd(ix0,jx0);
 780             dy00             = _mm256_sub_pd(iy0,jy0);
 781             dz00             = _mm256_sub_pd(iz0,jz0);
 782             dx11             = _mm256_sub_pd(ix1,jx1);
 783             dy11             = _mm256_sub_pd(iy1,jy1);
 784             dz11             = _mm256_sub_pd(iz1,jz1);
 785             dx12             = _mm256_sub_pd(ix1,jx2);
 786             dy12             = _mm256_sub_pd(iy1,jy2);
 787             dz12             = _mm256_sub_pd(iz1,jz2);
 788             dx13             = _mm256_sub_pd(ix1,jx3);
 789             dy13             = _mm256_sub_pd(iy1,jy3);
 790             dz13             = _mm256_sub_pd(iz1,jz3);
 791             dx21             = _mm256_sub_pd(ix2,jx1);
 792             dy21             = _mm256_sub_pd(iy2,jy1);
 793             dz21             = _mm256_sub_pd(iz2,jz1);
 794             dx22             = _mm256_sub_pd(ix2,jx2);
 795             dy22             = _mm256_sub_pd(iy2,jy2);
 796             dz22             = _mm256_sub_pd(iz2,jz2);
 797             dx23             = _mm256_sub_pd(ix2,jx3);
 798             dy23             = _mm256_sub_pd(iy2,jy3);
 799             dz23             = _mm256_sub_pd(iz2,jz3);
 800             dx31             = _mm256_sub_pd(ix3,jx1);
 801             dy31             = _mm256_sub_pd(iy3,jy1);
 802             dz31             = _mm256_sub_pd(iz3,jz1);
 803             dx32             = _mm256_sub_pd(ix3,jx2);
 804             dy32             = _mm256_sub_pd(iy3,jy2);
 805             dz32             = _mm256_sub_pd(iz3,jz2);
 806             dx33             = _mm256_sub_pd(ix3,jx3);
 807             dy33             = _mm256_sub_pd(iy3,jy3);
 808             dz33             = _mm256_sub_pd(iz3,jz3);
 809
 810             /* Calculate squared distance and things based on it */
 811             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 812             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 813             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 814             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 815             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 816             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 817             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 818             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 819             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 820             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 821
 822             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 823             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 824             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 825             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 826             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 827             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 828             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 829             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 830             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 831
 832             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 833             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 834             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 835             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 836             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 837             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 838             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 839             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 840             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 841             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 842
 843             fjx0             = _mm256_setzero_pd();
 844             fjy0             = _mm256_setzero_pd();
 845             fjz0             = _mm256_setzero_pd();
 846             fjx1             = _mm256_setzero_pd();
 847             fjy1             = _mm256_setzero_pd();
 848             fjz1             = _mm256_setzero_pd();
 849             fjx2             = _mm256_setzero_pd();
 850             fjy2             = _mm256_setzero_pd();
 851             fjz2             = _mm256_setzero_pd();
 852             fjx3             = _mm256_setzero_pd();
 853             fjy3             = _mm256_setzero_pd();
 854             fjz3             = _mm256_setzero_pd();
 855
 856             /**************************
 857              * CALCULATE INTERACTIONS *
 858              **************************/
 859
 860             /* LENNARD-JONES DISPERSION/REPULSION */
 861
 862             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 863             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 864             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 865             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 866             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 867
 868             /* Update potential sum for this i atom from the interaction with this j atom. */
 869             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 870             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 871
 872             fscal            = fvdw;
 873
 874             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 875
 876             /* Calculate temporary vectorial force */
 877             tx               = _mm256_mul_pd(fscal,dx00);
 878             ty               = _mm256_mul_pd(fscal,dy00);
 879             tz               = _mm256_mul_pd(fscal,dz00);
 880
 881             /* Update vectorial force */
 882             fix0             = _mm256_add_pd(fix0,tx);
 883             fiy0             = _mm256_add_pd(fiy0,ty);
 884             fiz0             = _mm256_add_pd(fiz0,tz);
 885
 886             fjx0             = _mm256_add_pd(fjx0,tx);
 887             fjy0             = _mm256_add_pd(fjy0,ty);
 888             fjz0             = _mm256_add_pd(fjz0,tz);
 889
 890             /**************************
 891              * CALCULATE INTERACTIONS *
 892              **************************/
 893
 894             r11              = _mm256_mul_pd(rsq11,rinv11);
 895             r11              = _mm256_andnot_pd(dummy_mask,r11);
 896
 897             /* EWALD ELECTROSTATICS */
 898
 899             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 900             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 901             ewitab           = _mm256_cvttpd_epi32(ewrt);
 902             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 903             ewitab           = _mm_slli_epi32(ewitab,2);
 904             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 905             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 906             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 907             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 908             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 909             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 910             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 911             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
 912             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 913
 914             /* Update potential sum for this i atom from the interaction with this j atom. */
 915             velec            = _mm256_andnot_pd(dummy_mask,velec);
 916             velecsum         = _mm256_add_pd(velecsum,velec);
 917
 918             fscal            = felec;
 919
 920             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 921
 922             /* Calculate temporary vectorial force */
 923             tx               = _mm256_mul_pd(fscal,dx11);
 924             ty               = _mm256_mul_pd(fscal,dy11);
 925             tz               = _mm256_mul_pd(fscal,dz11);
 926
 927             /* Update vectorial force */
 928             fix1             = _mm256_add_pd(fix1,tx);
 929             fiy1             = _mm256_add_pd(fiy1,ty);
 930             fiz1             = _mm256_add_pd(fiz1,tz);
 931
 932             fjx1             = _mm256_add_pd(fjx1,tx);
 933             fjy1             = _mm256_add_pd(fjy1,ty);
 934             fjz1             = _mm256_add_pd(fjz1,tz);
 935
 936             /**************************
 937              * CALCULATE INTERACTIONS *
 938              **************************/
 939
 940             r12              = _mm256_mul_pd(rsq12,rinv12);
 941             r12              = _mm256_andnot_pd(dummy_mask,r12);
 942
 943             /* EWALD ELECTROSTATICS */
 944
 945             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 946             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 947             ewitab           = _mm256_cvttpd_epi32(ewrt);
 948             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 949             ewitab           = _mm_slli_epi32(ewitab,2);
 950             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 951             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 952             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 953             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 954             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 955             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 956             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 957             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
 958             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 959
 960             /* Update potential sum for this i atom from the interaction with this j atom. */
