src/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecCSTab_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_ElecCSTab_VdwLJ_GeomW4W4_VF_avx_256_double
  38  * Electrostatics interaction: CubicSplineTable
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCSTab_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          vfitab;
 105     __m128i          ifour       = _mm_set1_epi32(4);
 106     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 107     real             *vftab;
 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     vftab            = kernel_data->table_elec->data;
 131     vftabscale       = _mm256_set1_pd(kernel_data->table_elec->scale);
 132
 133     /* Setup water-specific parameters */
 134     inr              = nlist->iinr[0];
 135     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 136     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 137     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 138     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 139
 140     jq1              = _mm256_set1_pd(charge[inr+1]);
 141     jq2              = _mm256_set1_pd(charge[inr+2]);
 142     jq3              = _mm256_set1_pd(charge[inr+3]);
 143     vdwjidx0A        = 2*vdwtype[inr+0];
 144     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 145     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 146     qq11             = _mm256_mul_pd(iq1,jq1);
 147     qq12             = _mm256_mul_pd(iq1,jq2);
 148     qq13             = _mm256_mul_pd(iq1,jq3);
 149     qq21             = _mm256_mul_pd(iq2,jq1);
 150     qq22             = _mm256_mul_pd(iq2,jq2);
 151     qq23             = _mm256_mul_pd(iq2,jq3);
 152     qq31             = _mm256_mul_pd(iq3,jq1);
 153     qq32             = _mm256_mul_pd(iq3,jq2);
 154     qq33             = _mm256_mul_pd(iq3,jq3);
 155
 156     /* Avoid stupid compiler warnings */
 157     jnrA = jnrB = jnrC = jnrD = 0;
 158     j_coord_offsetA = 0;
 159     j_coord_offsetB = 0;
 160     j_coord_offsetC = 0;
 161     j_coord_offsetD = 0;
 162
 163     outeriter        = 0;
 164     inneriter        = 0;
 165
 166     for(iidx=0;iidx<4*DIM;iidx++)
 167     {
 168         scratch[iidx] = 0.0;
 169     }
 170
 171     /* Start outer loop over neighborlists */
 172     for(iidx=0; iidx<nri; iidx++)
 173     {
 174         /* Load shift vector for this list */
 175         i_shift_offset   = DIM*shiftidx[iidx];
 176
 177         /* Load limits for loop over neighbors */
 178         j_index_start    = jindex[iidx];
 179         j_index_end      = jindex[iidx+1];
 180
 181         /* Get outer coordinate index */
 182         inr              = iinr[iidx];
 183         i_coord_offset   = DIM*inr;
 184
 185         /* Load i particle coords and add shift vector */
 186         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 187                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 188
 189         fix0             = _mm256_setzero_pd();
 190         fiy0             = _mm256_setzero_pd();
 191         fiz0             = _mm256_setzero_pd();
 192         fix1             = _mm256_setzero_pd();
 193         fiy1             = _mm256_setzero_pd();
 194         fiz1             = _mm256_setzero_pd();
 195         fix2             = _mm256_setzero_pd();
 196         fiy2             = _mm256_setzero_pd();
 197         fiz2             = _mm256_setzero_pd();
 198         fix3             = _mm256_setzero_pd();
 199         fiy3             = _mm256_setzero_pd();
 200         fiz3             = _mm256_setzero_pd();
 201
 202         /* Reset potential sums */
 203         velecsum         = _mm256_setzero_pd();
 204         vvdwsum          = _mm256_setzero_pd();
 205
 206         /* Start inner kernel loop */
 207         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 208         {
 209
 210             /* Get j neighbor index, and coordinate index */
 211             jnrA             = jjnr[jidx];
 212             jnrB             = jjnr[jidx+1];
 213             jnrC             = jjnr[jidx+2];
 214             jnrD             = jjnr[jidx+3];
 215             j_coord_offsetA  = DIM*jnrA;
 216             j_coord_offsetB  = DIM*jnrB;
 217             j_coord_offsetC  = DIM*jnrC;
 218             j_coord_offsetD  = DIM*jnrD;
 219
 220             /* load j atom coordinates */
 221             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 222                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 223                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 224                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 225
 226             /* Calculate displacement vector */
 227             dx00             = _mm256_sub_pd(ix0,jx0);
 228             dy00             = _mm256_sub_pd(iy0,jy0);
 229             dz00             = _mm256_sub_pd(iz0,jz0);
 230             dx11             = _mm256_sub_pd(ix1,jx1);
 231             dy11             = _mm256_sub_pd(iy1,jy1);
 232             dz11             = _mm256_sub_pd(iz1,jz1);
 233             dx12             = _mm256_sub_pd(ix1,jx2);
 234             dy12             = _mm256_sub_pd(iy1,jy2);
 235             dz12             = _mm256_sub_pd(iz1,jz2);
 236             dx13             = _mm256_sub_pd(ix1,jx3);
 237             dy13             = _mm256_sub_pd(iy1,jy3);
 238             dz13             = _mm256_sub_pd(iz1,jz3);
 239             dx21             = _mm256_sub_pd(ix2,jx1);
 240             dy21             = _mm256_sub_pd(iy2,jy1);
 241             dz21             = _mm256_sub_pd(iz2,jz1);
 242             dx22             = _mm256_sub_pd(ix2,jx2);
 243             dy22             = _mm256_sub_pd(iy2,jy2);
 244             dz22             = _mm256_sub_pd(iz2,jz2);
 245             dx23             = _mm256_sub_pd(ix2,jx3);
 246             dy23             = _mm256_sub_pd(iy2,jy3);
 247             dz23             = _mm256_sub_pd(iz2,jz3);
 248             dx31             = _mm256_sub_pd(ix3,jx1);
 249             dy31             = _mm256_sub_pd(iy3,jy1);
 250             dz31             = _mm256_sub_pd(iz3,jz1);
 251             dx32             = _mm256_sub_pd(ix3,jx2);
 252             dy32             = _mm256_sub_pd(iy3,jy2);
 253             dz32             = _mm256_sub_pd(iz3,jz2);
 254             dx33             = _mm256_sub_pd(ix3,jx3);
 255             dy33             = _mm256_sub_pd(iy3,jy3);
 256             dz33             = _mm256_sub_pd(iz3,jz3);
 257
 258             /* Calculate squared distance and things based on it */
 259             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 260             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 261             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 262             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 263             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 264             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 265             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 266             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 267             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 268             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 269
 270             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 271             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 272             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 273             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 274             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 275             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 276             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 277             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 278             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 279
 280             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 281
 282             fjx0             = _mm256_setzero_pd();
 283             fjy0             = _mm256_setzero_pd();
 284             fjz0             = _mm256_setzero_pd();
 285             fjx1             = _mm256_setzero_pd();
 286             fjy1             = _mm256_setzero_pd();
 287             fjz1             = _mm256_setzero_pd();
 288             fjx2             = _mm256_setzero_pd();
 289             fjy2             = _mm256_setzero_pd();
 290             fjz2             = _mm256_setzero_pd();
 291             fjx3             = _mm256_setzero_pd();
 292             fjy3             = _mm256_setzero_pd();
 293             fjz3             = _mm256_setzero_pd();
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             /* LENNARD-JONES DISPERSION/REPULSION */
 300
 301             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 302             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 303             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 304             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 305             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 306
 307             /* Update potential sum for this i atom from the interaction with this j atom. */
 308             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 309
 310             fscal            = fvdw;
 311
 312             /* Calculate temporary vectorial force */
 313             tx               = _mm256_mul_pd(fscal,dx00);
 314             ty               = _mm256_mul_pd(fscal,dy00);
 315             tz               = _mm256_mul_pd(fscal,dz00);
 316
 317             /* Update vectorial force */
 318             fix0             = _mm256_add_pd(fix0,tx);
 319             fiy0             = _mm256_add_pd(fiy0,ty);
 320             fiz0             = _mm256_add_pd(fiz0,tz);
 321
 322             fjx0             = _mm256_add_pd(fjx0,tx);
 323             fjy0             = _mm256_add_pd(fjy0,ty);
 324             fjz0             = _mm256_add_pd(fjz0,tz);
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             r11              = _mm256_mul_pd(rsq11,rinv11);
 331
 332             /* Calculate table index by multiplying r with table scale and truncate to integer */
 333             rt               = _mm256_mul_pd(r11,vftabscale);
 334             vfitab           = _mm256_cvttpd_epi32(rt);
 335             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 336             vfitab           = _mm_slli_epi32(vfitab,2);
 337
 338             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 339             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 340             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 341             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 342             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 343             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 344             Heps             = _mm256_mul_pd(vfeps,H);
 345             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 346             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 347             velec            = _mm256_mul_pd(qq11,VV);
 348             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 349             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
 350
 351             /* Update potential sum for this i atom from the interaction with this j atom. */
 352             velecsum         = _mm256_add_pd(velecsum,velec);
 353
 354             fscal            = felec;
 355
 356             /* Calculate temporary vectorial force */
 357             tx               = _mm256_mul_pd(fscal,dx11);
 358             ty               = _mm256_mul_pd(fscal,dy11);
 359             tz               = _mm256_mul_pd(fscal,dz11);
 360
 361             /* Update vectorial force */
 362             fix1             = _mm256_add_pd(fix1,tx);
 363             fiy1             = _mm256_add_pd(fiy1,ty);
 364             fiz1             = _mm256_add_pd(fiz1,tz);
 365
 366             fjx1             = _mm256_add_pd(fjx1,tx);
 367             fjy1             = _mm256_add_pd(fjy1,ty);
 368             fjz1             = _mm256_add_pd(fjz1,tz);
 369
 370             /**************************
 371              * CALCULATE INTERACTIONS *
