src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEw_VdwLJ_GeomW4W4_avx_256_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_double.h"
  34 #include "kernelutil_x86_avx_256_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_double
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  63     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  64     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  65     real             rcutoff_scalar;
  66     real             *shiftvec,*fshift,*x,*f;
  67     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  68     real             scratch[4*DIM];
  69     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  70     real *           vdwioffsetptr0;
  71     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  72     real *           vdwioffsetptr1;
  73     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     real *           vdwioffsetptr3;
  77     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  78     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  79     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  81     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  82     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  83     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  84     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  85     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  86     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  87     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  88     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  89     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  90     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  91     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  92     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  93     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  94     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  95     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  96     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     int              nvdwtype;
  99     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 100     int              *vdwtype;
 101     real             *vdwparam;
 102     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 103     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 104     __m128i          ewitab;
 105     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 106     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 107     real             *ewtab;
 108     __m256d          dummy_mask,cutoff_mask;
 109     __m128           tmpmask0,tmpmask1;
 110     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 111     __m256d          one     = _mm256_set1_pd(1.0);
 112     __m256d          two     = _mm256_set1_pd(2.0);
 113     x                = xx[0];
 114     f                = ff[0];
 115
 116     nri              = nlist->nri;
 117     iinr             = nlist->iinr;
 118     jindex           = nlist->jindex;
 119     jjnr             = nlist->jjnr;
 120     shiftidx         = nlist->shift;
 121     gid              = nlist->gid;
 122     shiftvec         = fr->shift_vec[0];
 123     fshift           = fr->fshift[0];
 124     facel            = _mm256_set1_pd(fr->epsfac);
 125     charge           = mdatoms->chargeA;
 126     nvdwtype         = fr->ntype;
 127     vdwparam         = fr->nbfp;
 128     vdwtype          = mdatoms->typeA;
 129
 130     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 131     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
 132     beta2            = _mm256_mul_pd(beta,beta);
 133     beta3            = _mm256_mul_pd(beta,beta2);
 134
 135     ewtab            = fr->ic->tabq_coul_FDV0;
 136     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 137     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 138
 139     /* Setup water-specific parameters */
 140     inr              = nlist->iinr[0];
 141     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 142     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 143     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 144     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 145
 146     jq1              = _mm256_set1_pd(charge[inr+1]);
 147     jq2              = _mm256_set1_pd(charge[inr+2]);
 148     jq3              = _mm256_set1_pd(charge[inr+3]);
 149     vdwjidx0A        = 2*vdwtype[inr+0];
 150     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 151     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 152     qq11             = _mm256_mul_pd(iq1,jq1);
 153     qq12             = _mm256_mul_pd(iq1,jq2);
 154     qq13             = _mm256_mul_pd(iq1,jq3);
 155     qq21             = _mm256_mul_pd(iq2,jq1);
 156     qq22             = _mm256_mul_pd(iq2,jq2);
 157     qq23             = _mm256_mul_pd(iq2,jq3);
 158     qq31             = _mm256_mul_pd(iq3,jq1);
 159     qq32             = _mm256_mul_pd(iq3,jq2);
 160     qq33             = _mm256_mul_pd(iq3,jq3);
 161
 162     /* Avoid stupid compiler warnings */
 163     jnrA = jnrB = jnrC = jnrD = 0;
 164     j_coord_offsetA = 0;
 165     j_coord_offsetB = 0;
 166     j_coord_offsetC = 0;
 167     j_coord_offsetD = 0;
 168
 169     outeriter        = 0;
 170     inneriter        = 0;
 171
 172     for(iidx=0;iidx<4*DIM;iidx++)
 173     {
 174         scratch[iidx] = 0.0;
 175     }
 176
 177     /* Start outer loop over neighborlists */
 178     for(iidx=0; iidx<nri; iidx++)
 179     {
 180         /* Load shift vector for this list */
 181         i_shift_offset   = DIM*shiftidx[iidx];
 182
 183         /* Load limits for loop over neighbors */
 184         j_index_start    = jindex[iidx];
 185         j_index_end      = jindex[iidx+1];
 186
 187         /* Get outer coordinate index */
 188         inr              = iinr[iidx];
 189         i_coord_offset   = DIM*inr;
 190
 191         /* Load i particle coords and add shift vector */
 192         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 193                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 194
 195         fix0             = _mm256_setzero_pd();
 196         fiy0             = _mm256_setzero_pd();
 197         fiz0             = _mm256_setzero_pd();
 198         fix1             = _mm256_setzero_pd();
 199         fiy1             = _mm256_setzero_pd();
 200         fiz1             = _mm256_setzero_pd();
 201         fix2             = _mm256_setzero_pd();
 202         fiy2             = _mm256_setzero_pd();
 203         fiz2             = _mm256_setzero_pd();
 204         fix3             = _mm256_setzero_pd();
 205         fiy3             = _mm256_setzero_pd();
 206         fiz3             = _mm256_setzero_pd();
 207
 208         /* Reset potential sums */
 209         velecsum         = _mm256_setzero_pd();
 210         vvdwsum          = _mm256_setzero_pd();
 211
 212         /* Start inner kernel loop */
 213         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 214         {
 215
 216             /* Get j neighbor index, and coordinate index */
 217             jnrA             = jjnr[jidx];
 218             jnrB             = jjnr[jidx+1];
 219             jnrC             = jjnr[jidx+2];
 220             jnrD             = jjnr[jidx+3];
 221             j_coord_offsetA  = DIM*jnrA;
 222             j_coord_offsetB  = DIM*jnrB;
 223             j_coord_offsetC  = DIM*jnrC;
 224             j_coord_offsetD  = DIM*jnrD;
 225
 226             /* load j atom coordinates */
 227             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 228                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 229                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 230                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 231
 232             /* Calculate displacement vector */
 233             dx00             = _mm256_sub_pd(ix0,jx0);
 234             dy00             = _mm256_sub_pd(iy0,jy0);
 235             dz00             = _mm256_sub_pd(iz0,jz0);
 236             dx11             = _mm256_sub_pd(ix1,jx1);
 237             dy11             = _mm256_sub_pd(iy1,jy1);
 238             dz11             = _mm256_sub_pd(iz1,jz1);
 239             dx12             = _mm256_sub_pd(ix1,jx2);
 240             dy12             = _mm256_sub_pd(iy1,jy2);
 241             dz12             = _mm256_sub_pd(iz1,jz2);
 242             dx13             = _mm256_sub_pd(ix1,jx3);
 243             dy13             = _mm256_sub_pd(iy1,jy3);
 244             dz13             = _mm256_sub_pd(iz1,jz3);
 245             dx21             = _mm256_sub_pd(ix2,jx1);
 246             dy21             = _mm256_sub_pd(iy2,jy1);
 247             dz21             = _mm256_sub_pd(iz2,jz1);
 248             dx22             = _mm256_sub_pd(ix2,jx2);
 249             dy22             = _mm256_sub_pd(iy2,jy2);
 250             dz22             = _mm256_sub_pd(iz2,jz2);
 251             dx23             = _mm256_sub_pd(ix2,jx3);
 252             dy23             = _mm256_sub_pd(iy2,jy3);
 253             dz23             = _mm256_sub_pd(iz2,jz3);
 254             dx31             = _mm256_sub_pd(ix3,jx1);
 255             dy31             = _mm256_sub_pd(iy3,jy1);
 256             dz31             = _mm256_sub_pd(iz3,jz1);
 257             dx32             = _mm256_sub_pd(ix3,jx2);
 258             dy32             = _mm256_sub_pd(iy3,jy2);
 259             dz32             = _mm256_sub_pd(iz3,jz2);
 260             dx33             = _mm256_sub_pd(ix3,jx3);
 261             dy33             = _mm256_sub_pd(iy3,jy3);
 262             dz33             = _mm256_sub_pd(iz3,jz3);
 263
 264             /* Calculate squared distance and things based on it */
 265             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 266             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 267             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 268             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 269             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 270             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 271             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 272             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 273             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 274             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 275
 276             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 277             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 278             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 279             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 280             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 281             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 282             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 283             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 284             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 285
 286             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 287             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 288             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 289             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 290             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 291             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 292             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 293             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 294             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 295             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 296
 297             fjx0             = _mm256_setzero_pd();
 298             fjy0             = _mm256_setzero_pd();
 299             fjz0             = _mm256_setzero_pd();
 300             fjx1             = _mm256_setzero_pd();
 301             fjy1             = _mm256_setzero_pd();
 302             fjz1             = _mm256_setzero_pd();
 303             fjx2             = _mm256_setzero_pd();
 304             fjy2             = _mm256_setzero_pd();
 305             fjz2             = _mm256_setzero_pd();
 306             fjx3             = _mm256_setzero_pd();
 307             fjy3             = _mm256_setzero_pd();
 308             fjz3             = _mm256_setzero_pd();
 309
 310             /**************************
 311              * CALCULATE INTERACTIONS *
 312              **************************/
 313
 314             /* LENNARD-JONES DISPERSION/REPULSION */
 315
 316             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 317             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 318             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 319             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 320             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 321
 322             /* Update potential sum for this i atom from the interaction with this j atom. */
 323             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 324
 325             fscal            = fvdw;
 326
 327             /* Calculate temporary vectorial force */
 328             tx               = _mm256_mul_pd(fscal,dx00);
 329             ty               = _mm256_mul_pd(fscal,dy00);
 330             tz               = _mm256_mul_pd(fscal,dz00);
 331
 332             /* Update vectorial force */
 333             fix0             = _mm256_add_pd(fix0,tx);
 334             fiy0             = _mm256_add_pd(fiy0,ty);
 335             fiz0             = _mm256_add_pd(fiz0,tz);
 336
 337             fjx0             = _mm256_add_pd(fjx0,tx);
 338             fjy0             = _mm256_add_pd(fjy0,ty);
 339             fjz0             = _mm256_add_pd(fjz0,tz);
 340
 341             /**************************
 342              * CALCULATE INTERACTIONS *
 343              **************************/
 344
 345             r11              = _mm256_mul_pd(rsq11,rinv11);
 346
 347             /* EWALD ELECTROSTATICS */
 348
 349             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 350             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 351             ewitab           = _mm256_cvttpd_epi32(ewrt);
 352             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 353             ewitab           = _mm_slli_epi32(ewitab,2);
 354             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 355             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 356             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 357             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 358             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 359             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 360             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 361             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
 362             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 363
 364             /* Update potential sum for this i atom from the interaction with this j atom. */
 365             velecsum         = _mm256_add_pd(velecsum,velec);
 366
 367             fscal            = felec;
 368
 369             /* Calculate temporary vectorial force */
 370             tx               = _mm256_mul_pd(fscal,dx11);
 371             ty               = _mm256_mul_pd(fscal,dy11);
 372             tz               = _mm256_mul_pd(fscal,dz11);
 373
 374             /* Update vectorial force */
 375             fix1             = _mm256_add_pd(fix1,tx);
 376             fiy1             = _mm256_add_pd(fiy1,ty);
 377             fiz1             = _mm256_add_pd(fiz1,tz);
 378
 379             fjx1             = _mm256_add_pd(fjx1,tx);
 380             fjy1             = _mm256_add_pd(fjy1,ty);
 381             fjz1             = _mm256_add_pd(fjz1,tz);
 382
 383             /**************************
 384              * CALCULATE INTERACTIONS *
 385              **************************/
 386
 387             r12              = _mm256_mul_pd(rsq12,rinv12);
 388
 389             /* EWALD ELECTROSTATICS */
 390
 391             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 392             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 393             ewitab           = _mm256_cvttpd_epi32(ewrt);
 394             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 395             ewitab           = _mm_slli_epi32(ewitab,2);