 961             velec            = _mm256_andnot_pd(dummy_mask,velec);
 962             velecsum         = _mm256_add_pd(velecsum,velec);
 963
 964             fscal            = felec;
 965
 966             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 967
 968             /* Calculate temporary vectorial force */
 969             tx               = _mm256_mul_pd(fscal,dx12);
 970             ty               = _mm256_mul_pd(fscal,dy12);
 971             tz               = _mm256_mul_pd(fscal,dz12);
 972
 973             /* Update vectorial force */
 974             fix1             = _mm256_add_pd(fix1,tx);
 975             fiy1             = _mm256_add_pd(fiy1,ty);
 976             fiz1             = _mm256_add_pd(fiz1,tz);
 977
 978             fjx2             = _mm256_add_pd(fjx2,tx);
 979             fjy2             = _mm256_add_pd(fjy2,ty);
 980             fjz2             = _mm256_add_pd(fjz2,tz);
 981
 982             /**************************
 983              * CALCULATE INTERACTIONS *
 984              **************************/
 985
 986             r13              = _mm256_mul_pd(rsq13,rinv13);
 987             r13              = _mm256_andnot_pd(dummy_mask,r13);
 988
 989             /* EWALD ELECTROSTATICS */
 990
 991             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 992             ewrt             = _mm256_mul_pd(r13,ewtabscale);
 993             ewitab           = _mm256_cvttpd_epi32(ewrt);
 994             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 995             ewitab           = _mm_slli_epi32(ewitab,2);
 996             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 997             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 998             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 999             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1000             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1001             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1002             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1003             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(rinv13,velec));
1004             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
1005
1006             /* Update potential sum for this i atom from the interaction with this j atom. */
1007             velec            = _mm256_andnot_pd(dummy_mask,velec);
1008             velecsum         = _mm256_add_pd(velecsum,velec);
1009
1010             fscal            = felec;
1011
1012             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm256_mul_pd(fscal,dx13);
1016             ty               = _mm256_mul_pd(fscal,dy13);
1017             tz               = _mm256_mul_pd(fscal,dz13);
1018
1019             /* Update vectorial force */
1020             fix1             = _mm256_add_pd(fix1,tx);
1021             fiy1             = _mm256_add_pd(fiy1,ty);
1022             fiz1             = _mm256_add_pd(fiz1,tz);
1023
1024             fjx3             = _mm256_add_pd(fjx3,tx);
1025             fjy3             = _mm256_add_pd(fjy3,ty);
1026             fjz3             = _mm256_add_pd(fjz3,tz);
1027
1028             /**************************
1029              * CALCULATE INTERACTIONS *
1030              **************************/
1031
1032             r21              = _mm256_mul_pd(rsq21,rinv21);
1033             r21              = _mm256_andnot_pd(dummy_mask,r21);
1034
1035             /* EWALD ELECTROSTATICS */
1036
1037             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1038             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1039             ewitab           = _mm256_cvttpd_epi32(ewrt);
1040             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1041             ewitab           = _mm_slli_epi32(ewitab,2);
1042             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1043             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1044             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1045             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1046             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1047             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1048             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1049             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
1050             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1051
1052             /* Update potential sum for this i atom from the interaction with this j atom. */
1053             velec            = _mm256_andnot_pd(dummy_mask,velec);
1054             velecsum         = _mm256_add_pd(velecsum,velec);
1055
1056             fscal            = felec;
1057
1058             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1059
1060             /* Calculate temporary vectorial force */
1061             tx               = _mm256_mul_pd(fscal,dx21);
1062             ty               = _mm256_mul_pd(fscal,dy21);
1063             tz               = _mm256_mul_pd(fscal,dz21);
1064
1065             /* Update vectorial force */
1066             fix2             = _mm256_add_pd(fix2,tx);
1067             fiy2             = _mm256_add_pd(fiy2,ty);
1068             fiz2             = _mm256_add_pd(fiz2,tz);
1069
1070             fjx1             = _mm256_add_pd(fjx1,tx);
1071             fjy1             = _mm256_add_pd(fjy1,ty);
1072             fjz1             = _mm256_add_pd(fjz1,tz);
1073
1074             /**************************
1075              * CALCULATE INTERACTIONS *
1076              **************************/
1077
1078             r22              = _mm256_mul_pd(rsq22,rinv22);
1079             r22              = _mm256_andnot_pd(dummy_mask,r22);
1080
1081             /* EWALD ELECTROSTATICS */
1082
1083             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1084             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1085             ewitab           = _mm256_cvttpd_epi32(ewrt);
1086             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1087             ewitab           = _mm_slli_epi32(ewitab,2);
1088             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1089             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1090             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1091             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1092             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1093             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1094             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1095             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
1096             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1097
1098             /* Update potential sum for this i atom from the interaction with this j atom. */
1099             velec            = _mm256_andnot_pd(dummy_mask,velec);
1100             velecsum         = _mm256_add_pd(velecsum,velec);
1101
1102             fscal            = felec;
1103
1104             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1105
1106             /* Calculate temporary vectorial force */
1107             tx               = _mm256_mul_pd(fscal,dx22);
1108             ty               = _mm256_mul_pd(fscal,dy22);
1109             tz               = _mm256_mul_pd(fscal,dz22);
1110
1111             /* Update vectorial force */
1112             fix2             = _mm256_add_pd(fix2,tx);
1113             fiy2             = _mm256_add_pd(fiy2,ty);
1114             fiz2             = _mm256_add_pd(fiz2,tz);
1115
1116             fjx2             = _mm256_add_pd(fjx2,tx);
1117             fjy2             = _mm256_add_pd(fjy2,ty);
1118             fjz2             = _mm256_add_pd(fjz2,tz);
1119
1120             /**************************
1121              * CALCULATE INTERACTIONS *
1122              **************************/
1123
1124             r23              = _mm256_mul_pd(rsq23,rinv23);
1125             r23              = _mm256_andnot_pd(dummy_mask,r23);
1126
1127             /* EWALD ELECTROSTATICS */
1128
1129             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1130             ewrt             = _mm256_mul_pd(r23,ewtabscale);
1131             ewitab           = _mm256_cvttpd_epi32(ewrt);
1132             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1133             ewitab           = _mm_slli_epi32(ewitab,2);
1134             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1135             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1136             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1137             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1138             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1139             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1140             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1141             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(rinv23,velec));
1142             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
1143
1144             /* Update potential sum for this i atom from the interaction with this j atom. */
1145             velec            = _mm256_andnot_pd(dummy_mask,velec);
1146             velecsum         = _mm256_add_pd(velecsum,velec);
1147
1148             fscal            = felec;
1149
1150             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1151
1152             /* Calculate temporary vectorial force */
1153             tx               = _mm256_mul_pd(fscal,dx23);
1154             ty               = _mm256_mul_pd(fscal,dy23);
1155             tz               = _mm256_mul_pd(fscal,dz23);
1156
1157             /* Update vectorial force */
1158             fix2             = _mm256_add_pd(fix2,tx);
1159             fiy2             = _mm256_add_pd(fiy2,ty);