 372              **************************/
 373
 374             r12              = _mm256_mul_pd(rsq12,rinv12);
 375
 376             /* Calculate table index by multiplying r with table scale and truncate to integer */
 377             rt               = _mm256_mul_pd(r12,vftabscale);
 378             vfitab           = _mm256_cvttpd_epi32(rt);
 379             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 380             vfitab           = _mm_slli_epi32(vfitab,2);
 381
 382             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 383             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 384             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 385             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 386             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 387             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 388             Heps             = _mm256_mul_pd(vfeps,H);
 389             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 390             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 391             velec            = _mm256_mul_pd(qq12,VV);
 392             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 393             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
 394
 395             /* Update potential sum for this i atom from the interaction with this j atom. */
 396             velecsum         = _mm256_add_pd(velecsum,velec);
 397
 398             fscal            = felec;
 399
 400             /* Calculate temporary vectorial force */
 401             tx               = _mm256_mul_pd(fscal,dx12);
 402             ty               = _mm256_mul_pd(fscal,dy12);
 403             tz               = _mm256_mul_pd(fscal,dz12);
 404
 405             /* Update vectorial force */
 406             fix1             = _mm256_add_pd(fix1,tx);
 407             fiy1             = _mm256_add_pd(fiy1,ty);
 408             fiz1             = _mm256_add_pd(fiz1,tz);
 409
 410             fjx2             = _mm256_add_pd(fjx2,tx);
 411             fjy2             = _mm256_add_pd(fjy2,ty);
 412             fjz2             = _mm256_add_pd(fjz2,tz);
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             r13              = _mm256_mul_pd(rsq13,rinv13);
 419
 420             /* Calculate table index by multiplying r with table scale and truncate to integer */
 421             rt               = _mm256_mul_pd(r13,vftabscale);
 422             vfitab           = _mm256_cvttpd_epi32(rt);
 423             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 424             vfitab           = _mm_slli_epi32(vfitab,2);
 425
 426             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 427             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 428             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 429             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 430             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 431             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 432             Heps             = _mm256_mul_pd(vfeps,H);
 433             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 434             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 435             velec            = _mm256_mul_pd(qq13,VV);
 436             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 437             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
 438
 439             /* Update potential sum for this i atom from the interaction with this j atom. */
 440             velecsum         = _mm256_add_pd(velecsum,velec);
 441
 442             fscal            = felec;
 443
 444             /* Calculate temporary vectorial force */
 445             tx               = _mm256_mul_pd(fscal,dx13);
 446             ty               = _mm256_mul_pd(fscal,dy13);
 447             tz               = _mm256_mul_pd(fscal,dz13);
 448
 449             /* Update vectorial force */
 450             fix1             = _mm256_add_pd(fix1,tx);
 451             fiy1             = _mm256_add_pd(fiy1,ty);
 452             fiz1             = _mm256_add_pd(fiz1,tz);
 453
 454             fjx3             = _mm256_add_pd(fjx3,tx);
 455             fjy3             = _mm256_add_pd(fjy3,ty);
 456             fjz3             = _mm256_add_pd(fjz3,tz);
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             r21              = _mm256_mul_pd(rsq21,rinv21);
 463
 464             /* Calculate table index by multiplying r with table scale and truncate to integer */
 465             rt               = _mm256_mul_pd(r21,vftabscale);
 466             vfitab           = _mm256_cvttpd_epi32(rt);
 467             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 468             vfitab           = _mm_slli_epi32(vfitab,2);
 469
 470             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 471             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 472             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 473             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 474             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 475             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 476             Heps             = _mm256_mul_pd(vfeps,H);
 477             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 478             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 479             velec            = _mm256_mul_pd(qq21,VV);
 480             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 481             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
 482
 483             /* Update potential sum for this i atom from the interaction with this j atom. */
 484             velecsum         = _mm256_add_pd(velecsum,velec);
 485
 486             fscal            = felec;
 487
 488             /* Calculate temporary vectorial force */
 489             tx               = _mm256_mul_pd(fscal,dx21);
 490             ty               = _mm256_mul_pd(fscal,dy21);
 491             tz               = _mm256_mul_pd(fscal,dz21);
 492
 493             /* Update vectorial force */
 494             fix2             = _mm256_add_pd(fix2,tx);
 495             fiy2             = _mm256_add_pd(fiy2,ty);
 496             fiz2             = _mm256_add_pd(fiz2,tz);
 497
 498             fjx1             = _mm256_add_pd(fjx1,tx);
 499             fjy1             = _mm256_add_pd(fjy1,ty);
 500             fjz1             = _mm256_add_pd(fjz1,tz);
 501
 502             /**************************
 503              * CALCULATE INTERACTIONS *
 504              **************************/
 505
 506             r22              = _mm256_mul_pd(rsq22,rinv22);
 507
 508             /* Calculate table index by multiplying r with table scale and truncate to integer */
 509             rt               = _mm256_mul_pd(r22,vftabscale);
 510             vfitab           = _mm256_cvttpd_epi32(rt);
 511             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 512             vfitab           = _mm_slli_epi32(vfitab,2);
 513
 514             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 515             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 516             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 517             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 518             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 519             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 520             Heps             = _mm256_mul_pd(vfeps,H);
 521             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 522             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 523             velec            = _mm256_mul_pd(qq22,VV);
 524             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 525             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velecsum         = _mm256_add_pd(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm256_mul_pd(fscal,dx22);
 534             ty               = _mm256_mul_pd(fscal,dy22);
 535             tz               = _mm256_mul_pd(fscal,dz22);
 536
 537             /* Update vectorial force */
 538             fix2             = _mm256_add_pd(fix2,tx);
 539             fiy2             = _mm256_add_pd(fiy2,ty);
 540             fiz2             = _mm256_add_pd(fiz2,tz);
 541
 542             fjx2             = _mm256_add_pd(fjx2,tx);
 543             fjy2             = _mm256_add_pd(fjy2,ty);
 544             fjz2             = _mm256_add_pd(fjz2,tz);
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             r23              = _mm256_mul_pd(rsq23,rinv23);
 551
 552             /* Calculate table index by multiplying r with table scale and truncate to integer */
 553             rt               = _mm256_mul_pd(r23,vftabscale);
 554             vfitab           = _mm256_cvttpd_epi32(rt);
 555             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 556             vfitab           = _mm_slli_epi32(vfitab,2);
 557
 558             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 559             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 560             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 561             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 562             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 563             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 564             Heps             = _mm256_mul_pd(vfeps,H);
 565             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 566             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 567             velec            = _mm256_mul_pd(qq23,VV);
 568             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 569             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
 570
 571             /* Update potential sum for this i atom from the interaction with this j atom. */
 572             velecsum         = _mm256_add_pd(velecsum,velec);
 573
 574             fscal            = felec;
 575
 576             /* Calculate temporary vectorial force */
 577             tx               = _mm256_mul_pd(fscal,dx23);
 578             ty               = _mm256_mul_pd(fscal,dy23);
 579             tz               = _mm256_mul_pd(fscal,dz23);
 580
 581             /* Update vectorial force */
 582             fix2             = _mm256_add_pd(fix2,tx);
 583             fiy2             = _mm256_add_pd(fiy2,ty);
 584             fiz2             = _mm256_add_pd(fiz2,tz);
 585
 586             fjx3             = _mm256_add_pd(fjx3,tx);
 587             fjy3             = _mm256_add_pd(fjy3,ty);
 588             fjz3             = _mm256_add_pd(fjz3,tz);
 589
 590             /**************************
 591              * CALCULATE INTERACTIONS *
 592              **************************/
 593
 594             r31              = _mm256_mul_pd(rsq31,rinv31);
 595
 596             /* Calculate table index by multiplying r with table scale and truncate to integer */
 597             rt               = _mm256_mul_pd(r31,vftabscale);
 598             vfitab           = _mm256_cvttpd_epi32(rt);
 599             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 600             vfitab           = _mm_slli_epi32(vfitab,2);
 601
 602             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 603             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 604             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 605             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 606             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 607             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 608             Heps             = _mm256_mul_pd(vfeps,H);
 609             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 610             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 611             velec            = _mm256_mul_pd(qq31,VV);
 612             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 613             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
 614
 615             /* Update potential sum for this i atom from the interaction with this j atom. */