 396             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 397             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 398             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 399             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 400             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 401             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 402             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 403             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
 404             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm256_add_pd(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm256_mul_pd(fscal,dx12);
 413             ty               = _mm256_mul_pd(fscal,dy12);
 414             tz               = _mm256_mul_pd(fscal,dz12);
 415
 416             /* Update vectorial force */
 417             fix1             = _mm256_add_pd(fix1,tx);
 418             fiy1             = _mm256_add_pd(fiy1,ty);
 419             fiz1             = _mm256_add_pd(fiz1,tz);
 420
 421             fjx2             = _mm256_add_pd(fjx2,tx);
 422             fjy2             = _mm256_add_pd(fjy2,ty);
 423             fjz2             = _mm256_add_pd(fjz2,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             r13              = _mm256_mul_pd(rsq13,rinv13);
 430
 431             /* EWALD ELECTROSTATICS */
 432
 433             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 434             ewrt             = _mm256_mul_pd(r13,ewtabscale);
 435             ewitab           = _mm256_cvttpd_epi32(ewrt);
 436             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 437             ewitab           = _mm_slli_epi32(ewitab,2);
 438             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 439             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 440             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 441             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 442             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 443             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 444             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 445             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(rinv13,velec));
 446             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
 447
 448             /* Update potential sum for this i atom from the interaction with this j atom. */
 449             velecsum         = _mm256_add_pd(velecsum,velec);
 450
 451             fscal            = felec;
 452
 453             /* Calculate temporary vectorial force */
 454             tx               = _mm256_mul_pd(fscal,dx13);
 455             ty               = _mm256_mul_pd(fscal,dy13);
 456             tz               = _mm256_mul_pd(fscal,dz13);
 457
 458             /* Update vectorial force */
 459             fix1             = _mm256_add_pd(fix1,tx);
 460             fiy1             = _mm256_add_pd(fiy1,ty);
 461             fiz1             = _mm256_add_pd(fiz1,tz);
 462
 463             fjx3             = _mm256_add_pd(fjx3,tx);
 464             fjy3             = _mm256_add_pd(fjy3,ty);
 465             fjz3             = _mm256_add_pd(fjz3,tz);
 466
 467             /**************************
 468              * CALCULATE INTERACTIONS *
 469              **************************/
 470
 471             r21              = _mm256_mul_pd(rsq21,rinv21);
 472
 473             /* EWALD ELECTROSTATICS */
 474
 475             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 476             ewrt             = _mm256_mul_pd(r21,ewtabscale);
 477             ewitab           = _mm256_cvttpd_epi32(ewrt);
 478             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 479             ewitab           = _mm_slli_epi32(ewitab,2);
 480             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 481             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 482             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 483             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 484             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 485             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 486             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 487             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
 488             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velecsum         = _mm256_add_pd(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm256_mul_pd(fscal,dx21);
 497             ty               = _mm256_mul_pd(fscal,dy21);
 498             tz               = _mm256_mul_pd(fscal,dz21);
 499
 500             /* Update vectorial force */
 501             fix2             = _mm256_add_pd(fix2,tx);
 502             fiy2             = _mm256_add_pd(fiy2,ty);
 503             fiz2             = _mm256_add_pd(fiz2,tz);
 504
 505             fjx1             = _mm256_add_pd(fjx1,tx);
 506             fjy1             = _mm256_add_pd(fjy1,ty);
 507             fjz1             = _mm256_add_pd(fjz1,tz);
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             r22              = _mm256_mul_pd(rsq22,rinv22);
 514
 515             /* EWALD ELECTROSTATICS */
 516
 517             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 518             ewrt             = _mm256_mul_pd(r22,ewtabscale);
 519             ewitab           = _mm256_cvttpd_epi32(ewrt);
 520             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 521             ewitab           = _mm_slli_epi32(ewitab,2);
 522             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 523             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 524             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 525             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 526             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 527             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 528             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 529             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
 530             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
 531
 532             /* Update potential sum for this i atom from the interaction with this j atom. */
 533             velecsum         = _mm256_add_pd(velecsum,velec);
 534
 535             fscal            = felec;
 536
 537             /* Calculate temporary vectorial force */
 538             tx               = _mm256_mul_pd(fscal,dx22);
 539             ty               = _mm256_mul_pd(fscal,dy22);
 540             tz               = _mm256_mul_pd(fscal,dz22);
 541
 542             /* Update vectorial force */
 543             fix2             = _mm256_add_pd(fix2,tx);
 544             fiy2             = _mm256_add_pd(fiy2,ty);
 545             fiz2             = _mm256_add_pd(fiz2,tz);
 546
 547             fjx2             = _mm256_add_pd(fjx2,tx);
 548             fjy2             = _mm256_add_pd(fjy2,ty);
 549             fjz2             = _mm256_add_pd(fjz2,tz);
 550
 551             /**************************
 552              * CALCULATE INTERACTIONS *
 553              **************************/
 554
 555             r23              = _mm256_mul_pd(rsq23,rinv23);
 556
 557             /* EWALD ELECTROSTATICS */
 558
 559             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 560             ewrt             = _mm256_mul_pd(r23,ewtabscale);
 561             ewitab           = _mm256_cvttpd_epi32(ewrt);
 562             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 563             ewitab           = _mm_slli_epi32(ewitab,2);
 564             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 565             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 566             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 567             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 568             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 569             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 570             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 571             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(rinv23,velec));
 572             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
 573
 574             /* Update potential sum for this i atom from the interaction with this j atom. */
 575             velecsum         = _mm256_add_pd(velecsum,velec);
 576
 577             fscal            = felec;
 578
 579             /* Calculate temporary vectorial force */
 580             tx               = _mm256_mul_pd(fscal,dx23);
 581             ty               = _mm256_mul_pd(fscal,dy23);
 582             tz               = _mm256_mul_pd(fscal,dz23);
 583
 584             /* Update vectorial force */
 585             fix2             = _mm256_add_pd(fix2,tx);
 586             fiy2             = _mm256_add_pd(fiy2,ty);
 587             fiz2             = _mm256_add_pd(fiz2,tz);
 588
 589             fjx3             = _mm256_add_pd(fjx3,tx);
 590             fjy3             = _mm256_add_pd(fjy3,ty);
 591             fjz3             = _mm256_add_pd(fjz3,tz);
 592
 593             /**************************
 594              * CALCULATE INTERACTIONS *
 595              **************************/
 596
 597             r31              = _mm256_mul_pd(rsq31,rinv31);
 598
 599             /* EWALD ELECTROSTATICS */
 600
 601             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 602             ewrt             = _mm256_mul_pd(r31,ewtabscale);
 603             ewitab           = _mm256_cvttpd_epi32(ewrt);
 604             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 605             ewitab           = _mm_slli_epi32(ewitab,2);
 606             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 607             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 608             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 609             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 610             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 611             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 612             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 613             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(rinv31,velec));
 614             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
 615
 616             /* Update potential sum for this i atom from the interaction with this j atom. */
 617             velecsum         = _mm256_add_pd(velecsum,velec);
 618
 619             fscal            = felec;
 620
 621             /* Calculate temporary vectorial force */
 622             tx               = _mm256_mul_pd(fscal,dx31);
 623             ty               = _mm256_mul_pd(fscal,dy31);
 624             tz               = _mm256_mul_pd(fscal,dz31);
 625
 626             /* Update vectorial force */
 627             fix3             = _mm256_add_pd(fix3,tx);
 628             fiy3             = _mm256_add_pd(fiy3,ty);
 629             fiz3             = _mm256_add_pd(fiz3,tz);
 630
 631             fjx1             = _mm256_add_pd(fjx1,tx);
 632             fjy1             = _mm256_add_pd(fjy1,ty);
 633             fjz1             = _mm256_add_pd(fjz1,tz);
 634
 635             /**************************
 636              * CALCULATE INTERACTIONS *
 637              **************************/
 638
 639             r32              = _mm256_mul_pd(rsq32,rinv32);
 640
 641             /* EWALD ELECTROSTATICS */
 642
 643             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 644             ewrt             = _mm256_mul_pd(r32,ewtabscale);
 645             ewitab           = _mm256_cvttpd_epi32(ewrt);
 646             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 647             ewitab           = _mm_slli_epi32(ewitab,2);
 648             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 649             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 650             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 651             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 652             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 653             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 654             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 655             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(rinv32,velec));
 656             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
 657
 658             /* Update potential sum for this i atom from the interaction with this j atom. */
 659             velecsum         = _mm256_add_pd(velecsum,velec);
 660
 661             fscal            = felec;
 662
 663             /* Calculate temporary vectorial force */
 664             tx               = _mm256_mul_pd(fscal,dx32);
 665             ty               = _mm256_mul_pd(fscal,dy32);
 666             tz               = _mm256_mul_pd(fscal,dz32);
 667
 668             /* Update vectorial force */
 669             fix3             = _mm256_add_pd(fix3,tx);
 670             fiy3             = _mm256_add_pd(fiy3,ty);
 671             fiz3             = _mm256_add_pd(fiz3,tz);
 672
 673             fjx2             = _mm256_add_pd(fjx2,tx);
 674             fjy2             = _mm256_add_pd(fjy2,ty);
 675             fjz2             = _mm256_add_pd(fjz2,tz);
 676
 677             /**************************
 678              * CALCULATE INTERACTIONS *
 679              **************************/
 680
 681             r33              = _mm256_mul_pd(rsq33,rinv33);
 682
 683             /* EWALD ELECTROSTATICS */
 684
 685             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 686             ewrt             = _mm256_mul_pd(r33,ewtabscale);
 687             ewitab           = _mm256_cvttpd_epi32(ewrt);
 688             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 689             ewitab           = _mm_slli_epi32(ewitab,2);
 690             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 691             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 692             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 693             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 694             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 695             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 696             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 697             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(rinv33,velec));
 698             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
 699
 700             /* Update potential sum for this i atom from the interaction with this j atom. */
 701             velecsum         = _mm256_add_pd(velecsum,velec);
 702
 703             fscal            = felec;
 704
 705             /* Calculate temporary vectorial force */
 706             tx               = _mm256_mul_pd(fscal,dx33);
 707             ty               = _mm256_mul_pd(fscal,dy33);
 708             tz               = _mm256_mul_pd(fscal,dz33);
 709
 710             /* Update vectorial force */
 711             fix3             = _mm256_add_pd(fix3,tx);
 712             fiy3             = _mm256_add_pd(fiy3,ty);
 713             fiz3             = _mm256_add_pd(fiz3,tz);
 714
 715             fjx3             = _mm256_add_pd(fjx3,tx);
 716             fjy3             = _mm256_add_pd(fjy3,ty);
 717             fjz3             = _mm256_add_pd(fjz3,tz);
 718
 719             fjptrA             = f+j_coord_offsetA;
 720             fjptrB             = f+j_coord_offsetB;
 721             fjptrC             = f+j_coord_offsetC;
 722             fjptrD             = f+j_coord_offsetD;
 723
 724             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 725                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 726                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 727
 728             /* Inner loop uses 404 flops */
 729         }
 730
 731         if(jidx<j_index_end)
 732         {
 733
 734             /* Get j neighbor index, and coordinate index */
 735             jnrlistA         = jjnr[jidx];
 736             jnrlistB         = jjnr[jidx+1];
 737             jnrlistC         = jjnr[jidx+2];
 738             jnrlistD         = jjnr[jidx+3];
 739             /* Sign of each element will be negative for non-real atoms.