1160             fiz2             = _mm256_add_pd(fiz2,tz);
1161
1162             fjx3             = _mm256_add_pd(fjx3,tx);
1163             fjy3             = _mm256_add_pd(fjy3,ty);
1164             fjz3             = _mm256_add_pd(fjz3,tz);
1165
1166             /**************************
1167              * CALCULATE INTERACTIONS *
1168              **************************/
1169
1170             r31              = _mm256_mul_pd(rsq31,rinv31);
1171             r31              = _mm256_andnot_pd(dummy_mask,r31);
1172
1173             /* EWALD ELECTROSTATICS */
1174
1175             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1176             ewrt             = _mm256_mul_pd(r31,ewtabscale);
1177             ewitab           = _mm256_cvttpd_epi32(ewrt);
1178             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1179             ewitab           = _mm_slli_epi32(ewitab,2);
1180             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1181             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1182             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1183             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1184             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1185             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1186             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1187             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(rinv31,velec));
1188             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
1189
1190             /* Update potential sum for this i atom from the interaction with this j atom. */
1191             velec            = _mm256_andnot_pd(dummy_mask,velec);
1192             velecsum         = _mm256_add_pd(velecsum,velec);
1193
1194             fscal            = felec;
1195
1196             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1197
1198             /* Calculate temporary vectorial force */
1199             tx               = _mm256_mul_pd(fscal,dx31);
1200             ty               = _mm256_mul_pd(fscal,dy31);
1201             tz               = _mm256_mul_pd(fscal,dz31);
1202
1203             /* Update vectorial force */
1204             fix3             = _mm256_add_pd(fix3,tx);
1205             fiy3             = _mm256_add_pd(fiy3,ty);
1206             fiz3             = _mm256_add_pd(fiz3,tz);
1207
1208             fjx1             = _mm256_add_pd(fjx1,tx);
1209             fjy1             = _mm256_add_pd(fjy1,ty);
1210             fjz1             = _mm256_add_pd(fjz1,tz);
1211
1212             /**************************
1213              * CALCULATE INTERACTIONS *
1214              **************************/
1215
1216             r32              = _mm256_mul_pd(rsq32,rinv32);
1217             r32              = _mm256_andnot_pd(dummy_mask,r32);
1218
1219             /* EWALD ELECTROSTATICS */
1220
1221             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1222             ewrt             = _mm256_mul_pd(r32,ewtabscale);
1223             ewitab           = _mm256_cvttpd_epi32(ewrt);
1224             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1225             ewitab           = _mm_slli_epi32(ewitab,2);
1226             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1227             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1228             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1229             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1230             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1231             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1232             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1233             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(rinv32,velec));
1234             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
1235
1236             /* Update potential sum for this i atom from the interaction with this j atom. */
1237             velec            = _mm256_andnot_pd(dummy_mask,velec);
1238             velecsum         = _mm256_add_pd(velecsum,velec);
1239
1240             fscal            = felec;
1241
1242             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1243
1244             /* Calculate temporary vectorial force */
1245             tx               = _mm256_mul_pd(fscal,dx32);
1246             ty               = _mm256_mul_pd(fscal,dy32);
1247             tz               = _mm256_mul_pd(fscal,dz32);
1248
1249             /* Update vectorial force */
1250             fix3             = _mm256_add_pd(fix3,tx);
1251             fiy3             = _mm256_add_pd(fiy3,ty);
1252             fiz3             = _mm256_add_pd(fiz3,tz);
1253
1254             fjx2             = _mm256_add_pd(fjx2,tx);
1255             fjy2             = _mm256_add_pd(fjy2,ty);
1256             fjz2             = _mm256_add_pd(fjz2,tz);
1257
1258             /**************************
1259              * CALCULATE INTERACTIONS *
1260              **************************/
1261
1262             r33              = _mm256_mul_pd(rsq33,rinv33);
1263             r33              = _mm256_andnot_pd(dummy_mask,r33);
1264
1265             /* EWALD ELECTROSTATICS */
1266
1267             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1268             ewrt             = _mm256_mul_pd(r33,ewtabscale);
1269             ewitab           = _mm256_cvttpd_epi32(ewrt);
1270             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1271             ewitab           = _mm_slli_epi32(ewitab,2);
1272             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1273             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1274             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1275             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1276             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1277             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1278             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1279             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(rinv33,velec));
1280             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
1281
1282             /* Update potential sum for this i atom from the interaction with this j atom. */
1283             velec            = _mm256_andnot_pd(dummy_mask,velec);
1284             velecsum         = _mm256_add_pd(velecsum,velec);
1285
1286             fscal            = felec;
1287
1288             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1289
1290             /* Calculate temporary vectorial force */
1291             tx               = _mm256_mul_pd(fscal,dx33);
1292             ty               = _mm256_mul_pd(fscal,dy33);
1293             tz               = _mm256_mul_pd(fscal,dz33);
1294
1295             /* Update vectorial force */
1296             fix3             = _mm256_add_pd(fix3,tx);
1297             fiy3             = _mm256_add_pd(fiy3,ty);
1298             fiz3             = _mm256_add_pd(fiz3,tz);
1299
1300             fjx3             = _mm256_add_pd(fjx3,tx);
1301             fjy3             = _mm256_add_pd(fjy3,ty);
1302             fjz3             = _mm256_add_pd(fjz3,tz);
1303
1304             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1305             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1306             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1307             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1308
1309             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1310                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1311                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1312
1313             /* Inner loop uses 413 flops */
1314         }
1315
1316         /* End of innermost loop */
1317
1318         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1319                                                  f+i_coord_offset,fshift+i_shift_offset);
1320
1321         ggid                        = gid[iidx];
1322         /* Update potential energies */
1323         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1324         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1325
1326         /* Increment number of inner iterations */
1327         inneriter                  += j_index_end - j_index_start;
1328
1329         /* Outer loop uses 26 flops */
1330     }
1331
1332     /* Increment number of outer iterations */
1333     outeriter        += nri;
1334
1335     /* Update outer/inner flops */
1336
1337     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*413);
1338 }
1339 /*
1340  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_double
1341  * Electrostatics interaction: Ewald
1342  * VdW interaction:            LennardJones
1343  * Geometry:                   Water4-Water4
1344  * Calculate force/pot:        Force
1345  */
1346 void
1347 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_double
1348                     (t_nblist                    * gmx_restrict       nlist,
1349                      rvec                        * gmx_restrict          xx,
1350                      rvec                        * gmx_restrict          ff,
1351                      t_forcerec                  * gmx_restrict          fr,
1352                      t_mdatoms                   * gmx_restrict     mdatoms,
1353                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1354                      t_nrnb                      * gmx_restrict        nrnb)
1355 {
1356     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1357      * just 0 for non-waters.
1358      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1359      * jnr indices corresponding to data put in the four positions in the SIMD register.