 616             velecsum         = _mm256_add_pd(velecsum,velec);
 617
 618             fscal            = felec;
 619
 620             /* Calculate temporary vectorial force */
 621             tx               = _mm256_mul_pd(fscal,dx31);
 622             ty               = _mm256_mul_pd(fscal,dy31);
 623             tz               = _mm256_mul_pd(fscal,dz31);
 624
 625             /* Update vectorial force */
 626             fix3             = _mm256_add_pd(fix3,tx);
 627             fiy3             = _mm256_add_pd(fiy3,ty);
 628             fiz3             = _mm256_add_pd(fiz3,tz);
 629
 630             fjx1             = _mm256_add_pd(fjx1,tx);
 631             fjy1             = _mm256_add_pd(fjy1,ty);
 632             fjz1             = _mm256_add_pd(fjz1,tz);
 633
 634             /**************************
 635              * CALCULATE INTERACTIONS *
 636              **************************/
 637
 638             r32              = _mm256_mul_pd(rsq32,rinv32);
 639
 640             /* Calculate table index by multiplying r with table scale and truncate to integer */
 641             rt               = _mm256_mul_pd(r32,vftabscale);
 642             vfitab           = _mm256_cvttpd_epi32(rt);
 643             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 644             vfitab           = _mm_slli_epi32(vfitab,2);
 645
 646             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 647             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 648             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 649             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 650             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 651             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 652             Heps             = _mm256_mul_pd(vfeps,H);
 653             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 654             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 655             velec            = _mm256_mul_pd(qq32,VV);
 656             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 657             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
 658
 659             /* Update potential sum for this i atom from the interaction with this j atom. */
 660             velecsum         = _mm256_add_pd(velecsum,velec);
 661
 662             fscal            = felec;
 663
 664             /* Calculate temporary vectorial force */
 665             tx               = _mm256_mul_pd(fscal,dx32);
 666             ty               = _mm256_mul_pd(fscal,dy32);
 667             tz               = _mm256_mul_pd(fscal,dz32);
 668
 669             /* Update vectorial force */
 670             fix3             = _mm256_add_pd(fix3,tx);
 671             fiy3             = _mm256_add_pd(fiy3,ty);
 672             fiz3             = _mm256_add_pd(fiz3,tz);
 673
 674             fjx2             = _mm256_add_pd(fjx2,tx);
 675             fjy2             = _mm256_add_pd(fjy2,ty);
 676             fjz2             = _mm256_add_pd(fjz2,tz);
 677
 678             /**************************
 679              * CALCULATE INTERACTIONS *
 680              **************************/
 681
 682             r33              = _mm256_mul_pd(rsq33,rinv33);
 683
 684             /* Calculate table index by multiplying r with table scale and truncate to integer */
 685             rt               = _mm256_mul_pd(r33,vftabscale);
 686             vfitab           = _mm256_cvttpd_epi32(rt);
 687             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 688             vfitab           = _mm_slli_epi32(vfitab,2);
 689
 690             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 691             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 692             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 693             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 694             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 695             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 696             Heps             = _mm256_mul_pd(vfeps,H);
 697             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 698             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 699             velec            = _mm256_mul_pd(qq33,VV);
 700             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 701             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
 702
 703             /* Update potential sum for this i atom from the interaction with this j atom. */
 704             velecsum         = _mm256_add_pd(velecsum,velec);
 705
 706             fscal            = felec;
 707
 708             /* Calculate temporary vectorial force */
 709             tx               = _mm256_mul_pd(fscal,dx33);
 710             ty               = _mm256_mul_pd(fscal,dy33);
 711             tz               = _mm256_mul_pd(fscal,dz33);
 712
 713             /* Update vectorial force */
 714             fix3             = _mm256_add_pd(fix3,tx);
 715             fiy3             = _mm256_add_pd(fiy3,ty);
 716             fiz3             = _mm256_add_pd(fiz3,tz);
 717
 718             fjx3             = _mm256_add_pd(fjx3,tx);
 719             fjy3             = _mm256_add_pd(fjy3,ty);
 720             fjz3             = _mm256_add_pd(fjz3,tz);
 721
 722             fjptrA             = f+j_coord_offsetA;
 723             fjptrB             = f+j_coord_offsetB;
 724             fjptrC             = f+j_coord_offsetC;
 725             fjptrD             = f+j_coord_offsetD;
 726
 727             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 728                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 729                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 730
 731             /* Inner loop uses 422 flops */
 732         }
 733
 734         if(jidx<j_index_end)
 735         {
 736
 737             /* Get j neighbor index, and coordinate index */
 738             jnrlistA         = jjnr[jidx];
 739             jnrlistB         = jjnr[jidx+1];
 740             jnrlistC         = jjnr[jidx+2];
 741             jnrlistD         = jjnr[jidx+3];
 742             /* Sign of each element will be negative for non-real atoms.
 743              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 744              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 745              */
 746             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 747
 748             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 749             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 750             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 751
 752             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 753             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 754             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 755             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 756             j_coord_offsetA  = DIM*jnrA;
 757             j_coord_offsetB  = DIM*jnrB;
 758             j_coord_offsetC  = DIM*jnrC;
 759             j_coord_offsetD  = DIM*jnrD;
 760
 761             /* load j atom coordinates */
 762             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 763                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 764                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 765                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 766
 767             /* Calculate displacement vector */
 768             dx00             = _mm256_sub_pd(ix0,jx0);
 769             dy00             = _mm256_sub_pd(iy0,jy0);
 770             dz00             = _mm256_sub_pd(iz0,jz0);
 771             dx11             = _mm256_sub_pd(ix1,jx1);
 772             dy11             = _mm256_sub_pd(iy1,jy1);
 773             dz11             = _mm256_sub_pd(iz1,jz1);
 774             dx12             = _mm256_sub_pd(ix1,jx2);
 775             dy12             = _mm256_sub_pd(iy1,jy2);
 776             dz12             = _mm256_sub_pd(iz1,jz2);
 777             dx13             = _mm256_sub_pd(ix1,jx3);
 778             dy13             = _mm256_sub_pd(iy1,jy3);
 779             dz13             = _mm256_sub_pd(iz1,jz3);
 780             dx21             = _mm256_sub_pd(ix2,jx1);
 781             dy21             = _mm256_sub_pd(iy2,jy1);
 782             dz21             = _mm256_sub_pd(iz2,jz1);
 783             dx22             = _mm256_sub_pd(ix2,jx2);
 784             dy22             = _mm256_sub_pd(iy2,jy2);
 785             dz22             = _mm256_sub_pd(iz2,jz2);
 786             dx23             = _mm256_sub_pd(ix2,jx3);
 787             dy23             = _mm256_sub_pd(iy2,jy3);
 788             dz23             = _mm256_sub_pd(iz2,jz3);
 789             dx31             = _mm256_sub_pd(ix3,jx1);
 790             dy31             = _mm256_sub_pd(iy3,jy1);
 791             dz31             = _mm256_sub_pd(iz3,jz1);
 792             dx32             = _mm256_sub_pd(ix3,jx2);
 793             dy32             = _mm256_sub_pd(iy3,jy2);
 794             dz32             = _mm256_sub_pd(iz3,jz2);
 795             dx33             = _mm256_sub_pd(ix3,jx3);
 796             dy33             = _mm256_sub_pd(iy3,jy3);
 797             dz33             = _mm256_sub_pd(iz3,jz3);
 798
 799             /* Calculate squared distance and things based on it */
 800             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 801             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 802             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 803             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 804             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 805             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 806             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 807             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 808             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 809             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 810
 811             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 812             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 813             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 814             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 815             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 816             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 817             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 818             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 819             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 820
 821             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 822
 823             fjx0             = _mm256_setzero_pd();
 824             fjy0             = _mm256_setzero_pd();
 825             fjz0             = _mm256_setzero_pd();
 826             fjx1             = _mm256_setzero_pd();
 827             fjy1             = _mm256_setzero_pd();
 828             fjz1             = _mm256_setzero_pd();
 829             fjx2             = _mm256_setzero_pd();
 830             fjy2             = _mm256_setzero_pd();
 831             fjz2             = _mm256_setzero_pd();
 832             fjx3             = _mm256_setzero_pd();
 833             fjy3             = _mm256_setzero_pd();
 834             fjz3             = _mm256_setzero_pd();
 835
 836             /**************************
 837              * CALCULATE INTERACTIONS *
 838              **************************/
 839
 840             /* LENNARD-JONES DISPERSION/REPULSION */
 841
 842             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 843             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 844             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 845             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 846             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 847
 848             /* Update potential sum for this i atom from the interaction with this j atom. */