 740              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 741              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 742              */
 743             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 744
 745             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 746             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 747             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 748
 749             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 750             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 751             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 752             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 753             j_coord_offsetA  = DIM*jnrA;
 754             j_coord_offsetB  = DIM*jnrB;
 755             j_coord_offsetC  = DIM*jnrC;
 756             j_coord_offsetD  = DIM*jnrD;
 757
 758             /* load j atom coordinates */
 759             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 760                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 761                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 762                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 763
 764             /* Calculate displacement vector */
 765             dx00             = _mm256_sub_pd(ix0,jx0);
 766             dy00             = _mm256_sub_pd(iy0,jy0);
 767             dz00             = _mm256_sub_pd(iz0,jz0);
 768             dx11             = _mm256_sub_pd(ix1,jx1);
 769             dy11             = _mm256_sub_pd(iy1,jy1);
 770             dz11             = _mm256_sub_pd(iz1,jz1);
 771             dx12             = _mm256_sub_pd(ix1,jx2);
 772             dy12             = _mm256_sub_pd(iy1,jy2);
 773             dz12             = _mm256_sub_pd(iz1,jz2);
 774             dx13             = _mm256_sub_pd(ix1,jx3);
 775             dy13             = _mm256_sub_pd(iy1,jy3);
 776             dz13             = _mm256_sub_pd(iz1,jz3);
 777             dx21             = _mm256_sub_pd(ix2,jx1);
 778             dy21             = _mm256_sub_pd(iy2,jy1);
 779             dz21             = _mm256_sub_pd(iz2,jz1);
 780             dx22             = _mm256_sub_pd(ix2,jx2);
 781             dy22             = _mm256_sub_pd(iy2,jy2);
 782             dz22             = _mm256_sub_pd(iz2,jz2);
 783             dx23             = _mm256_sub_pd(ix2,jx3);
 784             dy23             = _mm256_sub_pd(iy2,jy3);
 785             dz23             = _mm256_sub_pd(iz2,jz3);
 786             dx31             = _mm256_sub_pd(ix3,jx1);
 787             dy31             = _mm256_sub_pd(iy3,jy1);
 788             dz31             = _mm256_sub_pd(iz3,jz1);
 789             dx32             = _mm256_sub_pd(ix3,jx2);
 790             dy32             = _mm256_sub_pd(iy3,jy2);
 791             dz32             = _mm256_sub_pd(iz3,jz2);
 792             dx33             = _mm256_sub_pd(ix3,jx3);
 793             dy33             = _mm256_sub_pd(iy3,jy3);
 794             dz33             = _mm256_sub_pd(iz3,jz3);
 795
 796             /* Calculate squared distance and things based on it */
 797             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 798             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 799             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 800             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 801             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 802             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 803             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 804             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 805             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 806             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 807
 808             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 809             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 810             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 811             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 812             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 813             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 814             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 815             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 816             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 817
 818             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 819             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 820             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 821             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 822             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 823             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 824             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 825             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 826             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 827             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 828
 829             fjx0             = _mm256_setzero_pd();
 830             fjy0             = _mm256_setzero_pd();
 831             fjz0             = _mm256_setzero_pd();
 832             fjx1             = _mm256_setzero_pd();
 833             fjy1             = _mm256_setzero_pd();
 834             fjz1             = _mm256_setzero_pd();
 835             fjx2             = _mm256_setzero_pd();
 836             fjy2             = _mm256_setzero_pd();
 837             fjz2             = _mm256_setzero_pd();
 838             fjx3             = _mm256_setzero_pd();
 839             fjy3             = _mm256_setzero_pd();
 840             fjz3             = _mm256_setzero_pd();
 841
 842             /**************************
 843              * CALCULATE INTERACTIONS *
 844              **************************/
 845
 846             /* LENNARD-JONES DISPERSION/REPULSION */
 847
 848             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 849             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 850             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 851             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 852             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 853
 854             /* Update potential sum for this i atom from the interaction with this j atom. */
 855             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 856             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 857
 858             fscal            = fvdw;
 859
 860             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 861
 862             /* Calculate temporary vectorial force */
 863             tx               = _mm256_mul_pd(fscal,dx00);
 864             ty               = _mm256_mul_pd(fscal,dy00);
 865             tz               = _mm256_mul_pd(fscal,dz00);
 866
 867             /* Update vectorial force */
 868             fix0             = _mm256_add_pd(fix0,tx);
 869             fiy0             = _mm256_add_pd(fiy0,ty);
 870             fiz0             = _mm256_add_pd(fiz0,tz);
 871
 872             fjx0             = _mm256_add_pd(fjx0,tx);
 873             fjy0             = _mm256_add_pd(fjy0,ty);
 874             fjz0             = _mm256_add_pd(fjz0,tz);
 875
 876             /**************************
 877              * CALCULATE INTERACTIONS *
 878              **************************/
 879
 880             r11              = _mm256_mul_pd(rsq11,rinv11);
 881             r11              = _mm256_andnot_pd(dummy_mask,r11);
 882
 883             /* EWALD ELECTROSTATICS */
 884
 885             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 886             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 887             ewitab           = _mm256_cvttpd_epi32(ewrt);
 888             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 889             ewitab           = _mm_slli_epi32(ewitab,2);
 890             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 891             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 892             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 893             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 894             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 895             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 896             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 897             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
 898             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 899
 900             /* Update potential sum for this i atom from the interaction with this j atom. */
 901             velec            = _mm256_andnot_pd(dummy_mask,velec);
 902             velecsum         = _mm256_add_pd(velecsum,velec);
 903
 904             fscal            = felec;
 905
 906             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 907
 908             /* Calculate temporary vectorial force */
 909             tx               = _mm256_mul_pd(fscal,dx11);
 910             ty               = _mm256_mul_pd(fscal,dy11);
 911             tz               = _mm256_mul_pd(fscal,dz11);
 912
 913             /* Update vectorial force */
 914             fix1             = _mm256_add_pd(fix1,tx);
 915             fiy1             = _mm256_add_pd(fiy1,ty);
 916             fiz1             = _mm256_add_pd(fiz1,tz);
 917
 918             fjx1             = _mm256_add_pd(fjx1,tx);
 919             fjy1             = _mm256_add_pd(fjy1,ty);
 920             fjz1             = _mm256_add_pd(fjz1,tz);
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             r12              = _mm256_mul_pd(rsq12,rinv12);
 927             r12              = _mm256_andnot_pd(dummy_mask,r12);
 928
 929             /* EWALD ELECTROSTATICS */
 930
 931             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 932             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 933             ewitab           = _mm256_cvttpd_epi32(ewrt);
 934             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 935             ewitab           = _mm_slli_epi32(ewitab,2);
 936             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 937             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 938             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 939             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 940             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 941             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 942             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 943             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
 944             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm256_andnot_pd(dummy_mask,velec);
 948             velecsum         = _mm256_add_pd(velecsum,velec);
 949
 950             fscal            = felec;
 951
 952             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 953
 954             /* Calculate temporary vectorial force */
 955             tx               = _mm256_mul_pd(fscal,dx12);
 956             ty               = _mm256_mul_pd(fscal,dy12);
 957             tz               = _mm256_mul_pd(fscal,dz12);
 958
 959             /* Update vectorial force */
 960             fix1             = _mm256_add_pd(fix1,tx);
 961             fiy1             = _mm256_add_pd(fiy1,ty);
 962             fiz1             = _mm256_add_pd(fiz1,tz);
 963
 964             fjx2             = _mm256_add_pd(fjx2,tx);
 965             fjy2             = _mm256_add_pd(fjy2,ty);
 966             fjz2             = _mm256_add_pd(fjz2,tz);
 967
 968             /**************************
 969              * CALCULATE INTERACTIONS *
 970              **************************/
 971
 972             r13              = _mm256_mul_pd(rsq13,rinv13);
 973             r13              = _mm256_andnot_pd(dummy_mask,r13);
 974
 975             /* EWALD ELECTROSTATICS */
 976
 977             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 978             ewrt             = _mm256_mul_pd(r13,ewtabscale);
 979             ewitab           = _mm256_cvttpd_epi32(ewrt);
 980             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 981             ewitab           = _mm_slli_epi32(ewitab,2);
 982             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 983             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 984             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 985             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 986             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 987             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 988             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 989             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(rinv13,velec));
 990             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
 991
 992             /* Update potential sum for this i atom from the interaction with this j atom. */
 993             velec            = _mm256_andnot_pd(dummy_mask,velec);
 994             velecsum         = _mm256_add_pd(velecsum,velec);
 995
 996             fscal            = felec;
 997
 998             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 999
1000             /* Calculate temporary vectorial force */
1001             tx               = _mm256_mul_pd(fscal,dx13);
1002             ty               = _mm256_mul_pd(fscal,dy13);
1003             tz               = _mm256_mul_pd(fscal,dz13);
1004
1005             /* Update vectorial force */
1006             fix1             = _mm256_add_pd(fix1,tx);
1007             fiy1             = _mm256_add_pd(fiy1,ty);
1008             fiz1             = _mm256_add_pd(fiz1,tz);
1009
1010             fjx3             = _mm256_add_pd(fjx3,tx);
1011             fjy3             = _mm256_add_pd(fjy3,ty);
1012             fjz3             = _mm256_add_pd(fjz3,tz);
1013
1014             /**************************
1015              * CALCULATE INTERACTIONS *
1016              **************************/
1017
1018             r21              = _mm256_mul_pd(rsq21,rinv21);
1019             r21              = _mm256_andnot_pd(dummy_mask,r21);
1020
1021             /* EWALD ELECTROSTATICS */
1022
1023             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1024             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1025             ewitab           = _mm256_cvttpd_epi32(ewrt);
1026             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1027             ewitab           = _mm_slli_epi32(ewitab,2);
1028             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1029             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1030             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1031             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1032             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1033             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1034             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1035             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