1360      */
1361     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1362     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1363     int              jnrA,jnrB,jnrC,jnrD;
1364     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1365     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1366     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1367     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1368     real             rcutoff_scalar;
1369     real             *shiftvec,*fshift,*x,*f;
1370     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1371     real             scratch[4*DIM];
1372     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1373     real *           vdwioffsetptr0;
1374     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1375     real *           vdwioffsetptr1;
1376     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1377     real *           vdwioffsetptr2;
1378     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1379     real *           vdwioffsetptr3;
1380     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1381     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1382     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1383     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1384     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1385     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1386     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1387     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1388     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1389     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1390     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1391     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1392     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1393     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1394     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1395     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1396     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1397     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1398     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1399     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1400     real             *charge;
1401     int              nvdwtype;
1402     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1403     int              *vdwtype;
1404     real             *vdwparam;
1405     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1406     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1407     __m128i          ewitab;
1408     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1409     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1410     real             *ewtab;
1411     __m256d          dummy_mask,cutoff_mask;
1412     __m128           tmpmask0,tmpmask1;
1413     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1414     __m256d          one     = _mm256_set1_pd(1.0);
1415     __m256d          two     = _mm256_set1_pd(2.0);
1416     x                = xx[0];
1417     f                = ff[0];
1418
1419     nri              = nlist->nri;
1420     iinr             = nlist->iinr;
1421     jindex           = nlist->jindex;
1422     jjnr             = nlist->jjnr;
1423     shiftidx         = nlist->shift;
1424     gid              = nlist->gid;
1425     shiftvec         = fr->shift_vec[0];
1426     fshift           = fr->fshift[0];
1427     facel            = _mm256_set1_pd(fr->epsfac);
1428     charge           = mdatoms->chargeA;
1429     nvdwtype         = fr->ntype;
1430     vdwparam         = fr->nbfp;
1431     vdwtype          = mdatoms->typeA;
1432
1433     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
1434     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1435     beta2            = _mm256_mul_pd(beta,beta);
1436     beta3            = _mm256_mul_pd(beta,beta2);
1437
1438     ewtab            = fr->ic->tabq_coul_F;
1439     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
1440     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1441
1442     /* Setup water-specific parameters */
1443     inr              = nlist->iinr[0];
1444     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1445     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1446     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1447     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1448
1449     jq1              = _mm256_set1_pd(charge[inr+1]);
1450     jq2              = _mm256_set1_pd(charge[inr+2]);
1451     jq3              = _mm256_set1_pd(charge[inr+3]);
1452     vdwjidx0A        = 2*vdwtype[inr+0];
1453     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1454     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1455     qq11             = _mm256_mul_pd(iq1,jq1);
1456     qq12             = _mm256_mul_pd(iq1,jq2);
1457     qq13             = _mm256_mul_pd(iq1,jq3);
1458     qq21             = _mm256_mul_pd(iq2,jq1);
1459     qq22             = _mm256_mul_pd(iq2,jq2);
1460     qq23             = _mm256_mul_pd(iq2,jq3);
1461     qq31             = _mm256_mul_pd(iq3,jq1);
1462     qq32             = _mm256_mul_pd(iq3,jq2);
1463     qq33             = _mm256_mul_pd(iq3,jq3);
1464
1465     /* Avoid stupid compiler warnings */
1466     jnrA = jnrB = jnrC = jnrD = 0;
1467     j_coord_offsetA = 0;
1468     j_coord_offsetB = 0;
1469     j_coord_offsetC = 0;
1470     j_coord_offsetD = 0;
1471
1472     outeriter        = 0;
1473     inneriter        = 0;
1474
1475     for(iidx=0;iidx<4*DIM;iidx++)
1476     {
1477         scratch[iidx] = 0.0;
1478     }
1479
1480     /* Start outer loop over neighborlists */
1481     for(iidx=0; iidx<nri; iidx++)
1482     {
1483         /* Load shift vector for this list */
1484         i_shift_offset   = DIM*shiftidx[iidx];
1485
1486         /* Load limits for loop over neighbors */
1487         j_index_start    = jindex[iidx];
1488         j_index_end      = jindex[iidx+1];
1489
1490         /* Get outer coordinate index */
1491         inr              = iinr[iidx];
1492         i_coord_offset   = DIM*inr;
1493
1494         /* Load i particle coords and add shift vector */
1495         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1496                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1497
1498         fix0             = _mm256_setzero_pd();
1499         fiy0             = _mm256_setzero_pd();
1500         fiz0             = _mm256_setzero_pd();
1501         fix1             = _mm256_setzero_pd();
1502         fiy1             = _mm256_setzero_pd();
1503         fiz1             = _mm256_setzero_pd();
1504         fix2             = _mm256_setzero_pd();
1505         fiy2             = _mm256_setzero_pd();
1506         fiz2             = _mm256_setzero_pd();
1507         fix3             = _mm256_setzero_pd();
1508         fiy3             = _mm256_setzero_pd();
1509         fiz3             = _mm256_setzero_pd();
1510
1511         /* Start inner kernel loop */
1512         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1513         {
1514
1515             /* Get j neighbor index, and coordinate index */
1516             jnrA             = jjnr[jidx];
1517             jnrB             = jjnr[jidx+1];
1518             jnrC             = jjnr[jidx+2];
1519             jnrD             = jjnr[jidx+3];
1520             j_coord_offsetA  = DIM*jnrA;
1521             j_coord_offsetB  = DIM*jnrB;
1522             j_coord_offsetC  = DIM*jnrC;
1523             j_coord_offsetD  = DIM*jnrD;
1524
1525             /* load j atom coordinates */
1526             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1527                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1528                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1529                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1530
1531             /* Calculate displacement vector */
1532             dx00             = _mm256_sub_pd(ix0,jx0);
1533             dy00             = _mm256_sub_pd(iy0,jy0);
1534             dz00             = _mm256_sub_pd(iz0,jz0);
1535             dx11             = _mm256_sub_pd(ix1,jx1);
1536             dy11             = _mm256_sub_pd(iy1,jy1);
1537             dz11             = _mm256_sub_pd(iz1,jz1);
1538             dx12             = _mm256_sub_pd(ix1,jx2);
1539             dy12             = _mm256_sub_pd(iy1,jy2);
1540             dz12             = _mm256_sub_pd(iz1,jz2);
1541             dx13             = _mm256_sub_pd(ix1,jx3);
1542             dy13             = _mm256_sub_pd(iy1,jy3);
1543             dz13             = _mm256_sub_pd(iz1,jz3);
1544             dx21             = _mm256_sub_pd(ix2,jx1);
1545             dy21             = _mm256_sub_pd(iy2,jy1);
1546             dz21             = _mm256_sub_pd(iz2,jz1);
1547             dx22             = _mm256_sub_pd(ix2,jx2);
1548             dy22             = _mm256_sub_pd(iy2,jy2);
1549             dz22             = _mm256_sub_pd(iz2,jz2);
1550             dx23             = _mm256_sub_pd(ix2,jx3);
1551             dy23             = _mm256_sub_pd(iy2,jy3);
1552             dz23             = _mm256_sub_pd(iz2,jz3);
1553             dx31             = _mm256_sub_pd(ix3,jx1);
1554             dy31             = _mm256_sub_pd(iy3,jy1);
1555             dz31             = _mm256_sub_pd(iz3,jz1);
1556             dx32             = _mm256_sub_pd(ix3,jx2);
1557             dy32             = _mm256_sub_pd(iy3,jy2);
1558             dz32             = _mm256_sub_pd(iz3,jz2);
1559             dx33             = _mm256_sub_pd(ix3,jx3);
1560             dy33             = _mm256_sub_pd(iy3,jy3);
1561             dz33             = _mm256_sub_pd(iz3,jz3);
1562
1563             /* Calculate squared distance and things based on it */
1564             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1565             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1566             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1567             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1568             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1569             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1570             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1571             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1572             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1573             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1574