 849             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 850             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 851
 852             fscal            = fvdw;
 853
 854             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 855
 856             /* Calculate temporary vectorial force */
 857             tx               = _mm256_mul_pd(fscal,dx00);
 858             ty               = _mm256_mul_pd(fscal,dy00);
 859             tz               = _mm256_mul_pd(fscal,dz00);
 860
 861             /* Update vectorial force */
 862             fix0             = _mm256_add_pd(fix0,tx);
 863             fiy0             = _mm256_add_pd(fiy0,ty);
 864             fiz0             = _mm256_add_pd(fiz0,tz);
 865
 866             fjx0             = _mm256_add_pd(fjx0,tx);
 867             fjy0             = _mm256_add_pd(fjy0,ty);
 868             fjz0             = _mm256_add_pd(fjz0,tz);
 869
 870             /**************************
 871              * CALCULATE INTERACTIONS *
 872              **************************/
 873
 874             r11              = _mm256_mul_pd(rsq11,rinv11);
 875             r11              = _mm256_andnot_pd(dummy_mask,r11);
 876
 877             /* Calculate table index by multiplying r with table scale and truncate to integer */
 878             rt               = _mm256_mul_pd(r11,vftabscale);
 879             vfitab           = _mm256_cvttpd_epi32(rt);
 880             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 881             vfitab           = _mm_slli_epi32(vfitab,2);
 882
 883             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 884             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 885             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 886             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 887             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 888             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 889             Heps             = _mm256_mul_pd(vfeps,H);
 890             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 891             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 892             velec            = _mm256_mul_pd(qq11,VV);
 893             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 894             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
 895
 896             /* Update potential sum for this i atom from the interaction with this j atom. */
 897             velec            = _mm256_andnot_pd(dummy_mask,velec);
 898             velecsum         = _mm256_add_pd(velecsum,velec);
 899
 900             fscal            = felec;
 901
 902             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 903
 904             /* Calculate temporary vectorial force */
 905             tx               = _mm256_mul_pd(fscal,dx11);
 906             ty               = _mm256_mul_pd(fscal,dy11);
 907             tz               = _mm256_mul_pd(fscal,dz11);
 908
 909             /* Update vectorial force */
 910             fix1             = _mm256_add_pd(fix1,tx);
 911             fiy1             = _mm256_add_pd(fiy1,ty);
 912             fiz1             = _mm256_add_pd(fiz1,tz);
 913
 914             fjx1             = _mm256_add_pd(fjx1,tx);
 915             fjy1             = _mm256_add_pd(fjy1,ty);
 916             fjz1             = _mm256_add_pd(fjz1,tz);
 917
 918             /**************************
 919              * CALCULATE INTERACTIONS *
 920              **************************/
 921
 922             r12              = _mm256_mul_pd(rsq12,rinv12);
 923             r12              = _mm256_andnot_pd(dummy_mask,r12);
 924
 925             /* Calculate table index by multiplying r with table scale and truncate to integer */
 926             rt               = _mm256_mul_pd(r12,vftabscale);
 927             vfitab           = _mm256_cvttpd_epi32(rt);
 928             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 929             vfitab           = _mm_slli_epi32(vfitab,2);
 930
 931             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 932             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 933             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 934             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 935             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 936             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 937             Heps             = _mm256_mul_pd(vfeps,H);
 938             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 939             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 940             velec            = _mm256_mul_pd(qq12,VV);
 941             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 942             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
 943
 944             /* Update potential sum for this i atom from the interaction with this j atom. */
 945             velec            = _mm256_andnot_pd(dummy_mask,velec);
 946             velecsum         = _mm256_add_pd(velecsum,velec);
 947
 948             fscal            = felec;
 949
 950             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 951
 952             /* Calculate temporary vectorial force */
 953             tx               = _mm256_mul_pd(fscal,dx12);
 954             ty               = _mm256_mul_pd(fscal,dy12);
 955             tz               = _mm256_mul_pd(fscal,dz12);
 956
 957             /* Update vectorial force */
 958             fix1             = _mm256_add_pd(fix1,tx);
 959             fiy1             = _mm256_add_pd(fiy1,ty);
 960             fiz1             = _mm256_add_pd(fiz1,tz);
 961
 962             fjx2             = _mm256_add_pd(fjx2,tx);
 963             fjy2             = _mm256_add_pd(fjy2,ty);
 964             fjz2             = _mm256_add_pd(fjz2,tz);
 965
 966             /**************************
 967              * CALCULATE INTERACTIONS *
 968              **************************/
 969
 970             r13              = _mm256_mul_pd(rsq13,rinv13);
 971             r13              = _mm256_andnot_pd(dummy_mask,r13);
 972
 973             /* Calculate table index by multiplying r with table scale and truncate to integer */
 974             rt               = _mm256_mul_pd(r13,vftabscale);
 975             vfitab           = _mm256_cvttpd_epi32(rt);
 976             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 977             vfitab           = _mm_slli_epi32(vfitab,2);
 978
 979             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 980             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 981             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 982             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 983             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 984             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 985             Heps             = _mm256_mul_pd(vfeps,H);
 986             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 987             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 988             velec            = _mm256_mul_pd(qq13,VV);
 989             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 990             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
 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             /* Calculate table index by multiplying r with table scale and truncate to integer */
1022             rt               = _mm256_mul_pd(r21,vftabscale);
1023             vfitab           = _mm256_cvttpd_epi32(rt);
1024             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1025             vfitab           = _mm_slli_epi32(vfitab,2);
1026
1027             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1028             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1029             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1030             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1031             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1032             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1033             Heps             = _mm256_mul_pd(vfeps,H);
1034             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1035             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1036             velec            = _mm256_mul_pd(qq21,VV);
1037             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1038             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
1039
1040             /* Update potential sum for this i atom from the interaction with this j atom. */
1041             velec            = _mm256_andnot_pd(dummy_mask,velec);
1042             velecsum         = _mm256_add_pd(velecsum,velec);
1043
1044             fscal            = felec;
1045
1046             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1047
1048             /* Calculate temporary vectorial force */
1049             tx               = _mm256_mul_pd(fscal,dx21);
1050             ty               = _mm256_mul_pd(fscal,dy21);
1051             tz               = _mm256_mul_pd(fscal,dz21);
1052
1053             /* Update vectorial force */
1054             fix2             = _mm256_add_pd(fix2,tx);
1055             fiy2             = _mm256_add_pd(fiy2,ty);
1056             fiz2             = _mm256_add_pd(fiz2,tz);
1057
1058             fjx1             = _mm256_add_pd(fjx1,tx);
1059             fjy1             = _mm256_add_pd(fjy1,ty);
1060             fjz1             = _mm256_add_pd(fjz1,tz);
1061
1062             /**************************
1063              * CALCULATE INTERACTIONS *
1064              **************************/
1065
1066             r22              = _mm256_mul_pd(rsq22,rinv22);
1067             r22              = _mm256_andnot_pd(dummy_mask,r22);
1068
1069             /* Calculate table index by multiplying r with table scale and truncate to integer */
1070             rt               = _mm256_mul_pd(r22,vftabscale);
1071             vfitab           = _mm256_cvttpd_epi32(rt);
1072             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1073             vfitab           = _mm_slli_epi32(vfitab,2);
1074
1075             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1076             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1077             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1078             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1079             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1080             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1081             Heps             = _mm256_mul_pd(vfeps,H);
1082             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1083             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1084             velec            = _mm256_mul_pd(qq22,VV);
1085             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1086             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
1087
1088             /* Update potential sum for this i atom from the interaction with this j atom. */
1089             velec            = _mm256_andnot_pd(dummy_mask,velec);
1090             velecsum         = _mm256_add_pd(velecsum,velec);
1091
1092             fscal            = felec;
1093
1094             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1095
1096             /* Calculate temporary vectorial force */
1097             tx               = _mm256_mul_pd(fscal,dx22);
1098             ty               = _mm256_mul_pd(fscal,dy22);
1099             tz               = _mm256_mul_pd(fscal,dz22);
1100
1101             /* Update vectorial force */
1102             fix2             = _mm256_add_pd(fix2,tx);
1103             fiy2             = _mm256_add_pd(fiy2,ty);
1104             fiz2             = _mm256_add_pd(fiz2,tz);
1105
1106             fjx2             = _mm256_add_pd(fjx2,tx);
1107             fjy2             = _mm256_add_pd(fjy2,ty);
1108             fjz2             = _mm256_add_pd(fjz2,tz);
1109
1110             /**************************
1111              * CALCULATE INTERACTIONS *
1112              **************************/