1036             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1037
1038             /* Update potential sum for this i atom from the interaction with this j atom. */
1039             velec            = _mm256_andnot_pd(dummy_mask,velec);
1040             velecsum         = _mm256_add_pd(velecsum,velec);
1041
1042             fscal            = felec;
1043
1044             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1045
1046             /* Calculate temporary vectorial force */
1047             tx               = _mm256_mul_pd(fscal,dx21);
1048             ty               = _mm256_mul_pd(fscal,dy21);
1049             tz               = _mm256_mul_pd(fscal,dz21);
1050
1051             /* Update vectorial force */
1052             fix2             = _mm256_add_pd(fix2,tx);
1053             fiy2             = _mm256_add_pd(fiy2,ty);
1054             fiz2             = _mm256_add_pd(fiz2,tz);
1055
1056             fjx1             = _mm256_add_pd(fjx1,tx);
1057             fjy1             = _mm256_add_pd(fjy1,ty);
1058             fjz1             = _mm256_add_pd(fjz1,tz);
1059
1060             /**************************
1061              * CALCULATE INTERACTIONS *
1062              **************************/
1063
1064             r22              = _mm256_mul_pd(rsq22,rinv22);
1065             r22              = _mm256_andnot_pd(dummy_mask,r22);
1066
1067             /* EWALD ELECTROSTATICS */
1068
1069             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1070             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1071             ewitab           = _mm256_cvttpd_epi32(ewrt);
1072             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1073             ewitab           = _mm_slli_epi32(ewitab,2);
1074             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1075             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1076             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1077             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1078             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1079             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1080             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1081             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
1082             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1083
1084             /* Update potential sum for this i atom from the interaction with this j atom. */
1085             velec            = _mm256_andnot_pd(dummy_mask,velec);
1086             velecsum         = _mm256_add_pd(velecsum,velec);
1087
1088             fscal            = felec;
1089
1090             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1091
1092             /* Calculate temporary vectorial force */
1093             tx               = _mm256_mul_pd(fscal,dx22);
1094             ty               = _mm256_mul_pd(fscal,dy22);
1095             tz               = _mm256_mul_pd(fscal,dz22);
1096
1097             /* Update vectorial force */
1098             fix2             = _mm256_add_pd(fix2,tx);
1099             fiy2             = _mm256_add_pd(fiy2,ty);
1100             fiz2             = _mm256_add_pd(fiz2,tz);
1101
1102             fjx2             = _mm256_add_pd(fjx2,tx);
1103             fjy2             = _mm256_add_pd(fjy2,ty);
1104             fjz2             = _mm256_add_pd(fjz2,tz);
1105
1106             /**************************
1107              * CALCULATE INTERACTIONS *
1108              **************************/
1109
1110             r23              = _mm256_mul_pd(rsq23,rinv23);
1111             r23              = _mm256_andnot_pd(dummy_mask,r23);
1112
1113             /* EWALD ELECTROSTATICS */
1114
1115             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1116             ewrt             = _mm256_mul_pd(r23,ewtabscale);
1117             ewitab           = _mm256_cvttpd_epi32(ewrt);
1118             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1119             ewitab           = _mm_slli_epi32(ewitab,2);
1120             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1121             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1122             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1123             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1124             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1125             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1126             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1127             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(rinv23,velec));
1128             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
1129
1130             /* Update potential sum for this i atom from the interaction with this j atom. */
1131             velec            = _mm256_andnot_pd(dummy_mask,velec);
1132             velecsum         = _mm256_add_pd(velecsum,velec);
1133
1134             fscal            = felec;
1135
1136             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1137
1138             /* Calculate temporary vectorial force */
1139             tx               = _mm256_mul_pd(fscal,dx23);
1140             ty               = _mm256_mul_pd(fscal,dy23);
1141             tz               = _mm256_mul_pd(fscal,dz23);
1142
1143             /* Update vectorial force */
1144             fix2             = _mm256_add_pd(fix2,tx);
1145             fiy2             = _mm256_add_pd(fiy2,ty);
1146             fiz2             = _mm256_add_pd(fiz2,tz);
1147
1148             fjx3             = _mm256_add_pd(fjx3,tx);
1149             fjy3             = _mm256_add_pd(fjy3,ty);
1150             fjz3             = _mm256_add_pd(fjz3,tz);
1151
1152             /**************************
1153              * CALCULATE INTERACTIONS *
1154              **************************/
1155
1156             r31              = _mm256_mul_pd(rsq31,rinv31);
1157             r31              = _mm256_andnot_pd(dummy_mask,r31);
1158
1159             /* EWALD ELECTROSTATICS */
1160
1161             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1162             ewrt             = _mm256_mul_pd(r31,ewtabscale);
1163             ewitab           = _mm256_cvttpd_epi32(ewrt);
1164             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1165             ewitab           = _mm_slli_epi32(ewitab,2);
1166             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1167             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1168             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1169             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1170             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1171             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1172             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1173             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(rinv31,velec));
1174             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
1175
1176             /* Update potential sum for this i atom from the interaction with this j atom. */
1177             velec            = _mm256_andnot_pd(dummy_mask,velec);
1178             velecsum         = _mm256_add_pd(velecsum,velec);
1179
1180             fscal            = felec;
1181
1182             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1183
1184             /* Calculate temporary vectorial force */
1185             tx               = _mm256_mul_pd(fscal,dx31);
1186             ty               = _mm256_mul_pd(fscal,dy31);
1187             tz               = _mm256_mul_pd(fscal,dz31);
1188
1189             /* Update vectorial force */
1190             fix3             = _mm256_add_pd(fix3,tx);
1191             fiy3             = _mm256_add_pd(fiy3,ty);
1192             fiz3             = _mm256_add_pd(fiz3,tz);
1193
1194             fjx1             = _mm256_add_pd(fjx1,tx);
1195             fjy1             = _mm256_add_pd(fjy1,ty);
1196             fjz1             = _mm256_add_pd(fjz1,tz);
1197
1198             /**************************
1199              * CALCULATE INTERACTIONS *
1200              **************************/
1201
1202             r32              = _mm256_mul_pd(rsq32,rinv32);
1203             r32              = _mm256_andnot_pd(dummy_mask,r32);
1204
1205             /* EWALD ELECTROSTATICS */
1206
1207             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1208             ewrt             = _mm256_mul_pd(r32,ewtabscale);
1209             ewitab           = _mm256_cvttpd_epi32(ewrt);
1210             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1211             ewitab           = _mm_slli_epi32(ewitab,2);
1212             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1213             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1214             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1215             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1216             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1217             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1218             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1219             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(rinv32,velec));
1220             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
1221
1222             /* Update potential sum for this i atom from the interaction with this j atom. */
1223             velec            = _mm256_andnot_pd(dummy_mask,velec);
1224             velecsum         = _mm256_add_pd(velecsum,velec);
1225
1226             fscal            = felec;
1227
1228             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1229
1230             /* Calculate temporary vectorial force */
1231             tx               = _mm256_mul_pd(fscal,dx32);
1232             ty               = _mm256_mul_pd(fscal,dy32);
1233             tz               = _mm256_mul_pd(fscal,dz32);
1234
1235             /* Update vectorial force */
1236             fix3             = _mm256_add_pd(fix3,tx);
1237             fiy3             = _mm256_add_pd(fiy3,ty);
1238             fiz3             = _mm256_add_pd(fiz3,tz);
1239
1240             fjx2             = _mm256_add_pd(fjx2,tx);
1241             fjy2             = _mm256_add_pd(fjy2,ty);
1242             fjz2             = _mm256_add_pd(fjz2,tz);
1243
1244             /**************************
1245              * CALCULATE INTERACTIONS *
1246              **************************/
1247
1248             r33              = _mm256_mul_pd(rsq33,rinv33);
1249             r33              = _mm256_andnot_pd(dummy_mask,r33);
1250
1251             /* EWALD ELECTROSTATICS */
1252
1253             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1254             ewrt             = _mm256_mul_pd(r33,ewtabscale);
1255             ewitab           = _mm256_cvttpd_epi32(ewrt);
1256             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1257             ewitab           = _mm_slli_epi32(ewitab,2);
1258             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1259             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1260             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1261             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1262             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1263             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1264             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1265             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(rinv33,velec));
1266             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
1267
1268             /* Update potential sum for this i atom from the interaction with this j atom. */
1269             velec            = _mm256_andnot_pd(dummy_mask,velec);
1270             velecsum         = _mm256_add_pd(velecsum,velec);
1271
1272             fscal            = felec;
1273
1274             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1275
1276             /* Calculate temporary vectorial force */
1277             tx               = _mm256_mul_pd(fscal,dx33);
1278             ty               = _mm256_mul_pd(fscal,dy33);
1279             tz               = _mm256_mul_pd(fscal,dz33);
1280
1281             /* Update vectorial force */
1282             fix3             = _mm256_add_pd(fix3,tx);
1283             fiy3             = _mm256_add_pd(fiy3,ty);
1284             fiz3             = _mm256_add_pd(fiz3,tz);
1285
1286             fjx3             = _mm256_add_pd(fjx3,tx);
1287             fjy3             = _mm256_add_pd(fjy3,ty);
1288             fjz3             = _mm256_add_pd(fjz3,tz);
1289
1290             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1291             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1292             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1293             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1294
1295             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1296                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1297                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1298
1299             /* Inner loop uses 413 flops */
1300         }
1301
1302         /* End of innermost loop */
1303
1304         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1305                                                  f+i_coord_offset,fshift+i_shift_offset);
1306
1307         ggid                        = gid[iidx];
1308         /* Update potential energies */
1309         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1310         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1311
1312         /* Increment number of inner iterations */
1313         inneriter                  += j_index_end - j_index_start;
1314
1315         /* Outer loop uses 26 flops */
1316     }
1317
1318     /* Increment number of outer iterations */
1319     outeriter        += nri;
1320
1321     /* Update outer/inner flops */
1322
1323     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*413);
1324 }
1325 /*
1326  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_double
1327  * Electrostatics interaction: Ewald
1328  * VdW interaction:            LennardJones
1329  * Geometry:                   Water4-Water4
1330  * Calculate force/pot:        Force
1331  */
1332 void
1333 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_double
1334                     (t_nblist * gmx_restrict                nlist,
1335                      rvec * gmx_restrict                    xx,
1336                      rvec * gmx_restrict                    ff,
1337                      t_forcerec * gmx_restrict              fr,
1338                      t_mdatoms * gmx_restrict               mdatoms,
1339                      nb_kernel_data_t * gmx_restrict        kernel_data,
1340                      t_nrnb * gmx_restrict                  nrnb)
1341 {
1342     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1343      * just 0 for non-waters.