1575             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1576             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1577             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1578             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1579             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1580             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1581             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1582             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1583             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1584
1585             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1586             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1587             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1588             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1589             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1590             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1591             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1592             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1593             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1594             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1595
1596             fjx0             = _mm256_setzero_pd();
1597             fjy0             = _mm256_setzero_pd();
1598             fjz0             = _mm256_setzero_pd();
1599             fjx1             = _mm256_setzero_pd();
1600             fjy1             = _mm256_setzero_pd();
1601             fjz1             = _mm256_setzero_pd();
1602             fjx2             = _mm256_setzero_pd();
1603             fjy2             = _mm256_setzero_pd();
1604             fjz2             = _mm256_setzero_pd();
1605             fjx3             = _mm256_setzero_pd();
1606             fjy3             = _mm256_setzero_pd();
1607             fjz3             = _mm256_setzero_pd();
1608
1609             /**************************
1610              * CALCULATE INTERACTIONS *
1611              **************************/
1612
1613             /* LENNARD-JONES DISPERSION/REPULSION */
1614
1615             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1616             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1617
1618             fscal            = fvdw;
1619
1620             /* Calculate temporary vectorial force */
1621             tx               = _mm256_mul_pd(fscal,dx00);
1622             ty               = _mm256_mul_pd(fscal,dy00);
1623             tz               = _mm256_mul_pd(fscal,dz00);
1624
1625             /* Update vectorial force */
1626             fix0             = _mm256_add_pd(fix0,tx);
1627             fiy0             = _mm256_add_pd(fiy0,ty);
1628             fiz0             = _mm256_add_pd(fiz0,tz);
1629
1630             fjx0             = _mm256_add_pd(fjx0,tx);
1631             fjy0             = _mm256_add_pd(fjy0,ty);
1632             fjz0             = _mm256_add_pd(fjz0,tz);
1633
1634             /**************************
1635              * CALCULATE INTERACTIONS *
1636              **************************/
1637
1638             r11              = _mm256_mul_pd(rsq11,rinv11);
1639
1640             /* EWALD ELECTROSTATICS */
1641
1642             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1643             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1644             ewitab           = _mm256_cvttpd_epi32(ewrt);
1645             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1646             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1647                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1648                                             &ewtabF,&ewtabFn);
1649             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1650             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1651
1652             fscal            = felec;
1653
1654             /* Calculate temporary vectorial force */
1655             tx               = _mm256_mul_pd(fscal,dx11);
1656             ty               = _mm256_mul_pd(fscal,dy11);
1657             tz               = _mm256_mul_pd(fscal,dz11);
1658
1659             /* Update vectorial force */
1660             fix1             = _mm256_add_pd(fix1,tx);
1661             fiy1             = _mm256_add_pd(fiy1,ty);
1662             fiz1             = _mm256_add_pd(fiz1,tz);
1663
1664             fjx1             = _mm256_add_pd(fjx1,tx);
1665             fjy1             = _mm256_add_pd(fjy1,ty);
1666             fjz1             = _mm256_add_pd(fjz1,tz);
1667
1668             /**************************
1669              * CALCULATE INTERACTIONS *
1670              **************************/
1671
1672             r12              = _mm256_mul_pd(rsq12,rinv12);
1673
1674             /* EWALD ELECTROSTATICS */
1675
1676             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1677             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1678             ewitab           = _mm256_cvttpd_epi32(ewrt);
1679             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1680             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1681                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1682                                             &ewtabF,&ewtabFn);
1683             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1684             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1685
1686             fscal            = felec;
1687
1688             /* Calculate temporary vectorial force */
1689             tx               = _mm256_mul_pd(fscal,dx12);
1690             ty               = _mm256_mul_pd(fscal,dy12);
1691             tz               = _mm256_mul_pd(fscal,dz12);
1692
1693             /* Update vectorial force */
1694             fix1             = _mm256_add_pd(fix1,tx);
1695             fiy1             = _mm256_add_pd(fiy1,ty);
1696             fiz1             = _mm256_add_pd(fiz1,tz);
1697
1698             fjx2             = _mm256_add_pd(fjx2,tx);
1699             fjy2             = _mm256_add_pd(fjy2,ty);
1700             fjz2             = _mm256_add_pd(fjz2,tz);
1701
1702             /**************************
1703              * CALCULATE INTERACTIONS *
1704              **************************/
1705
1706             r13              = _mm256_mul_pd(rsq13,rinv13);
1707
1708             /* EWALD ELECTROSTATICS */
1709
1710             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1711             ewrt             = _mm256_mul_pd(r13,ewtabscale);
1712             ewitab           = _mm256_cvttpd_epi32(ewrt);
1713             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1714             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1715                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1716                                             &ewtabF,&ewtabFn);
1717             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1718             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
1719
1720             fscal            = felec;
1721
1722             /* Calculate temporary vectorial force */
1723             tx               = _mm256_mul_pd(fscal,dx13);
1724             ty               = _mm256_mul_pd(fscal,dy13);
1725             tz               = _mm256_mul_pd(fscal,dz13);
1726
1727             /* Update vectorial force */
1728             fix1             = _mm256_add_pd(fix1,tx);
1729             fiy1             = _mm256_add_pd(fiy1,ty);
1730             fiz1             = _mm256_add_pd(fiz1,tz);
1731
1732             fjx3             = _mm256_add_pd(fjx3,tx);
1733             fjy3             = _mm256_add_pd(fjy3,ty);
1734             fjz3             = _mm256_add_pd(fjz3,tz);
1735
1736             /**************************
1737              * CALCULATE INTERACTIONS *
1738              **************************/
1739
1740             r21              = _mm256_mul_pd(rsq21,rinv21);
1741
1742             /* EWALD ELECTROSTATICS */
1743
1744             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1745             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1746             ewitab           = _mm256_cvttpd_epi32(ewrt);
1747             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1748             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1749                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1750                                             &ewtabF,&ewtabFn);
1751             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1752             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1753
1754             fscal            = felec;
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm256_mul_pd(fscal,dx21);
1758             ty               = _mm256_mul_pd(fscal,dy21);
1759             tz               = _mm256_mul_pd(fscal,dz21);
1760
1761             /* Update vectorial force */
1762             fix2             = _mm256_add_pd(fix2,tx);
1763             fiy2             = _mm256_add_pd(fiy2,ty);
1764             fiz2             = _mm256_add_pd(fiz2,tz);
1765
1766             fjx1             = _mm256_add_pd(fjx1,tx);
1767             fjy1             = _mm256_add_pd(fjy1,ty);
1768             fjz1             = _mm256_add_pd(fjz1,tz);
1769
1770             /**************************
1771              * CALCULATE INTERACTIONS *
1772              **************************/
1773
1774             r22              = _mm256_mul_pd(rsq22,rinv22);
1775
1776             /* EWALD ELECTROSTATICS */
1777
1778             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1779             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1780             ewitab           = _mm256_cvttpd_epi32(ewrt);
1781             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1782             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1783                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1784                                             &ewtabF,&ewtabFn);
1785             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1786             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1787
1788             fscal            = felec;
1789
1790             /* Calculate temporary vectorial force */
1791             tx               = _mm256_mul_pd(fscal,dx22);
1792             ty               = _mm256_mul_pd(fscal,dy22);
1793             tz               = _mm256_mul_pd(fscal,dz22);