1113
1114             r23              = _mm256_mul_pd(rsq23,rinv23);
1115             r23              = _mm256_andnot_pd(dummy_mask,r23);
1116
1117             /* Calculate table index by multiplying r with table scale and truncate to integer */
1118             rt               = _mm256_mul_pd(r23,vftabscale);
1119             vfitab           = _mm256_cvttpd_epi32(rt);
1120             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1121             vfitab           = _mm_slli_epi32(vfitab,2);
1122
1123             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1124             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1125             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1126             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1127             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1128             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1129             Heps             = _mm256_mul_pd(vfeps,H);
1130             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1131             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1132             velec            = _mm256_mul_pd(qq23,VV);
1133             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1134             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
1135
1136             /* Update potential sum for this i atom from the interaction with this j atom. */
1137             velec            = _mm256_andnot_pd(dummy_mask,velec);
1138             velecsum         = _mm256_add_pd(velecsum,velec);
1139
1140             fscal            = felec;
1141
1142             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1143
1144             /* Calculate temporary vectorial force */
1145             tx               = _mm256_mul_pd(fscal,dx23);
1146             ty               = _mm256_mul_pd(fscal,dy23);
1147             tz               = _mm256_mul_pd(fscal,dz23);
1148
1149             /* Update vectorial force */
1150             fix2             = _mm256_add_pd(fix2,tx);
1151             fiy2             = _mm256_add_pd(fiy2,ty);
1152             fiz2             = _mm256_add_pd(fiz2,tz);
1153
1154             fjx3             = _mm256_add_pd(fjx3,tx);
1155             fjy3             = _mm256_add_pd(fjy3,ty);
1156             fjz3             = _mm256_add_pd(fjz3,tz);
1157
1158             /**************************
1159              * CALCULATE INTERACTIONS *
1160              **************************/
1161
1162             r31              = _mm256_mul_pd(rsq31,rinv31);
1163             r31              = _mm256_andnot_pd(dummy_mask,r31);
1164
1165             /* Calculate table index by multiplying r with table scale and truncate to integer */
1166             rt               = _mm256_mul_pd(r31,vftabscale);
1167             vfitab           = _mm256_cvttpd_epi32(rt);
1168             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1169             vfitab           = _mm_slli_epi32(vfitab,2);
1170
1171             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1172             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1173             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1174             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1175             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1176             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1177             Heps             = _mm256_mul_pd(vfeps,H);
1178             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1179             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1180             velec            = _mm256_mul_pd(qq31,VV);
1181             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1182             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
1183
1184             /* Update potential sum for this i atom from the interaction with this j atom. */
1185             velec            = _mm256_andnot_pd(dummy_mask,velec);
1186             velecsum         = _mm256_add_pd(velecsum,velec);
1187
1188             fscal            = felec;
1189
1190             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1191
1192             /* Calculate temporary vectorial force */
1193             tx               = _mm256_mul_pd(fscal,dx31);
1194             ty               = _mm256_mul_pd(fscal,dy31);
1195             tz               = _mm256_mul_pd(fscal,dz31);
1196
1197             /* Update vectorial force */
1198             fix3             = _mm256_add_pd(fix3,tx);
1199             fiy3             = _mm256_add_pd(fiy3,ty);
1200             fiz3             = _mm256_add_pd(fiz3,tz);
1201
1202             fjx1             = _mm256_add_pd(fjx1,tx);
1203             fjy1             = _mm256_add_pd(fjy1,ty);
1204             fjz1             = _mm256_add_pd(fjz1,tz);
1205
1206             /**************************
1207              * CALCULATE INTERACTIONS *
1208              **************************/
1209
1210             r32              = _mm256_mul_pd(rsq32,rinv32);
1211             r32              = _mm256_andnot_pd(dummy_mask,r32);
1212
1213             /* Calculate table index by multiplying r with table scale and truncate to integer */
1214             rt               = _mm256_mul_pd(r32,vftabscale);
1215             vfitab           = _mm256_cvttpd_epi32(rt);
1216             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1217             vfitab           = _mm_slli_epi32(vfitab,2);
1218
1219             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1220             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1221             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1222             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1223             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1224             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1225             Heps             = _mm256_mul_pd(vfeps,H);
1226             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1227             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1228             velec            = _mm256_mul_pd(qq32,VV);
1229             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1230             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
1231
1232             /* Update potential sum for this i atom from the interaction with this j atom. */
1233             velec            = _mm256_andnot_pd(dummy_mask,velec);
1234             velecsum         = _mm256_add_pd(velecsum,velec);
1235
1236             fscal            = felec;
1237
1238             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1239
1240             /* Calculate temporary vectorial force */
1241             tx               = _mm256_mul_pd(fscal,dx32);
1242             ty               = _mm256_mul_pd(fscal,dy32);
1243             tz               = _mm256_mul_pd(fscal,dz32);
1244
1245             /* Update vectorial force */
1246             fix3             = _mm256_add_pd(fix3,tx);
1247             fiy3             = _mm256_add_pd(fiy3,ty);
1248             fiz3             = _mm256_add_pd(fiz3,tz);
1249
1250             fjx2             = _mm256_add_pd(fjx2,tx);
1251             fjy2             = _mm256_add_pd(fjy2,ty);
1252             fjz2             = _mm256_add_pd(fjz2,tz);
1253
1254             /**************************
1255              * CALCULATE INTERACTIONS *
1256              **************************/
1257
1258             r33              = _mm256_mul_pd(rsq33,rinv33);
1259             r33              = _mm256_andnot_pd(dummy_mask,r33);
1260
1261             /* Calculate table index by multiplying r with table scale and truncate to integer */
1262             rt               = _mm256_mul_pd(r33,vftabscale);
1263             vfitab           = _mm256_cvttpd_epi32(rt);
1264             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1265             vfitab           = _mm_slli_epi32(vfitab,2);
1266
1267             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1268             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1269             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1270             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1271             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1272             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1273             Heps             = _mm256_mul_pd(vfeps,H);
1274             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1275             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1276             velec            = _mm256_mul_pd(qq33,VV);
1277             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1278             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
1279
1280             /* Update potential sum for this i atom from the interaction with this j atom. */
1281             velec            = _mm256_andnot_pd(dummy_mask,velec);
1282             velecsum         = _mm256_add_pd(velecsum,velec);
1283
1284             fscal            = felec;
1285
1286             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1287
1288             /* Calculate temporary vectorial force */
1289             tx               = _mm256_mul_pd(fscal,dx33);
1290             ty               = _mm256_mul_pd(fscal,dy33);
1291             tz               = _mm256_mul_pd(fscal,dz33);
1292
1293             /* Update vectorial force */
1294             fix3             = _mm256_add_pd(fix3,tx);
1295             fiy3             = _mm256_add_pd(fiy3,ty);
1296             fiz3             = _mm256_add_pd(fiz3,tz);
1297
1298             fjx3             = _mm256_add_pd(fjx3,tx);
1299             fjy3             = _mm256_add_pd(fjy3,ty);
1300             fjz3             = _mm256_add_pd(fjz3,tz);
1301
1302             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1303             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1304             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1305             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1306
1307             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1308                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1309                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1310
1311             /* Inner loop uses 431 flops */
1312         }
1313
1314         /* End of innermost loop */
1315
1316         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1317                                                  f+i_coord_offset,fshift+i_shift_offset);
1318
1319         ggid                        = gid[iidx];
1320         /* Update potential energies */
1321         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1322         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1323
1324         /* Increment number of inner iterations */
1325         inneriter                  += j_index_end - j_index_start;
1326
1327         /* Outer loop uses 26 flops */
1328     }
1329
1330     /* Increment number of outer iterations */
1331     outeriter        += nri;
1332
1333     /* Update outer/inner flops */
1334
1335     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*431);
1336 }
1337 /*
1338  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_avx_256_double
1339  * Electrostatics interaction: CubicSplineTable
1340  * VdW interaction:            LennardJones
1341  * Geometry:                   Water4-Water4
1342  * Calculate force/pot:        Force
1343  */
1344 void
1345 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_avx_256_double
1346                     (t_nblist * gmx_restrict                nlist,
1347                      rvec * gmx_restrict                    xx,
1348                      rvec * gmx_restrict                    ff,
1349                      t_forcerec * gmx_restrict              fr,
1350                      t_mdatoms * gmx_restrict               mdatoms,
1351                      nb_kernel_data_t * gmx_restrict        kernel_data,
1352                      t_nrnb * gmx_restrict                  nrnb)
1353 {
1354     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1355      * just 0 for non-waters.