1344      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1345      * jnr indices corresponding to data put in the four positions in the SIMD register.
1346      */
1347     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1348     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1349     int              jnrA,jnrB,jnrC,jnrD;
1350     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1351     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1352     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1353     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1354     real             rcutoff_scalar;
1355     real             *shiftvec,*fshift,*x,*f;
1356     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1357     real             scratch[4*DIM];
1358     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1359     real *           vdwioffsetptr0;
1360     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1361     real *           vdwioffsetptr1;
1362     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1363     real *           vdwioffsetptr2;
1364     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1365     real *           vdwioffsetptr3;
1366     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1367     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1368     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1369     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1370     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1371     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1372     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1373     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1374     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1375     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1376     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1377     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1378     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1379     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1380     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1381     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1382     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1383     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1384     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1385     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1386     real             *charge;
1387     int              nvdwtype;
1388     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1389     int              *vdwtype;
1390     real             *vdwparam;
1391     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1392     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1393     __m128i          ewitab;
1394     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1395     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1396     real             *ewtab;
1397     __m256d          dummy_mask,cutoff_mask;
1398     __m128           tmpmask0,tmpmask1;
1399     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1400     __m256d          one     = _mm256_set1_pd(1.0);
1401     __m256d          two     = _mm256_set1_pd(2.0);
1402     x                = xx[0];
1403     f                = ff[0];
1404
1405     nri              = nlist->nri;
1406     iinr             = nlist->iinr;
1407     jindex           = nlist->jindex;
1408     jjnr             = nlist->jjnr;
1409     shiftidx         = nlist->shift;
1410     gid              = nlist->gid;
1411     shiftvec         = fr->shift_vec[0];
1412     fshift           = fr->fshift[0];
1413     facel            = _mm256_set1_pd(fr->epsfac);
1414     charge           = mdatoms->chargeA;
1415     nvdwtype         = fr->ntype;
1416     vdwparam         = fr->nbfp;
1417     vdwtype          = mdatoms->typeA;
1418
1419     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
1420     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff);
1421     beta2            = _mm256_mul_pd(beta,beta);
1422     beta3            = _mm256_mul_pd(beta,beta2);
1423
1424     ewtab            = fr->ic->tabq_coul_F;
1425     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
1426     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1427
1428     /* Setup water-specific parameters */
1429     inr              = nlist->iinr[0];
1430     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1431     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1432     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1433     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1434
1435     jq1              = _mm256_set1_pd(charge[inr+1]);
1436     jq2              = _mm256_set1_pd(charge[inr+2]);
1437     jq3              = _mm256_set1_pd(charge[inr+3]);
1438     vdwjidx0A        = 2*vdwtype[inr+0];
1439     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1440     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1441     qq11             = _mm256_mul_pd(iq1,jq1);
1442     qq12             = _mm256_mul_pd(iq1,jq2);
1443     qq13             = _mm256_mul_pd(iq1,jq3);
1444     qq21             = _mm256_mul_pd(iq2,jq1);
1445     qq22             = _mm256_mul_pd(iq2,jq2);
1446     qq23             = _mm256_mul_pd(iq2,jq3);
1447     qq31             = _mm256_mul_pd(iq3,jq1);
1448     qq32             = _mm256_mul_pd(iq3,jq2);
1449     qq33             = _mm256_mul_pd(iq3,jq3);
1450
1451     /* Avoid stupid compiler warnings */
1452     jnrA = jnrB = jnrC = jnrD = 0;
1453     j_coord_offsetA = 0;
1454     j_coord_offsetB = 0;
1455     j_coord_offsetC = 0;
1456     j_coord_offsetD = 0;
1457
1458     outeriter        = 0;
1459     inneriter        = 0;
1460
1461     for(iidx=0;iidx<4*DIM;iidx++)
1462     {
1463         scratch[iidx] = 0.0;
1464     }
1465
1466     /* Start outer loop over neighborlists */
1467     for(iidx=0; iidx<nri; iidx++)
1468     {
1469         /* Load shift vector for this list */
1470         i_shift_offset   = DIM*shiftidx[iidx];
1471
1472         /* Load limits for loop over neighbors */
1473         j_index_start    = jindex[iidx];
1474         j_index_end      = jindex[iidx+1];
1475
1476         /* Get outer coordinate index */
1477         inr              = iinr[iidx];
1478         i_coord_offset   = DIM*inr;
1479
1480         /* Load i particle coords and add shift vector */
1481         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1482                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1483
1484         fix0             = _mm256_setzero_pd();
1485         fiy0             = _mm256_setzero_pd();
1486         fiz0             = _mm256_setzero_pd();
1487         fix1             = _mm256_setzero_pd();
1488         fiy1             = _mm256_setzero_pd();
1489         fiz1             = _mm256_setzero_pd();
1490         fix2             = _mm256_setzero_pd();
1491         fiy2             = _mm256_setzero_pd();
1492         fiz2             = _mm256_setzero_pd();
1493         fix3             = _mm256_setzero_pd();
1494         fiy3             = _mm256_setzero_pd();
1495         fiz3             = _mm256_setzero_pd();
1496
1497         /* Start inner kernel loop */
1498         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1499         {
1500
1501             /* Get j neighbor index, and coordinate index */
1502             jnrA             = jjnr[jidx];
1503             jnrB             = jjnr[jidx+1];
1504             jnrC             = jjnr[jidx+2];
1505             jnrD             = jjnr[jidx+3];
1506             j_coord_offsetA  = DIM*jnrA;
1507             j_coord_offsetB  = DIM*jnrB;
1508             j_coord_offsetC  = DIM*jnrC;
1509             j_coord_offsetD  = DIM*jnrD;
1510
1511             /* load j atom coordinates */
1512             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1513                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1514                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1515                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1516
1517             /* Calculate displacement vector */
1518             dx00             = _mm256_sub_pd(ix0,jx0);
1519             dy00             = _mm256_sub_pd(iy0,jy0);
1520             dz00             = _mm256_sub_pd(iz0,jz0);
1521             dx11             = _mm256_sub_pd(ix1,jx1);
1522             dy11             = _mm256_sub_pd(iy1,jy1);
1523             dz11             = _mm256_sub_pd(iz1,jz1);
1524             dx12             = _mm256_sub_pd(ix1,jx2);
1525             dy12             = _mm256_sub_pd(iy1,jy2);
1526             dz12             = _mm256_sub_pd(iz1,jz2);
1527             dx13             = _mm256_sub_pd(ix1,jx3);
1528             dy13             = _mm256_sub_pd(iy1,jy3);
1529             dz13             = _mm256_sub_pd(iz1,jz3);
1530             dx21             = _mm256_sub_pd(ix2,jx1);
1531             dy21             = _mm256_sub_pd(iy2,jy1);
1532             dz21             = _mm256_sub_pd(iz2,jz1);
1533             dx22             = _mm256_sub_pd(ix2,jx2);
1534             dy22             = _mm256_sub_pd(iy2,jy2);
1535             dz22             = _mm256_sub_pd(iz2,jz2);
1536             dx23             = _mm256_sub_pd(ix2,jx3);
1537             dy23             = _mm256_sub_pd(iy2,jy3);
1538             dz23             = _mm256_sub_pd(iz2,jz3);
1539             dx31             = _mm256_sub_pd(ix3,jx1);
1540             dy31             = _mm256_sub_pd(iy3,jy1);
1541             dz31             = _mm256_sub_pd(iz3,jz1);
1542             dx32             = _mm256_sub_pd(ix3,jx2);
1543             dy32             = _mm256_sub_pd(iy3,jy2);
1544             dz32             = _mm256_sub_pd(iz3,jz2);
1545             dx33             = _mm256_sub_pd(ix3,jx3);
1546             dy33             = _mm256_sub_pd(iy3,jy3);
1547             dz33             = _mm256_sub_pd(iz3,jz3);
1548
1549             /* Calculate squared distance and things based on it */
1550             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1551             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1552             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1553             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1554             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1555             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1556             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1557             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1558             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1559             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1560
1561             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1562             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1563             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1564             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1565             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1566             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1567             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1568             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1569             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1570
1571             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1572             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1573             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1574             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1575             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1576             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1577             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1578             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1579             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1580             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1581
1582             fjx0             = _mm256_setzero_pd();
1583             fjy0             = _mm256_setzero_pd();
1584             fjz0             = _mm256_setzero_pd();
1585             fjx1             = _mm256_setzero_pd();
1586             fjy1             = _mm256_setzero_pd();
1587             fjz1             = _mm256_setzero_pd();
1588             fjx2             = _mm256_setzero_pd();
1589             fjy2             = _mm256_setzero_pd();
1590             fjz2             = _mm256_setzero_pd();
1591             fjx3             = _mm256_setzero_pd();
1592             fjy3             = _mm256_setzero_pd();
1593             fjz3             = _mm256_setzero_pd();
1594
1595             /**************************
1596              * CALCULATE INTERACTIONS *
1597              **************************/
1598
1599             /* LENNARD-JONES DISPERSION/REPULSION */
1600
1601             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1602             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1603
1604             fscal            = fvdw;
1605
1606             /* Calculate temporary vectorial force */
1607             tx               = _mm256_mul_pd(fscal,dx00);
1608             ty               = _mm256_mul_pd(fscal,dy00);