1794
1795             /* Update vectorial force */
1796             fix2             = _mm256_add_pd(fix2,tx);
1797             fiy2             = _mm256_add_pd(fiy2,ty);
1798             fiz2             = _mm256_add_pd(fiz2,tz);
1799
1800             fjx2             = _mm256_add_pd(fjx2,tx);
1801             fjy2             = _mm256_add_pd(fjy2,ty);
1802             fjz2             = _mm256_add_pd(fjz2,tz);
1803
1804             /**************************
1805              * CALCULATE INTERACTIONS *
1806              **************************/
1807
1808             r23              = _mm256_mul_pd(rsq23,rinv23);
1809
1810             /* EWALD ELECTROSTATICS */
1811
1812             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1813             ewrt             = _mm256_mul_pd(r23,ewtabscale);
1814             ewitab           = _mm256_cvttpd_epi32(ewrt);
1815             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1816             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1817                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1818                                             &ewtabF,&ewtabFn);
1819             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1820             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
1821
1822             fscal            = felec;
1823
1824             /* Calculate temporary vectorial force */
1825             tx               = _mm256_mul_pd(fscal,dx23);
1826             ty               = _mm256_mul_pd(fscal,dy23);
1827             tz               = _mm256_mul_pd(fscal,dz23);
1828
1829             /* Update vectorial force */
1830             fix2             = _mm256_add_pd(fix2,tx);
1831             fiy2             = _mm256_add_pd(fiy2,ty);
1832             fiz2             = _mm256_add_pd(fiz2,tz);
1833
1834             fjx3             = _mm256_add_pd(fjx3,tx);
1835             fjy3             = _mm256_add_pd(fjy3,ty);
1836             fjz3             = _mm256_add_pd(fjz3,tz);
1837
1838             /**************************
1839              * CALCULATE INTERACTIONS *
1840              **************************/
1841
1842             r31              = _mm256_mul_pd(rsq31,rinv31);
1843
1844             /* EWALD ELECTROSTATICS */
1845
1846             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1847             ewrt             = _mm256_mul_pd(r31,ewtabscale);
1848             ewitab           = _mm256_cvttpd_epi32(ewrt);
1849             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1850             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1851                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1852                                             &ewtabF,&ewtabFn);
1853             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1854             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
1855
1856             fscal            = felec;
1857
1858             /* Calculate temporary vectorial force */
1859             tx               = _mm256_mul_pd(fscal,dx31);
1860             ty               = _mm256_mul_pd(fscal,dy31);
1861             tz               = _mm256_mul_pd(fscal,dz31);
1862
1863             /* Update vectorial force */
1864             fix3             = _mm256_add_pd(fix3,tx);
1865             fiy3             = _mm256_add_pd(fiy3,ty);
1866             fiz3             = _mm256_add_pd(fiz3,tz);
1867
1868             fjx1             = _mm256_add_pd(fjx1,tx);
1869             fjy1             = _mm256_add_pd(fjy1,ty);
1870             fjz1             = _mm256_add_pd(fjz1,tz);
1871
1872             /**************************
1873              * CALCULATE INTERACTIONS *
1874              **************************/
1875
1876             r32              = _mm256_mul_pd(rsq32,rinv32);
1877
1878             /* EWALD ELECTROSTATICS */
1879
1880             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1881             ewrt             = _mm256_mul_pd(r32,ewtabscale);
1882             ewitab           = _mm256_cvttpd_epi32(ewrt);
1883             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1884             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1885                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1886                                             &ewtabF,&ewtabFn);
1887             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1888             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
1889
1890             fscal            = felec;
1891
1892             /* Calculate temporary vectorial force */
1893             tx               = _mm256_mul_pd(fscal,dx32);
1894             ty               = _mm256_mul_pd(fscal,dy32);
1895             tz               = _mm256_mul_pd(fscal,dz32);
1896
1897             /* Update vectorial force */
1898             fix3             = _mm256_add_pd(fix3,tx);
1899             fiy3             = _mm256_add_pd(fiy3,ty);
1900             fiz3             = _mm256_add_pd(fiz3,tz);
1901
1902             fjx2             = _mm256_add_pd(fjx2,tx);
1903             fjy2             = _mm256_add_pd(fjy2,ty);
1904             fjz2             = _mm256_add_pd(fjz2,tz);
1905
1906             /**************************
1907              * CALCULATE INTERACTIONS *
1908              **************************/
1909
1910             r33              = _mm256_mul_pd(rsq33,rinv33);
1911
1912             /* EWALD ELECTROSTATICS */
1913
1914             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1915             ewrt             = _mm256_mul_pd(r33,ewtabscale);
1916             ewitab           = _mm256_cvttpd_epi32(ewrt);
1917             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1918             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1919                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1920                                             &ewtabF,&ewtabFn);
1921             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1922             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
1923
1924             fscal            = felec;
1925
1926             /* Calculate temporary vectorial force */
1927             tx               = _mm256_mul_pd(fscal,dx33);
1928             ty               = _mm256_mul_pd(fscal,dy33);
1929             tz               = _mm256_mul_pd(fscal,dz33);
1930
1931             /* Update vectorial force */
1932             fix3             = _mm256_add_pd(fix3,tx);
1933             fiy3             = _mm256_add_pd(fiy3,ty);
1934             fiz3             = _mm256_add_pd(fiz3,tz);
1935
1936             fjx3             = _mm256_add_pd(fjx3,tx);
1937             fjy3             = _mm256_add_pd(fjy3,ty);
1938             fjz3             = _mm256_add_pd(fjz3,tz);
1939
1940             fjptrA             = f+j_coord_offsetA;
1941             fjptrB             = f+j_coord_offsetB;
1942             fjptrC             = f+j_coord_offsetC;
1943             fjptrD             = f+j_coord_offsetD;
1944
1945             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1946                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1947                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1948
1949             /* Inner loop uses 354 flops */
1950         }
1951
1952         if(jidx<j_index_end)
1953         {
1954
1955             /* Get j neighbor index, and coordinate index */
1956             jnrlistA         = jjnr[jidx];
1957             jnrlistB         = jjnr[jidx+1];
1958             jnrlistC         = jjnr[jidx+2];
1959             jnrlistD         = jjnr[jidx+3];
1960             /* Sign of each element will be negative for non-real atoms.
1961              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1962              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1963              */
1964             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1965
1966             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1967             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1968             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1969
1970             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1971             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1972             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1973             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1974             j_coord_offsetA  = DIM*jnrA;
1975             j_coord_offsetB  = DIM*jnrB;
1976             j_coord_offsetC  = DIM*jnrC;
1977             j_coord_offsetD  = DIM*jnrD;
1978
1979             /* load j atom coordinates */
1980             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1981                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1982                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1983                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1984
1985             /* Calculate displacement vector */
1986             dx00             = _mm256_sub_pd(ix0,jx0);
1987             dy00             = _mm256_sub_pd(iy0,jy0);
1988             dz00             = _mm256_sub_pd(iz0,jz0);
1989             dx11             = _mm256_sub_pd(ix1,jx1);
1990             dy11             = _mm256_sub_pd(iy1,jy1);
1991             dz11             = _mm256_sub_pd(iz1,jz1);
1992             dx12             = _mm256_sub_pd(ix1,jx2);
1993             dy12             = _mm256_sub_pd(iy1,jy2);
1994             dz12             = _mm256_sub_pd(iz1,jz2);
1995             dx13             = _mm256_sub_pd(ix1,jx3);
1996             dy13             = _mm256_sub_pd(iy1,jy3);
1997             dz13             = _mm256_sub_pd(iz1,jz3);
1998             dx21             = _mm256_sub_pd(ix2,jx1);
1999             dy21             = _mm256_sub_pd(iy2,jy1);
2000             dz21             = _mm256_sub_pd(iz2,jz1);
2001             dx22             = _mm256_sub_pd(ix2,jx2);
2002             dy22             = _mm256_sub_pd(iy2,jy2);
2003             dz22             = _mm256_sub_pd(iz2,jz2);
2004             dx23             = _mm256_sub_pd(ix2,jx3);
2005             dy23             = _mm256_sub_pd(iy2,jy3);
2006             dz23             = _mm256_sub_pd(iz2,jz3);
2007             dx31             = _mm256_sub_pd(ix3,jx1);
2008             dy31             = _mm256_sub_pd(iy3,jy1);
2009             dz31             = _mm256_sub_pd(iz3,jz1);
2010             dx32             = _mm256_sub_pd(ix3,jx2);