1356      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1357      * jnr indices corresponding to data put in the four positions in the SIMD register.
1358      */
1359     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1360     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1361     int              jnrA,jnrB,jnrC,jnrD;
1362     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1363     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1364     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1365     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1366     real             rcutoff_scalar;
1367     real             *shiftvec,*fshift,*x,*f;
1368     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1369     real             scratch[4*DIM];
1370     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1371     real *           vdwioffsetptr0;
1372     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1373     real *           vdwioffsetptr1;
1374     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1375     real *           vdwioffsetptr2;
1376     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1377     real *           vdwioffsetptr3;
1378     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1379     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1380     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1381     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1382     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1383     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1384     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1385     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1386     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1387     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1388     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1389     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1390     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1391     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1392     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1393     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1394     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1395     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1396     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1397     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1398     real             *charge;
1399     int              nvdwtype;
1400     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1401     int              *vdwtype;
1402     real             *vdwparam;
1403     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1404     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1405     __m128i          vfitab;
1406     __m128i          ifour       = _mm_set1_epi32(4);
1407     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1408     real             *vftab;
1409     __m256d          dummy_mask,cutoff_mask;
1410     __m128           tmpmask0,tmpmask1;
1411     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1412     __m256d          one     = _mm256_set1_pd(1.0);
1413     __m256d          two     = _mm256_set1_pd(2.0);
1414     x                = xx[0];
1415     f                = ff[0];
1416
1417     nri              = nlist->nri;
1418     iinr             = nlist->iinr;
1419     jindex           = nlist->jindex;
1420     jjnr             = nlist->jjnr;
1421     shiftidx         = nlist->shift;
1422     gid              = nlist->gid;
1423     shiftvec         = fr->shift_vec[0];
1424     fshift           = fr->fshift[0];
1425     facel            = _mm256_set1_pd(fr->epsfac);
1426     charge           = mdatoms->chargeA;
1427     nvdwtype         = fr->ntype;
1428     vdwparam         = fr->nbfp;
1429     vdwtype          = mdatoms->typeA;
1430
1431     vftab            = kernel_data->table_elec->data;
1432     vftabscale       = _mm256_set1_pd(kernel_data->table_elec->scale);
1433
1434     /* Setup water-specific parameters */
1435     inr              = nlist->iinr[0];
1436     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1437     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1438     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1439     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1440
1441     jq1              = _mm256_set1_pd(charge[inr+1]);
1442     jq2              = _mm256_set1_pd(charge[inr+2]);
1443     jq3              = _mm256_set1_pd(charge[inr+3]);
1444     vdwjidx0A        = 2*vdwtype[inr+0];
1445     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1446     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1447     qq11             = _mm256_mul_pd(iq1,jq1);
1448     qq12             = _mm256_mul_pd(iq1,jq2);
1449     qq13             = _mm256_mul_pd(iq1,jq3);
1450     qq21             = _mm256_mul_pd(iq2,jq1);
1451     qq22             = _mm256_mul_pd(iq2,jq2);
1452     qq23             = _mm256_mul_pd(iq2,jq3);
1453     qq31             = _mm256_mul_pd(iq3,jq1);
1454     qq32             = _mm256_mul_pd(iq3,jq2);
1455     qq33             = _mm256_mul_pd(iq3,jq3);
1456
1457     /* Avoid stupid compiler warnings */
1458     jnrA = jnrB = jnrC = jnrD = 0;
1459     j_coord_offsetA = 0;
1460     j_coord_offsetB = 0;
1461     j_coord_offsetC = 0;
1462     j_coord_offsetD = 0;
1463
1464     outeriter        = 0;
1465     inneriter        = 0;
1466
1467     for(iidx=0;iidx<4*DIM;iidx++)
1468     {
1469         scratch[iidx] = 0.0;
1470     }
1471
1472     /* Start outer loop over neighborlists */
1473     for(iidx=0; iidx<nri; iidx++)
1474     {
1475         /* Load shift vector for this list */
1476         i_shift_offset   = DIM*shiftidx[iidx];
1477
1478         /* Load limits for loop over neighbors */
1479         j_index_start    = jindex[iidx];
1480         j_index_end      = jindex[iidx+1];
1481
1482         /* Get outer coordinate index */
1483         inr              = iinr[iidx];
1484         i_coord_offset   = DIM*inr;
1485
1486         /* Load i particle coords and add shift vector */
1487         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1488                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1489
1490         fix0             = _mm256_setzero_pd();
1491         fiy0             = _mm256_setzero_pd();
1492         fiz0             = _mm256_setzero_pd();
1493         fix1             = _mm256_setzero_pd();
1494         fiy1             = _mm256_setzero_pd();
1495         fiz1             = _mm256_setzero_pd();
1496         fix2             = _mm256_setzero_pd();
1497         fiy2             = _mm256_setzero_pd();
1498         fiz2             = _mm256_setzero_pd();
1499         fix3             = _mm256_setzero_pd();
1500         fiy3             = _mm256_setzero_pd();
1501         fiz3             = _mm256_setzero_pd();
1502
1503         /* Start inner kernel loop */
1504         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1505         {
1506
1507             /* Get j neighbor index, and coordinate index */
1508             jnrA             = jjnr[jidx];
1509             jnrB             = jjnr[jidx+1];
1510             jnrC             = jjnr[jidx+2];
1511             jnrD             = jjnr[jidx+3];
1512             j_coord_offsetA  = DIM*jnrA;
1513             j_coord_offsetB  = DIM*jnrB;
1514             j_coord_offsetC  = DIM*jnrC;
1515             j_coord_offsetD  = DIM*jnrD;
1516
1517             /* load j atom coordinates */
1518             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1519                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1520                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1521                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1522
1523             /* Calculate displacement vector */
1524             dx00             = _mm256_sub_pd(ix0,jx0);
1525             dy00             = _mm256_sub_pd(iy0,jy0);
1526             dz00             = _mm256_sub_pd(iz0,jz0);
1527             dx11             = _mm256_sub_pd(ix1,jx1);
1528             dy11             = _mm256_sub_pd(iy1,jy1);
1529             dz11             = _mm256_sub_pd(iz1,jz1);
1530             dx12             = _mm256_sub_pd(ix1,jx2);
1531             dy12             = _mm256_sub_pd(iy1,jy2);
1532             dz12             = _mm256_sub_pd(iz1,jz2);
1533             dx13             = _mm256_sub_pd(ix1,jx3);
1534             dy13             = _mm256_sub_pd(iy1,jy3);
1535             dz13             = _mm256_sub_pd(iz1,jz3);
1536             dx21             = _mm256_sub_pd(ix2,jx1);
1537             dy21             = _mm256_sub_pd(iy2,jy1);
1538             dz21             = _mm256_sub_pd(iz2,jz1);
1539             dx22             = _mm256_sub_pd(ix2,jx2);
1540             dy22             = _mm256_sub_pd(iy2,jy2);
1541             dz22             = _mm256_sub_pd(iz2,jz2);
1542             dx23             = _mm256_sub_pd(ix2,jx3);
1543             dy23             = _mm256_sub_pd(iy2,jy3);
1544             dz23             = _mm256_sub_pd(iz2,jz3);
1545             dx31             = _mm256_sub_pd(ix3,jx1);
1546             dy31             = _mm256_sub_pd(iy3,jy1);
1547             dz31             = _mm256_sub_pd(iz3,jz1);
1548             dx32             = _mm256_sub_pd(ix3,jx2);
1549             dy32             = _mm256_sub_pd(iy3,jy2);
1550             dz32             = _mm256_sub_pd(iz3,jz2);
1551             dx33             = _mm256_sub_pd(ix3,jx3);
1552             dy33             = _mm256_sub_pd(iy3,jy3);
1553             dz33             = _mm256_sub_pd(iz3,jz3);
1554
1555             /* Calculate squared distance and things based on it */
1556             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1557             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1558             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1559             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1560             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1561             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1562             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1563             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1564             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1565             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1566
1567             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1568             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1569             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1570             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1571             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1572             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1573             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1574             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1575             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1576
1577             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1578
1579             fjx0             = _mm256_setzero_pd();
1580             fjy0             = _mm256_setzero_pd();
1581             fjz0             = _mm256_setzero_pd();
1582             fjx1             = _mm256_setzero_pd();
1583             fjy1             = _mm256_setzero_pd();
1584             fjz1             = _mm256_setzero_pd();