1609             tz               = _mm256_mul_pd(fscal,dz00);
1610
1611             /* Update vectorial force */
1612             fix0             = _mm256_add_pd(fix0,tx);
1613             fiy0             = _mm256_add_pd(fiy0,ty);
1614             fiz0             = _mm256_add_pd(fiz0,tz);
1615
1616             fjx0             = _mm256_add_pd(fjx0,tx);
1617             fjy0             = _mm256_add_pd(fjy0,ty);
1618             fjz0             = _mm256_add_pd(fjz0,tz);
1619
1620             /**************************
1621              * CALCULATE INTERACTIONS *
1622              **************************/
1623
1624             r11              = _mm256_mul_pd(rsq11,rinv11);
1625
1626             /* EWALD ELECTROSTATICS */
1627
1628             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1629             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1630             ewitab           = _mm256_cvttpd_epi32(ewrt);
1631             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1632             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1633                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1634                                             &ewtabF,&ewtabFn);
1635             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1636             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1637
1638             fscal            = felec;
1639
1640             /* Calculate temporary vectorial force */
1641             tx               = _mm256_mul_pd(fscal,dx11);
1642             ty               = _mm256_mul_pd(fscal,dy11);
1643             tz               = _mm256_mul_pd(fscal,dz11);
1644
1645             /* Update vectorial force */
1646             fix1             = _mm256_add_pd(fix1,tx);
1647             fiy1             = _mm256_add_pd(fiy1,ty);
1648             fiz1             = _mm256_add_pd(fiz1,tz);
1649
1650             fjx1             = _mm256_add_pd(fjx1,tx);
1651             fjy1             = _mm256_add_pd(fjy1,ty);
1652             fjz1             = _mm256_add_pd(fjz1,tz);
1653
1654             /**************************
1655              * CALCULATE INTERACTIONS *
1656              **************************/
1657
1658             r12              = _mm256_mul_pd(rsq12,rinv12);
1659
1660             /* EWALD ELECTROSTATICS */
1661
1662             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1663             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1664             ewitab           = _mm256_cvttpd_epi32(ewrt);
1665             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1666             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1667                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1668                                             &ewtabF,&ewtabFn);
1669             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1670             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1671
1672             fscal            = felec;
1673
1674             /* Calculate temporary vectorial force */
1675             tx               = _mm256_mul_pd(fscal,dx12);
1676             ty               = _mm256_mul_pd(fscal,dy12);
1677             tz               = _mm256_mul_pd(fscal,dz12);
1678
1679             /* Update vectorial force */
1680             fix1             = _mm256_add_pd(fix1,tx);
1681             fiy1             = _mm256_add_pd(fiy1,ty);
1682             fiz1             = _mm256_add_pd(fiz1,tz);
1683
1684             fjx2             = _mm256_add_pd(fjx2,tx);
1685             fjy2             = _mm256_add_pd(fjy2,ty);
1686             fjz2             = _mm256_add_pd(fjz2,tz);
1687
1688             /**************************
1689              * CALCULATE INTERACTIONS *
1690              **************************/
1691
1692             r13              = _mm256_mul_pd(rsq13,rinv13);
1693
1694             /* EWALD ELECTROSTATICS */
1695
1696             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1697             ewrt             = _mm256_mul_pd(r13,ewtabscale);
1698             ewitab           = _mm256_cvttpd_epi32(ewrt);
1699             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1700             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1701                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1702                                             &ewtabF,&ewtabFn);
1703             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1704             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
1705
1706             fscal            = felec;
1707
1708             /* Calculate temporary vectorial force */
1709             tx               = _mm256_mul_pd(fscal,dx13);
1710             ty               = _mm256_mul_pd(fscal,dy13);
1711             tz               = _mm256_mul_pd(fscal,dz13);
1712
1713             /* Update vectorial force */
1714             fix1             = _mm256_add_pd(fix1,tx);
1715             fiy1             = _mm256_add_pd(fiy1,ty);
1716             fiz1             = _mm256_add_pd(fiz1,tz);
1717
1718             fjx3             = _mm256_add_pd(fjx3,tx);
1719             fjy3             = _mm256_add_pd(fjy3,ty);
1720             fjz3             = _mm256_add_pd(fjz3,tz);
1721
1722             /**************************
1723              * CALCULATE INTERACTIONS *
1724              **************************/
1725
1726             r21              = _mm256_mul_pd(rsq21,rinv21);
1727
1728             /* EWALD ELECTROSTATICS */
1729
1730             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1731             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1732             ewitab           = _mm256_cvttpd_epi32(ewrt);
1733             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1734             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1735                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1736                                             &ewtabF,&ewtabFn);
1737             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1738             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1739
1740             fscal            = felec;
1741
1742             /* Calculate temporary vectorial force */
1743             tx               = _mm256_mul_pd(fscal,dx21);
1744             ty               = _mm256_mul_pd(fscal,dy21);
1745             tz               = _mm256_mul_pd(fscal,dz21);
1746
1747             /* Update vectorial force */
1748             fix2             = _mm256_add_pd(fix2,tx);
1749             fiy2             = _mm256_add_pd(fiy2,ty);
1750             fiz2             = _mm256_add_pd(fiz2,tz);
1751
1752             fjx1             = _mm256_add_pd(fjx1,tx);
1753             fjy1             = _mm256_add_pd(fjy1,ty);
1754             fjz1             = _mm256_add_pd(fjz1,tz);
1755
1756             /**************************
1757              * CALCULATE INTERACTIONS *
1758              **************************/
1759
1760             r22              = _mm256_mul_pd(rsq22,rinv22);
1761
1762             /* EWALD ELECTROSTATICS */
1763
1764             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1765             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1766             ewitab           = _mm256_cvttpd_epi32(ewrt);
1767             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1768             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1769                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1770                                             &ewtabF,&ewtabFn);
1771             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1772             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1773
1774             fscal            = felec;
1775
1776             /* Calculate temporary vectorial force */
1777             tx               = _mm256_mul_pd(fscal,dx22);
1778             ty               = _mm256_mul_pd(fscal,dy22);
1779             tz               = _mm256_mul_pd(fscal,dz22);
1780
1781             /* Update vectorial force */
1782             fix2             = _mm256_add_pd(fix2,tx);
1783             fiy2             = _mm256_add_pd(fiy2,ty);
1784             fiz2             = _mm256_add_pd(fiz2,tz);
1785
1786             fjx2             = _mm256_add_pd(fjx2,tx);
1787             fjy2             = _mm256_add_pd(fjy2,ty);
1788             fjz2             = _mm256_add_pd(fjz2,tz);
1789
1790             /**************************
1791              * CALCULATE INTERACTIONS *
1792              **************************/
1793
1794             r23              = _mm256_mul_pd(rsq23,rinv23);
1795
1796             /* EWALD ELECTROSTATICS */
1797
1798             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1799             ewrt             = _mm256_mul_pd(r23,ewtabscale);
1800             ewitab           = _mm256_cvttpd_epi32(ewrt);
1801             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1802             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1803                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1804                                             &ewtabF,&ewtabFn);
1805             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1806             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
1807
1808             fscal            = felec;
1809
1810             /* Calculate temporary vectorial force */
1811             tx               = _mm256_mul_pd(fscal,dx23);
1812             ty               = _mm256_mul_pd(fscal,dy23);
1813             tz               = _mm256_mul_pd(fscal,dz23);
1814
1815             /* Update vectorial force */
1816             fix2             = _mm256_add_pd(fix2,tx);
1817             fiy2             = _mm256_add_pd(fiy2,ty);
1818             fiz2             = _mm256_add_pd(fiz2,tz);
1819
1820             fjx3             = _mm256_add_pd(fjx3,tx);
1821             fjy3             = _mm256_add_pd(fjy3,ty);
1822             fjz3             = _mm256_add_pd(fjz3,tz);
1823
1824             /**************************
1825              * CALCULATE INTERACTIONS *
1826              **************************/
1827
1828             r31              = _mm256_mul_pd(rsq31,rinv31);
1829
1830             /* EWALD ELECTROSTATICS */
1831
1832             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1833             ewrt             = _mm256_mul_pd(r31,ewtabscale);
1834             ewitab           = _mm256_cvttpd_epi32(ewrt);
1835             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1836             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1837                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1838                                             &ewtabF,&ewtabFn);
1839             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1840             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
1841
1842             fscal            = felec;
1843
1844             /* Calculate temporary vectorial force */
1845             tx               = _mm256_mul_pd(fscal,dx31);
1846             ty               = _mm256_mul_pd(fscal,dy31);
1847             tz               = _mm256_mul_pd(fscal,dz31);
1848
1849             /* Update vectorial force */
1850             fix3             = _mm256_add_pd(fix3,tx);
1851             fiy3             = _mm256_add_pd(fiy3,ty);
1852             fiz3             = _mm256_add_pd(fiz3,tz);
1853
1854             fjx1             = _mm256_add_pd(fjx1,tx);
1855             fjy1             = _mm256_add_pd(fjy1,ty);
1856             fjz1             = _mm256_add_pd(fjz1,tz);
1857
1858             /**************************
1859              * CALCULATE INTERACTIONS *
1860              **************************/
1861
1862             r32              = _mm256_mul_pd(rsq32,rinv32);
1863
1864             /* EWALD ELECTROSTATICS */
1865
1866             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1867             ewrt             = _mm256_mul_pd(r32,ewtabscale);
1868             ewitab           = _mm256_cvttpd_epi32(ewrt);
1869             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1870             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1871                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1872                                             &ewtabF,&ewtabFn);
1873             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1874             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
1875
1876             fscal            = felec;
1877
1878             /* Calculate temporary vectorial force */
1879             tx               = _mm256_mul_pd(fscal,dx32);
1880             ty               = _mm256_mul_pd(fscal,dy32);
1881             tz               = _mm256_mul_pd(fscal,dz32);
1882
1883             /* Update vectorial force */
1884             fix3             = _mm256_add_pd(fix3,tx);
1885             fiy3             = _mm256_add_pd(fiy3,ty);
1886             fiz3             = _mm256_add_pd(fiz3,tz);
1887
1888             fjx2             = _mm256_add_pd(fjx2,tx);
1889             fjy2             = _mm256_add_pd(fjy2,ty);
1890             fjz2             = _mm256_add_pd(fjz2,tz);