2011             dy32             = _mm256_sub_pd(iy3,jy2);
2012             dz32             = _mm256_sub_pd(iz3,jz2);
2013             dx33             = _mm256_sub_pd(ix3,jx3);
2014             dy33             = _mm256_sub_pd(iy3,jy3);
2015             dz33             = _mm256_sub_pd(iz3,jz3);
2016
2017             /* Calculate squared distance and things based on it */
2018             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2019             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2020             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2021             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
2022             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2023             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2024             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
2025             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
2026             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
2027             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
2028
2029             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
2030             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
2031             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
2032             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
2033             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
2034             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
2035             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
2036             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
2037             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
2038
2039             rinvsq00         = gmx_mm256_inv_pd(rsq00);
2040             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
2041             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
2042             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
2043             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
2044             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
2045             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
2046             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
2047             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
2048             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
2049
2050             fjx0             = _mm256_setzero_pd();
2051             fjy0             = _mm256_setzero_pd();
2052             fjz0             = _mm256_setzero_pd();
2053             fjx1             = _mm256_setzero_pd();
2054             fjy1             = _mm256_setzero_pd();
2055             fjz1             = _mm256_setzero_pd();
2056             fjx2             = _mm256_setzero_pd();
2057             fjy2             = _mm256_setzero_pd();
2058             fjz2             = _mm256_setzero_pd();
2059             fjx3             = _mm256_setzero_pd();
2060             fjy3             = _mm256_setzero_pd();
2061             fjz3             = _mm256_setzero_pd();
2062
2063             /**************************
2064              * CALCULATE INTERACTIONS *
2065              **************************/
2066
2067             /* LENNARD-JONES DISPERSION/REPULSION */
2068
2069             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2070             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2071
2072             fscal            = fvdw;
2073
2074             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2075
2076             /* Calculate temporary vectorial force */
2077             tx               = _mm256_mul_pd(fscal,dx00);
2078             ty               = _mm256_mul_pd(fscal,dy00);
2079             tz               = _mm256_mul_pd(fscal,dz00);
2080
2081             /* Update vectorial force */
2082             fix0             = _mm256_add_pd(fix0,tx);
2083             fiy0             = _mm256_add_pd(fiy0,ty);
2084             fiz0             = _mm256_add_pd(fiz0,tz);
2085
2086             fjx0             = _mm256_add_pd(fjx0,tx);
2087             fjy0             = _mm256_add_pd(fjy0,ty);
2088             fjz0             = _mm256_add_pd(fjz0,tz);
2089
2090             /**************************
2091              * CALCULATE INTERACTIONS *
2092              **************************/
2093
2094             r11              = _mm256_mul_pd(rsq11,rinv11);
2095             r11              = _mm256_andnot_pd(dummy_mask,r11);
2096
2097             /* EWALD ELECTROSTATICS */
2098
2099             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2100             ewrt             = _mm256_mul_pd(r11,ewtabscale);
2101             ewitab           = _mm256_cvttpd_epi32(ewrt);
2102             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2103             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2104                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2105                                             &ewtabF,&ewtabFn);
2106             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2107             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2108
2109             fscal            = felec;
2110
2111             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2112
2113             /* Calculate temporary vectorial force */
2114             tx               = _mm256_mul_pd(fscal,dx11);
2115             ty               = _mm256_mul_pd(fscal,dy11);
2116             tz               = _mm256_mul_pd(fscal,dz11);
2117
2118             /* Update vectorial force */
2119             fix1             = _mm256_add_pd(fix1,tx);
2120             fiy1             = _mm256_add_pd(fiy1,ty);
2121             fiz1             = _mm256_add_pd(fiz1,tz);
2122
2123             fjx1             = _mm256_add_pd(fjx1,tx);
2124             fjy1             = _mm256_add_pd(fjy1,ty);
2125             fjz1             = _mm256_add_pd(fjz1,tz);
2126
2127             /**************************
2128              * CALCULATE INTERACTIONS *
2129              **************************/
2130
2131             r12              = _mm256_mul_pd(rsq12,rinv12);
2132             r12              = _mm256_andnot_pd(dummy_mask,r12);
2133
2134             /* EWALD ELECTROSTATICS */
2135
2136             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2137             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2138             ewitab           = _mm256_cvttpd_epi32(ewrt);
2139             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2140             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2141                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2142                                             &ewtabF,&ewtabFn);
2143             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2144             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2145
2146             fscal            = felec;
2147
2148             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2149
2150             /* Calculate temporary vectorial force */
2151             tx               = _mm256_mul_pd(fscal,dx12);
2152             ty               = _mm256_mul_pd(fscal,dy12);
2153             tz               = _mm256_mul_pd(fscal,dz12);
2154
2155             /* Update vectorial force */
2156             fix1             = _mm256_add_pd(fix1,tx);
2157             fiy1             = _mm256_add_pd(fiy1,ty);
2158             fiz1             = _mm256_add_pd(fiz1,tz);
2159
2160             fjx2             = _mm256_add_pd(fjx2,tx);
2161             fjy2             = _mm256_add_pd(fjy2,ty);
2162             fjz2             = _mm256_add_pd(fjz2,tz);
2163
2164             /**************************
2165              * CALCULATE INTERACTIONS *
2166              **************************/
2167
2168             r13              = _mm256_mul_pd(rsq13,rinv13);
2169             r13              = _mm256_andnot_pd(dummy_mask,r13);
2170
2171             /* EWALD ELECTROSTATICS */
2172
2173             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2174             ewrt             = _mm256_mul_pd(r13,ewtabscale);
2175             ewitab           = _mm256_cvttpd_epi32(ewrt);
2176             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2177             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2178                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2179                                             &ewtabF,&ewtabFn);
2180             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2181             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
2182
2183             fscal            = felec;
2184
2185             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2186
2187             /* Calculate temporary vectorial force */
2188             tx               = _mm256_mul_pd(fscal,dx13);
2189             ty               = _mm256_mul_pd(fscal,dy13);
2190             tz               = _mm256_mul_pd(fscal,dz13);
2191
2192             /* Update vectorial force */
2193             fix1             = _mm256_add_pd(fix1,tx);
2194             fiy1             = _mm256_add_pd(fiy1,ty);
2195             fiz1             = _mm256_add_pd(fiz1,tz);
2196
2197             fjx3             = _mm256_add_pd(fjx3,tx);
2198             fjy3             = _mm256_add_pd(fjy3,ty);
2199             fjz3             = _mm256_add_pd(fjz3,tz);
2200
2201             /**************************
2202              * CALCULATE INTERACTIONS *
2203              **************************/
2204
2205             r21              = _mm256_mul_pd(rsq21,rinv21);
2206             r21              = _mm256_andnot_pd(dummy_mask,r21);
2207
2208             /* EWALD ELECTROSTATICS */
2209
2210             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2211             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2212             ewitab           = _mm256_cvttpd_epi32(ewrt);
2213             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2214             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2215                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2216                                             &ewtabF,&ewtabFn);
2217             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2218             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2219
2220             fscal            = felec;
2221
2222             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2223