1585             fjx2             = _mm256_setzero_pd();
1586             fjy2             = _mm256_setzero_pd();
1587             fjz2             = _mm256_setzero_pd();
1588             fjx3             = _mm256_setzero_pd();
1589             fjy3             = _mm256_setzero_pd();
1590             fjz3             = _mm256_setzero_pd();
1591
1592             /**************************
1593              * CALCULATE INTERACTIONS *
1594              **************************/
1595
1596             /* LENNARD-JONES DISPERSION/REPULSION */
1597
1598             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1599             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1600
1601             fscal            = fvdw;
1602
1603             /* Calculate temporary vectorial force */
1604             tx               = _mm256_mul_pd(fscal,dx00);
1605             ty               = _mm256_mul_pd(fscal,dy00);
1606             tz               = _mm256_mul_pd(fscal,dz00);
1607
1608             /* Update vectorial force */
1609             fix0             = _mm256_add_pd(fix0,tx);
1610             fiy0             = _mm256_add_pd(fiy0,ty);
1611             fiz0             = _mm256_add_pd(fiz0,tz);
1612
1613             fjx0             = _mm256_add_pd(fjx0,tx);
1614             fjy0             = _mm256_add_pd(fjy0,ty);
1615             fjz0             = _mm256_add_pd(fjz0,tz);
1616
1617             /**************************
1618              * CALCULATE INTERACTIONS *
1619              **************************/
1620
1621             r11              = _mm256_mul_pd(rsq11,rinv11);
1622
1623             /* Calculate table index by multiplying r with table scale and truncate to integer */
1624             rt               = _mm256_mul_pd(r11,vftabscale);
1625             vfitab           = _mm256_cvttpd_epi32(rt);
1626             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1627             vfitab           = _mm_slli_epi32(vfitab,2);
1628
1629             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1630             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1631             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1632             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1633             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1634             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1635             Heps             = _mm256_mul_pd(vfeps,H);
1636             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1637             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1638             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
1639
1640             fscal            = felec;
1641
1642             /* Calculate temporary vectorial force */
1643             tx               = _mm256_mul_pd(fscal,dx11);
1644             ty               = _mm256_mul_pd(fscal,dy11);
1645             tz               = _mm256_mul_pd(fscal,dz11);
1646
1647             /* Update vectorial force */
1648             fix1             = _mm256_add_pd(fix1,tx);
1649             fiy1             = _mm256_add_pd(fiy1,ty);
1650             fiz1             = _mm256_add_pd(fiz1,tz);
1651
1652             fjx1             = _mm256_add_pd(fjx1,tx);
1653             fjy1             = _mm256_add_pd(fjy1,ty);
1654             fjz1             = _mm256_add_pd(fjz1,tz);
1655
1656             /**************************
1657              * CALCULATE INTERACTIONS *
1658              **************************/
1659
1660             r12              = _mm256_mul_pd(rsq12,rinv12);
1661
1662             /* Calculate table index by multiplying r with table scale and truncate to integer */
1663             rt               = _mm256_mul_pd(r12,vftabscale);
1664             vfitab           = _mm256_cvttpd_epi32(rt);
1665             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1666             vfitab           = _mm_slli_epi32(vfitab,2);
1667
1668             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1669             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1670             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1671             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1672             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1673             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1674             Heps             = _mm256_mul_pd(vfeps,H);
1675             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1676             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1677             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
1678
1679             fscal            = felec;
1680
1681             /* Calculate temporary vectorial force */
1682             tx               = _mm256_mul_pd(fscal,dx12);
1683             ty               = _mm256_mul_pd(fscal,dy12);
1684             tz               = _mm256_mul_pd(fscal,dz12);
1685
1686             /* Update vectorial force */
1687             fix1             = _mm256_add_pd(fix1,tx);
1688             fiy1             = _mm256_add_pd(fiy1,ty);
1689             fiz1             = _mm256_add_pd(fiz1,tz);
1690
1691             fjx2             = _mm256_add_pd(fjx2,tx);
1692             fjy2             = _mm256_add_pd(fjy2,ty);
1693             fjz2             = _mm256_add_pd(fjz2,tz);
1694
1695             /**************************
1696              * CALCULATE INTERACTIONS *
1697              **************************/
1698
1699             r13              = _mm256_mul_pd(rsq13,rinv13);
1700
1701             /* Calculate table index by multiplying r with table scale and truncate to integer */
1702             rt               = _mm256_mul_pd(r13,vftabscale);
1703             vfitab           = _mm256_cvttpd_epi32(rt);
1704             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1705             vfitab           = _mm_slli_epi32(vfitab,2);
1706
1707             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1708             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1709             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1710             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1711             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1712             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1713             Heps             = _mm256_mul_pd(vfeps,H);
1714             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1715             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1716             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
1717
1718             fscal            = felec;
1719
1720             /* Calculate temporary vectorial force */
1721             tx               = _mm256_mul_pd(fscal,dx13);
1722             ty               = _mm256_mul_pd(fscal,dy13);
1723             tz               = _mm256_mul_pd(fscal,dz13);
1724
1725             /* Update vectorial force */
1726             fix1             = _mm256_add_pd(fix1,tx);
1727             fiy1             = _mm256_add_pd(fiy1,ty);
1728             fiz1             = _mm256_add_pd(fiz1,tz);
1729
1730             fjx3             = _mm256_add_pd(fjx3,tx);
1731             fjy3             = _mm256_add_pd(fjy3,ty);
1732             fjz3             = _mm256_add_pd(fjz3,tz);
1733
1734             /**************************
1735              * CALCULATE INTERACTIONS *
1736              **************************/
1737
1738             r21              = _mm256_mul_pd(rsq21,rinv21);
1739
1740             /* Calculate table index by multiplying r with table scale and truncate to integer */
1741             rt               = _mm256_mul_pd(r21,vftabscale);
1742             vfitab           = _mm256_cvttpd_epi32(rt);
1743             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1744             vfitab           = _mm_slli_epi32(vfitab,2);
1745
1746             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1747             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1748             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1749             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1750             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1751             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1752             Heps             = _mm256_mul_pd(vfeps,H);
1753             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1754             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1755             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
1756
1757             fscal            = felec;
1758
1759             /* Calculate temporary vectorial force */
1760             tx               = _mm256_mul_pd(fscal,dx21);
1761             ty               = _mm256_mul_pd(fscal,dy21);
1762             tz               = _mm256_mul_pd(fscal,dz21);
1763
1764             /* Update vectorial force */
1765             fix2             = _mm256_add_pd(fix2,tx);
1766             fiy2             = _mm256_add_pd(fiy2,ty);
1767             fiz2             = _mm256_add_pd(fiz2,tz);
1768
1769             fjx1             = _mm256_add_pd(fjx1,tx);
1770             fjy1             = _mm256_add_pd(fjy1,ty);
1771             fjz1             = _mm256_add_pd(fjz1,tz);
1772
1773             /**************************
1774              * CALCULATE INTERACTIONS *
1775              **************************/
1776
1777             r22              = _mm256_mul_pd(rsq22,rinv22);
1778
1779             /* Calculate table index by multiplying r with table scale and truncate to integer */
1780             rt               = _mm256_mul_pd(r22,vftabscale);
1781             vfitab           = _mm256_cvttpd_epi32(rt);
1782             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1783             vfitab           = _mm_slli_epi32(vfitab,2);
1784
1785             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1786             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1787             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1788             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1789             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1790             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1791             Heps             = _mm256_mul_pd(vfeps,H);
1792             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1793             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1794             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
1795
1796             fscal            = felec;
1797
1798             /* Calculate temporary vectorial force */
1799             tx               = _mm256_mul_pd(fscal,dx22);
1800             ty               = _mm256_mul_pd(fscal,dy22);
1801             tz               = _mm256_mul_pd(fscal,dz22);
1802
1803             /* Update vectorial force */
1804             fix2             = _mm256_add_pd(fix2,tx);
1805             fiy2             = _mm256_add_pd(fiy2,ty);
1806             fiz2             = _mm256_add_pd(fiz2,tz);
1807
1808             fjx2             = _mm256_add_pd(fjx2,tx);
1809             fjy2             = _mm256_add_pd(fjy2,ty);
1810             fjz2             = _mm256_add_pd(fjz2,tz);
1811
1812             /**************************
1813              * CALCULATE INTERACTIONS *
1814              **************************/
1815