1891
1892             /**************************
1893              * CALCULATE INTERACTIONS *
1894              **************************/
1895
1896             r33              = _mm256_mul_pd(rsq33,rinv33);
1897
1898             /* EWALD ELECTROSTATICS */
1899
1900             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1901             ewrt             = _mm256_mul_pd(r33,ewtabscale);
1902             ewitab           = _mm256_cvttpd_epi32(ewrt);
1903             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1904             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1905                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1906                                             &ewtabF,&ewtabFn);
1907             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1908             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
1909
1910             fscal            = felec;
1911
1912             /* Calculate temporary vectorial force */
1913             tx               = _mm256_mul_pd(fscal,dx33);
1914             ty               = _mm256_mul_pd(fscal,dy33);
1915             tz               = _mm256_mul_pd(fscal,dz33);
1916
1917             /* Update vectorial force */
1918             fix3             = _mm256_add_pd(fix3,tx);
1919             fiy3             = _mm256_add_pd(fiy3,ty);
1920             fiz3             = _mm256_add_pd(fiz3,tz);
1921
1922             fjx3             = _mm256_add_pd(fjx3,tx);
1923             fjy3             = _mm256_add_pd(fjy3,ty);
1924             fjz3             = _mm256_add_pd(fjz3,tz);
1925
1926             fjptrA             = f+j_coord_offsetA;
1927             fjptrB             = f+j_coord_offsetB;
1928             fjptrC             = f+j_coord_offsetC;
1929             fjptrD             = f+j_coord_offsetD;
1930
1931             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1932                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1933                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1934
1935             /* Inner loop uses 354 flops */
1936         }
1937
1938         if(jidx<j_index_end)
1939         {
1940
1941             /* Get j neighbor index, and coordinate index */
1942             jnrlistA         = jjnr[jidx];
1943             jnrlistB         = jjnr[jidx+1];
1944             jnrlistC         = jjnr[jidx+2];
1945             jnrlistD         = jjnr[jidx+3];
1946             /* Sign of each element will be negative for non-real atoms.
1947              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1948              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1949              */
1950             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1951
1952             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1953             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1954             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1955
1956             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1957             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1958             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1959             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1960             j_coord_offsetA  = DIM*jnrA;
1961             j_coord_offsetB  = DIM*jnrB;
1962             j_coord_offsetC  = DIM*jnrC;
1963             j_coord_offsetD  = DIM*jnrD;
1964
1965             /* load j atom coordinates */
1966             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1967                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1968                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1969                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1970
1971             /* Calculate displacement vector */
1972             dx00             = _mm256_sub_pd(ix0,jx0);
1973             dy00             = _mm256_sub_pd(iy0,jy0);
1974             dz00             = _mm256_sub_pd(iz0,jz0);
1975             dx11             = _mm256_sub_pd(ix1,jx1);
1976             dy11             = _mm256_sub_pd(iy1,jy1);
1977             dz11             = _mm256_sub_pd(iz1,jz1);
1978             dx12             = _mm256_sub_pd(ix1,jx2);
1979             dy12             = _mm256_sub_pd(iy1,jy2);
1980             dz12             = _mm256_sub_pd(iz1,jz2);
1981             dx13             = _mm256_sub_pd(ix1,jx3);
1982             dy13             = _mm256_sub_pd(iy1,jy3);
1983             dz13             = _mm256_sub_pd(iz1,jz3);
1984             dx21             = _mm256_sub_pd(ix2,jx1);
1985             dy21             = _mm256_sub_pd(iy2,jy1);
1986             dz21             = _mm256_sub_pd(iz2,jz1);
1987             dx22             = _mm256_sub_pd(ix2,jx2);
1988             dy22             = _mm256_sub_pd(iy2,jy2);
1989             dz22             = _mm256_sub_pd(iz2,jz2);
1990             dx23             = _mm256_sub_pd(ix2,jx3);
1991             dy23             = _mm256_sub_pd(iy2,jy3);
1992             dz23             = _mm256_sub_pd(iz2,jz3);
1993             dx31             = _mm256_sub_pd(ix3,jx1);
1994             dy31             = _mm256_sub_pd(iy3,jy1);
1995             dz31             = _mm256_sub_pd(iz3,jz1);
1996             dx32             = _mm256_sub_pd(ix3,jx2);
1997             dy32             = _mm256_sub_pd(iy3,jy2);
1998             dz32             = _mm256_sub_pd(iz3,jz2);
1999             dx33             = _mm256_sub_pd(ix3,jx3);
2000             dy33             = _mm256_sub_pd(iy3,jy3);
2001             dz33             = _mm256_sub_pd(iz3,jz3);
2002
2003             /* Calculate squared distance and things based on it */
2004             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2005             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2006             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2007             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
2008             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2009             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2010             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
2011             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
2012             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
2013             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
2014
2015             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
2016             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
2017             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
2018             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
2019             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
2020             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
2021             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
2022             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
2023             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
2024
2025             rinvsq00         = gmx_mm256_inv_pd(rsq00);
2026             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
2027             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
2028             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
2029             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
2030             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
2031             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
2032             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
2033             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
2034             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
2035
2036             fjx0             = _mm256_setzero_pd();
2037             fjy0             = _mm256_setzero_pd();
2038             fjz0             = _mm256_setzero_pd();
2039             fjx1             = _mm256_setzero_pd();
2040             fjy1             = _mm256_setzero_pd();
2041             fjz1             = _mm256_setzero_pd();
2042             fjx2             = _mm256_setzero_pd();
2043             fjy2             = _mm256_setzero_pd();
2044             fjz2             = _mm256_setzero_pd();
2045             fjx3             = _mm256_setzero_pd();
2046             fjy3             = _mm256_setzero_pd();
2047             fjz3             = _mm256_setzero_pd();
2048
2049             /**************************
2050              * CALCULATE INTERACTIONS *
2051              **************************/
2052
2053             /* LENNARD-JONES DISPERSION/REPULSION */
2054
2055             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2056             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2057
2058             fscal            = fvdw;
2059
2060             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2061
2062             /* Calculate temporary vectorial force */
2063             tx               = _mm256_mul_pd(fscal,dx00);
2064             ty               = _mm256_mul_pd(fscal,dy00);
2065             tz               = _mm256_mul_pd(fscal,dz00);
2066
2067             /* Update vectorial force */
2068             fix0             = _mm256_add_pd(fix0,tx);
2069             fiy0             = _mm256_add_pd(fiy0,ty);
2070             fiz0             = _mm256_add_pd(fiz0,tz);
2071
2072             fjx0             = _mm256_add_pd(fjx0,tx);
2073             fjy0             = _mm256_add_pd(fjy0,ty);
2074             fjz0             = _mm256_add_pd(fjz0,tz);
2075
2076             /**************************
2077              * CALCULATE INTERACTIONS *
2078              **************************/
2079
2080             r11              = _mm256_mul_pd(rsq11,rinv11);
2081             r11              = _mm256_andnot_pd(dummy_mask,r11);
2082
2083             /* EWALD ELECTROSTATICS */
2084
2085             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2086             ewrt             = _mm256_mul_pd(r11,ewtabscale);
2087             ewitab           = _mm256_cvttpd_epi32(ewrt);
2088             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2089             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2090                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2091                                             &ewtabF,&ewtabFn);
2092             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2093             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2094
2095             fscal            = felec;
2096
2097             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2098
2099             /* Calculate temporary vectorial force */
2100             tx               = _mm256_mul_pd(fscal,dx11);
2101             ty               = _mm256_mul_pd(fscal,dy11);
2102             tz               = _mm256_mul_pd(fscal,dz11);
2103
2104             /* Update vectorial force */
2105             fix1             = _mm256_add_pd(fix1,tx);
2106             fiy1             = _mm256_add_pd(fiy1,ty);
2107             fiz1             = _mm256_add_pd(fiz1,tz);
2108
2109             fjx1             = _mm256_add_pd(fjx1,tx);
2110             fjy1             = _mm256_add_pd(fjy1,ty);
2111             fjz1             = _mm256_add_pd(fjz1,tz);
2112
2113             /**************************
2114              * CALCULATE INTERACTIONS *
2115              **************************/
2116
2117             r12              = _mm256_mul_pd(rsq12,rinv12);
2118             r12              = _mm256_andnot_pd(dummy_mask,r12);
2119
2120             /* EWALD ELECTROSTATICS */
2121
2122             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2123             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2124             ewitab           = _mm256_cvttpd_epi32(ewrt);
2125             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2126             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2127                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2128                                             &ewtabF,&ewtabFn);
2129             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2130             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2131
2132             fscal            = felec;
2133
2134             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2135
2136             /* Calculate temporary vectorial force */
2137             tx               = _mm256_mul_pd(fscal,dx12);
2138             ty               = _mm256_mul_pd(fscal,dy12);
2139             tz               = _mm256_mul_pd(fscal,dz12);
2140
2141             /* Update vectorial force */
2142             fix1             = _mm256_add_pd(fix1,tx);
2143             fiy1             = _mm256_add_pd(fiy1,ty);
2144             fiz1             = _mm256_add_pd(fiz1,tz);
2145
2146             fjx2             = _mm256_add_pd(fjx2,tx);
2147             fjy2             = _mm256_add_pd(fjy2,ty);
2148             fjz2             = _mm256_add_pd(fjz2,tz);
2149
2150             /**************************
2151              * CALCULATE INTERACTIONS *
2152              **************************/
2153
2154             r13              = _mm256_mul_pd(rsq13,rinv13);