2224             /* Calculate temporary vectorial force */
2225             tx               = _mm256_mul_pd(fscal,dx21);
2226             ty               = _mm256_mul_pd(fscal,dy21);
2227             tz               = _mm256_mul_pd(fscal,dz21);
2228
2229             /* Update vectorial force */
2230             fix2             = _mm256_add_pd(fix2,tx);
2231             fiy2             = _mm256_add_pd(fiy2,ty);
2232             fiz2             = _mm256_add_pd(fiz2,tz);
2233
2234             fjx1             = _mm256_add_pd(fjx1,tx);
2235             fjy1             = _mm256_add_pd(fjy1,ty);
2236             fjz1             = _mm256_add_pd(fjz1,tz);
2237
2238             /**************************
2239              * CALCULATE INTERACTIONS *
2240              **************************/
2241
2242             r22              = _mm256_mul_pd(rsq22,rinv22);
2243             r22              = _mm256_andnot_pd(dummy_mask,r22);
2244
2245             /* EWALD ELECTROSTATICS */
2246
2247             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2248             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2249             ewitab           = _mm256_cvttpd_epi32(ewrt);
2250             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2251             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2252                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2253                                             &ewtabF,&ewtabFn);
2254             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2255             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2256
2257             fscal            = felec;
2258
2259             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2260
2261             /* Calculate temporary vectorial force */
2262             tx               = _mm256_mul_pd(fscal,dx22);
2263             ty               = _mm256_mul_pd(fscal,dy22);
2264             tz               = _mm256_mul_pd(fscal,dz22);
2265
2266             /* Update vectorial force */
2267             fix2             = _mm256_add_pd(fix2,tx);
2268             fiy2             = _mm256_add_pd(fiy2,ty);
2269             fiz2             = _mm256_add_pd(fiz2,tz);
2270
2271             fjx2             = _mm256_add_pd(fjx2,tx);
2272             fjy2             = _mm256_add_pd(fjy2,ty);
2273             fjz2             = _mm256_add_pd(fjz2,tz);
2274
2275             /**************************
2276              * CALCULATE INTERACTIONS *
2277              **************************/
2278
2279             r23              = _mm256_mul_pd(rsq23,rinv23);
2280             r23              = _mm256_andnot_pd(dummy_mask,r23);
2281
2282             /* EWALD ELECTROSTATICS */
2283
2284             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2285             ewrt             = _mm256_mul_pd(r23,ewtabscale);
2286             ewitab           = _mm256_cvttpd_epi32(ewrt);
2287             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2288             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2289                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2290                                             &ewtabF,&ewtabFn);
2291             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2292             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
2293
2294             fscal            = felec;
2295
2296             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2297
2298             /* Calculate temporary vectorial force */
2299             tx               = _mm256_mul_pd(fscal,dx23);
2300             ty               = _mm256_mul_pd(fscal,dy23);
2301             tz               = _mm256_mul_pd(fscal,dz23);
2302
2303             /* Update vectorial force */
2304             fix2             = _mm256_add_pd(fix2,tx);
2305             fiy2             = _mm256_add_pd(fiy2,ty);
2306             fiz2             = _mm256_add_pd(fiz2,tz);
2307
2308             fjx3             = _mm256_add_pd(fjx3,tx);
2309             fjy3             = _mm256_add_pd(fjy3,ty);
2310             fjz3             = _mm256_add_pd(fjz3,tz);
2311
2312             /**************************
2313              * CALCULATE INTERACTIONS *
2314              **************************/
2315
2316             r31              = _mm256_mul_pd(rsq31,rinv31);
2317             r31              = _mm256_andnot_pd(dummy_mask,r31);
2318
2319             /* EWALD ELECTROSTATICS */
2320
2321             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2322             ewrt             = _mm256_mul_pd(r31,ewtabscale);
2323             ewitab           = _mm256_cvttpd_epi32(ewrt);
2324             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2325             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2326                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2327                                             &ewtabF,&ewtabFn);
2328             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2329             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
2330
2331             fscal            = felec;
2332
2333             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2334
2335             /* Calculate temporary vectorial force */
2336             tx               = _mm256_mul_pd(fscal,dx31);
2337             ty               = _mm256_mul_pd(fscal,dy31);
2338             tz               = _mm256_mul_pd(fscal,dz31);
2339
2340             /* Update vectorial force */
2341             fix3             = _mm256_add_pd(fix3,tx);
2342             fiy3             = _mm256_add_pd(fiy3,ty);
2343             fiz3             = _mm256_add_pd(fiz3,tz);
2344
2345             fjx1             = _mm256_add_pd(fjx1,tx);
2346             fjy1             = _mm256_add_pd(fjy1,ty);
2347             fjz1             = _mm256_add_pd(fjz1,tz);
2348
2349             /**************************
2350              * CALCULATE INTERACTIONS *
2351              **************************/
2352
2353             r32              = _mm256_mul_pd(rsq32,rinv32);
2354             r32              = _mm256_andnot_pd(dummy_mask,r32);
2355
2356             /* EWALD ELECTROSTATICS */
2357
2358             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2359             ewrt             = _mm256_mul_pd(r32,ewtabscale);
2360             ewitab           = _mm256_cvttpd_epi32(ewrt);
2361             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2362             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2363                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2364                                             &ewtabF,&ewtabFn);
2365             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2366             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
2367
2368             fscal            = felec;
2369
2370             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2371
2372             /* Calculate temporary vectorial force */
2373             tx               = _mm256_mul_pd(fscal,dx32);
2374             ty               = _mm256_mul_pd(fscal,dy32);
2375             tz               = _mm256_mul_pd(fscal,dz32);
2376
2377             /* Update vectorial force */
2378             fix3             = _mm256_add_pd(fix3,tx);
2379             fiy3             = _mm256_add_pd(fiy3,ty);
2380             fiz3             = _mm256_add_pd(fiz3,tz);
2381
2382             fjx2             = _mm256_add_pd(fjx2,tx);
2383             fjy2             = _mm256_add_pd(fjy2,ty);
2384             fjz2             = _mm256_add_pd(fjz2,tz);
2385
2386             /**************************
2387              * CALCULATE INTERACTIONS *
2388              **************************/
2389
2390             r33              = _mm256_mul_pd(rsq33,rinv33);
2391             r33              = _mm256_andnot_pd(dummy_mask,r33);
2392
2393             /* EWALD ELECTROSTATICS */
2394
2395             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2396             ewrt             = _mm256_mul_pd(r33,ewtabscale);
2397             ewitab           = _mm256_cvttpd_epi32(ewrt);
2398             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2399             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2400                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2401                                             &ewtabF,&ewtabFn);
2402             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2403             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
2404
2405             fscal            = felec;
2406
2407             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2408
2409             /* Calculate temporary vectorial force */
2410             tx               = _mm256_mul_pd(fscal,dx33);
2411             ty               = _mm256_mul_pd(fscal,dy33);
2412             tz               = _mm256_mul_pd(fscal,dz33);
2413
2414             /* Update vectorial force */
2415             fix3             = _mm256_add_pd(fix3,tx);
2416             fiy3             = _mm256_add_pd(fiy3,ty);
2417             fiz3             = _mm256_add_pd(fiz3,tz);
2418
2419             fjx3             = _mm256_add_pd(fjx3,tx);
2420             fjy3             = _mm256_add_pd(fjy3,ty);
2421             fjz3             = _mm256_add_pd(fjz3,tz);
2422
2423             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2424             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2425             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2426             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2427
2428             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2429                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2430                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2431
2432             /* Inner loop uses 363 flops */
2433         }
2434
2435         /* End of innermost loop */
2436
2437         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2438                                                  f+i_coord_offset,fshift+i_shift_offset);
2439
2440         /* Increment number of inner iterations */
2441         inneriter                  += j_index_end - j_index_start;
2442
2443         /* Outer loop uses 24 flops */
2444     }
2445
2446     /* Increment number of outer iterations */
2447     outeriter        += nri;
2448
2449     /* Update outer/inner flops */
2450
2451     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*363);
2452 }