2155             r13              = _mm256_andnot_pd(dummy_mask,r13);
2156
2157             /* EWALD ELECTROSTATICS */
2158
2159             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2160             ewrt             = _mm256_mul_pd(r13,ewtabscale);
2161             ewitab           = _mm256_cvttpd_epi32(ewrt);
2162             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2163             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2164                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2165                                             &ewtabF,&ewtabFn);
2166             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2167             felec            = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
2168
2169             fscal            = felec;
2170
2171             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2172
2173             /* Calculate temporary vectorial force */
2174             tx               = _mm256_mul_pd(fscal,dx13);
2175             ty               = _mm256_mul_pd(fscal,dy13);
2176             tz               = _mm256_mul_pd(fscal,dz13);
2177
2178             /* Update vectorial force */
2179             fix1             = _mm256_add_pd(fix1,tx);
2180             fiy1             = _mm256_add_pd(fiy1,ty);
2181             fiz1             = _mm256_add_pd(fiz1,tz);
2182
2183             fjx3             = _mm256_add_pd(fjx3,tx);
2184             fjy3             = _mm256_add_pd(fjy3,ty);
2185             fjz3             = _mm256_add_pd(fjz3,tz);
2186
2187             /**************************
2188              * CALCULATE INTERACTIONS *
2189              **************************/
2190
2191             r21              = _mm256_mul_pd(rsq21,rinv21);
2192             r21              = _mm256_andnot_pd(dummy_mask,r21);
2193
2194             /* EWALD ELECTROSTATICS */
2195
2196             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2197             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2198             ewitab           = _mm256_cvttpd_epi32(ewrt);
2199             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2200             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2201                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2202                                             &ewtabF,&ewtabFn);
2203             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2204             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2205
2206             fscal            = felec;
2207
2208             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2209
2210             /* Calculate temporary vectorial force */
2211             tx               = _mm256_mul_pd(fscal,dx21);
2212             ty               = _mm256_mul_pd(fscal,dy21);
2213             tz               = _mm256_mul_pd(fscal,dz21);
2214
2215             /* Update vectorial force */
2216             fix2             = _mm256_add_pd(fix2,tx);
2217             fiy2             = _mm256_add_pd(fiy2,ty);
2218             fiz2             = _mm256_add_pd(fiz2,tz);
2219
2220             fjx1             = _mm256_add_pd(fjx1,tx);
2221             fjy1             = _mm256_add_pd(fjy1,ty);
2222             fjz1             = _mm256_add_pd(fjz1,tz);
2223
2224             /**************************
2225              * CALCULATE INTERACTIONS *
2226              **************************/
2227
2228             r22              = _mm256_mul_pd(rsq22,rinv22);
2229             r22              = _mm256_andnot_pd(dummy_mask,r22);
2230
2231             /* EWALD ELECTROSTATICS */
2232
2233             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2234             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2235             ewitab           = _mm256_cvttpd_epi32(ewrt);
2236             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2237             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2238                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2239                                             &ewtabF,&ewtabFn);
2240             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2241             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2242
2243             fscal            = felec;
2244
2245             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2246
2247             /* Calculate temporary vectorial force */
2248             tx               = _mm256_mul_pd(fscal,dx22);
2249             ty               = _mm256_mul_pd(fscal,dy22);
2250             tz               = _mm256_mul_pd(fscal,dz22);
2251
2252             /* Update vectorial force */
2253             fix2             = _mm256_add_pd(fix2,tx);
2254             fiy2             = _mm256_add_pd(fiy2,ty);
2255             fiz2             = _mm256_add_pd(fiz2,tz);
2256
2257             fjx2             = _mm256_add_pd(fjx2,tx);
2258             fjy2             = _mm256_add_pd(fjy2,ty);
2259             fjz2             = _mm256_add_pd(fjz2,tz);
2260
2261             /**************************
2262              * CALCULATE INTERACTIONS *
2263              **************************/
2264
2265             r23              = _mm256_mul_pd(rsq23,rinv23);
2266             r23              = _mm256_andnot_pd(dummy_mask,r23);
2267
2268             /* EWALD ELECTROSTATICS */
2269
2270             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2271             ewrt             = _mm256_mul_pd(r23,ewtabscale);
2272             ewitab           = _mm256_cvttpd_epi32(ewrt);
2273             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2274             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2275                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2276                                             &ewtabF,&ewtabFn);
2277             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2278             felec            = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
2279
2280             fscal            = felec;
2281
2282             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2283
2284             /* Calculate temporary vectorial force */
2285             tx               = _mm256_mul_pd(fscal,dx23);
2286             ty               = _mm256_mul_pd(fscal,dy23);
2287             tz               = _mm256_mul_pd(fscal,dz23);
2288
2289             /* Update vectorial force */
2290             fix2             = _mm256_add_pd(fix2,tx);
2291             fiy2             = _mm256_add_pd(fiy2,ty);
2292             fiz2             = _mm256_add_pd(fiz2,tz);
2293
2294             fjx3             = _mm256_add_pd(fjx3,tx);
2295             fjy3             = _mm256_add_pd(fjy3,ty);
2296             fjz3             = _mm256_add_pd(fjz3,tz);
2297
2298             /**************************
2299              * CALCULATE INTERACTIONS *
2300              **************************/
2301
2302             r31              = _mm256_mul_pd(rsq31,rinv31);
2303             r31              = _mm256_andnot_pd(dummy_mask,r31);
2304
2305             /* EWALD ELECTROSTATICS */
2306
2307             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2308             ewrt             = _mm256_mul_pd(r31,ewtabscale);
2309             ewitab           = _mm256_cvttpd_epi32(ewrt);
2310             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2311             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2312                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2313                                             &ewtabF,&ewtabFn);
2314             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2315             felec            = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
2316
2317             fscal            = felec;
2318
2319             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2320
2321             /* Calculate temporary vectorial force */
2322             tx               = _mm256_mul_pd(fscal,dx31);
2323             ty               = _mm256_mul_pd(fscal,dy31);
2324             tz               = _mm256_mul_pd(fscal,dz31);
2325
2326             /* Update vectorial force */
2327             fix3             = _mm256_add_pd(fix3,tx);
2328             fiy3             = _mm256_add_pd(fiy3,ty);
2329             fiz3             = _mm256_add_pd(fiz3,tz);
2330
2331             fjx1             = _mm256_add_pd(fjx1,tx);
2332             fjy1             = _mm256_add_pd(fjy1,ty);
2333             fjz1             = _mm256_add_pd(fjz1,tz);
2334
2335             /**************************
2336              * CALCULATE INTERACTIONS *
2337              **************************/
2338
2339             r32              = _mm256_mul_pd(rsq32,rinv32);
2340             r32              = _mm256_andnot_pd(dummy_mask,r32);
2341
2342             /* EWALD ELECTROSTATICS */
2343
2344             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2345             ewrt             = _mm256_mul_pd(r32,ewtabscale);
2346             ewitab           = _mm256_cvttpd_epi32(ewrt);
2347             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2348             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2349                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2350                                             &ewtabF,&ewtabFn);
2351             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2352             felec            = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
2353
2354             fscal            = felec;
2355
2356             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2357
2358             /* Calculate temporary vectorial force */
2359             tx               = _mm256_mul_pd(fscal,dx32);
2360             ty               = _mm256_mul_pd(fscal,dy32);
2361             tz               = _mm256_mul_pd(fscal,dz32);
2362
2363             /* Update vectorial force */
2364             fix3             = _mm256_add_pd(fix3,tx);
2365             fiy3             = _mm256_add_pd(fiy3,ty);
2366             fiz3             = _mm256_add_pd(fiz3,tz);
2367
2368             fjx2             = _mm256_add_pd(fjx2,tx);
2369             fjy2             = _mm256_add_pd(fjy2,ty);
2370             fjz2             = _mm256_add_pd(fjz2,tz);
2371
2372             /**************************
2373              * CALCULATE INTERACTIONS *
2374              **************************/
2375
2376             r33              = _mm256_mul_pd(rsq33,rinv33);
2377             r33              = _mm256_andnot_pd(dummy_mask,r33);
2378
2379             /* EWALD ELECTROSTATICS */
2380
2381             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2382             ewrt             = _mm256_mul_pd(r33,ewtabscale);
2383             ewitab           = _mm256_cvttpd_epi32(ewrt);
2384             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2385             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2386                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2387                                             &ewtabF,&ewtabFn);
2388             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2389             felec            = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
2390
2391             fscal            = felec;
2392
2393             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2394
2395             /* Calculate temporary vectorial force */
2396             tx               = _mm256_mul_pd(fscal,dx33);
2397             ty               = _mm256_mul_pd(fscal,dy33);
2398             tz               = _mm256_mul_pd(fscal,dz33);
2399
2400             /* Update vectorial force */
2401             fix3             = _mm256_add_pd(fix3,tx);
2402             fiy3             = _mm256_add_pd(fiy3,ty);
2403             fiz3             = _mm256_add_pd(fiz3,tz);
2404
2405             fjx3             = _mm256_add_pd(fjx3,tx);
2406             fjy3             = _mm256_add_pd(fjy3,ty);
2407             fjz3             = _mm256_add_pd(fjz3,tz);
2408
2409             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2410             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2411             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2412             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2413
2414             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2415                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2416                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2417
2418             /* Inner loop uses 363 flops */
2419         }
2420
2421         /* End of innermost loop */
2422
2423         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2424                                                  f+i_coord_offset,fshift+i_shift_offset);
2425
2426         /* Increment number of inner iterations */
2427         inneriter                  += j_index_end - j_index_start;
2428
2429         /* Outer loop uses 24 flops */
2430     }
2431
2432     /* Increment number of outer iterations */
2433     outeriter        += nri;
2434
2435     /* Update outer/inner flops */
2436
2437     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*363);
2438 }