1816             r23              = _mm256_mul_pd(rsq23,rinv23);
1817
1818             /* Calculate table index by multiplying r with table scale and truncate to integer */
1819             rt               = _mm256_mul_pd(r23,vftabscale);
1820             vfitab           = _mm256_cvttpd_epi32(rt);
1821             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1822             vfitab           = _mm_slli_epi32(vfitab,2);
1823
1824             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1825             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1826             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1827             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1828             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1829             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1830             Heps             = _mm256_mul_pd(vfeps,H);
1831             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1832             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1833             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
1834
1835             fscal            = felec;
1836
1837             /* Calculate temporary vectorial force */
1838             tx               = _mm256_mul_pd(fscal,dx23);
1839             ty               = _mm256_mul_pd(fscal,dy23);
1840             tz               = _mm256_mul_pd(fscal,dz23);
1841
1842             /* Update vectorial force */
1843             fix2             = _mm256_add_pd(fix2,tx);
1844             fiy2             = _mm256_add_pd(fiy2,ty);
1845             fiz2             = _mm256_add_pd(fiz2,tz);
1846
1847             fjx3             = _mm256_add_pd(fjx3,tx);
1848             fjy3             = _mm256_add_pd(fjy3,ty);
1849             fjz3             = _mm256_add_pd(fjz3,tz);
1850
1851             /**************************
1852              * CALCULATE INTERACTIONS *
1853              **************************/
1854
1855             r31              = _mm256_mul_pd(rsq31,rinv31);
1856
1857             /* Calculate table index by multiplying r with table scale and truncate to integer */
1858             rt               = _mm256_mul_pd(r31,vftabscale);
1859             vfitab           = _mm256_cvttpd_epi32(rt);
1860             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1861             vfitab           = _mm_slli_epi32(vfitab,2);
1862
1863             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1864             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1865             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1866             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1867             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1868             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1869             Heps             = _mm256_mul_pd(vfeps,H);
1870             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1871             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1872             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
1873
1874             fscal            = felec;
1875
1876             /* Calculate temporary vectorial force */
1877             tx               = _mm256_mul_pd(fscal,dx31);
1878             ty               = _mm256_mul_pd(fscal,dy31);
1879             tz               = _mm256_mul_pd(fscal,dz31);
1880
1881             /* Update vectorial force */
1882             fix3             = _mm256_add_pd(fix3,tx);
1883             fiy3             = _mm256_add_pd(fiy3,ty);
1884             fiz3             = _mm256_add_pd(fiz3,tz);
1885
1886             fjx1             = _mm256_add_pd(fjx1,tx);
1887             fjy1             = _mm256_add_pd(fjy1,ty);
1888             fjz1             = _mm256_add_pd(fjz1,tz);
1889
1890             /**************************
1891              * CALCULATE INTERACTIONS *
1892              **************************/
1893
1894             r32              = _mm256_mul_pd(rsq32,rinv32);
1895
1896             /* Calculate table index by multiplying r with table scale and truncate to integer */
1897             rt               = _mm256_mul_pd(r32,vftabscale);
1898             vfitab           = _mm256_cvttpd_epi32(rt);
1899             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1900             vfitab           = _mm_slli_epi32(vfitab,2);
1901
1902             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1903             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1904             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1905             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1906             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1907             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1908             Heps             = _mm256_mul_pd(vfeps,H);
1909             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1910             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1911             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
1912
1913             fscal            = felec;
1914
1915             /* Calculate temporary vectorial force */
1916             tx               = _mm256_mul_pd(fscal,dx32);
1917             ty               = _mm256_mul_pd(fscal,dy32);
1918             tz               = _mm256_mul_pd(fscal,dz32);
1919
1920             /* Update vectorial force */
1921             fix3             = _mm256_add_pd(fix3,tx);
1922             fiy3             = _mm256_add_pd(fiy3,ty);
1923             fiz3             = _mm256_add_pd(fiz3,tz);
1924
1925             fjx2             = _mm256_add_pd(fjx2,tx);
1926             fjy2             = _mm256_add_pd(fjy2,ty);
1927             fjz2             = _mm256_add_pd(fjz2,tz);
1928
1929             /**************************
1930              * CALCULATE INTERACTIONS *
1931              **************************/
1932
1933             r33              = _mm256_mul_pd(rsq33,rinv33);
1934
1935             /* Calculate table index by multiplying r with table scale and truncate to integer */
1936             rt               = _mm256_mul_pd(r33,vftabscale);
1937             vfitab           = _mm256_cvttpd_epi32(rt);
1938             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1939             vfitab           = _mm_slli_epi32(vfitab,2);
1940
1941             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1942             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1943             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1944             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1945             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1946             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1947             Heps             = _mm256_mul_pd(vfeps,H);
1948             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1949             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1950             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
1951
1952             fscal            = felec;
1953
1954             /* Calculate temporary vectorial force */
1955             tx               = _mm256_mul_pd(fscal,dx33);
1956             ty               = _mm256_mul_pd(fscal,dy33);
1957             tz               = _mm256_mul_pd(fscal,dz33);
1958
1959             /* Update vectorial force */
1960             fix3             = _mm256_add_pd(fix3,tx);
1961             fiy3             = _mm256_add_pd(fiy3,ty);
1962             fiz3             = _mm256_add_pd(fiz3,tz);
1963
1964             fjx3             = _mm256_add_pd(fjx3,tx);
1965             fjy3             = _mm256_add_pd(fjy3,ty);
1966             fjz3             = _mm256_add_pd(fjz3,tz);
1967
1968             fjptrA             = f+j_coord_offsetA;
1969             fjptrB             = f+j_coord_offsetB;
1970             fjptrC             = f+j_coord_offsetC;
1971             fjptrD             = f+j_coord_offsetD;
1972
1973             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1974                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1975                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1976
1977             /* Inner loop uses 381 flops */
1978         }
1979
1980         if(jidx<j_index_end)
1981         {
1982
1983             /* Get j neighbor index, and coordinate index */
1984             jnrlistA         = jjnr[jidx];
1985             jnrlistB         = jjnr[jidx+1];
1986             jnrlistC         = jjnr[jidx+2];
1987             jnrlistD         = jjnr[jidx+3];
1988             /* Sign of each element will be negative for non-real atoms.
1989              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1990              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1991              */
1992             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1993
1994             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1995             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1996             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1997
1998             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1999             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2000             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2001             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2002             j_coord_offsetA  = DIM*jnrA;
2003             j_coord_offsetB  = DIM*jnrB;
2004             j_coord_offsetC  = DIM*jnrC;
2005             j_coord_offsetD  = DIM*jnrD;
2006
2007             /* load j atom coordinates */
2008             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
2009                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2010                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2011                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
2012
2013             /* Calculate displacement vector */
2014             dx00             = _mm256_sub_pd(ix0,jx0);
2015             dy00             = _mm256_sub_pd(iy0,jy0);
2016             dz00             = _mm256_sub_pd(iz0,jz0);
2017             dx11             = _mm256_sub_pd(ix1,jx1);
2018             dy11             = _mm256_sub_pd(iy1,jy1);
2019             dz11             = _mm256_sub_pd(iz1,jz1);
2020             dx12             = _mm256_sub_pd(ix1,jx2);
2021             dy12             = _mm256_sub_pd(iy1,jy2);
2022             dz12             = _mm256_sub_pd(iz1,jz2);
2023             dx13             = _mm256_sub_pd(ix1,jx3);
2024             dy13             = _mm256_sub_pd(iy1,jy3);
2025             dz13             = _mm256_sub_pd(iz1,jz3);
2026             dx21             = _mm256_sub_pd(ix2,jx1);
2027             dy21             = _mm256_sub_pd(iy2,jy1);
2028             dz21             = _mm256_sub_pd(iz2,jz1);
2029             dx22             = _mm256_sub_pd(ix2,jx2);
2030             dy22             = _mm256_sub_pd(iy2,jy2);
2031             dz22             = _mm256_sub_pd(iz2,jz2);
2032             dx23             = _mm256_sub_pd(ix2,jx3);
2033             dy23             = _mm256_sub_pd(iy2,jy3);
2034             dz23             = _mm256_sub_pd(iz2,jz3);
2035             dx31             = _mm256_sub_pd(ix3,jx1);
2036             dy31             = _mm256_sub_pd(iy3,jy1);
2037             dz31             = _mm256_sub_pd(iz3,jz1);
2038             dx32             = _mm256_sub_pd(ix3,jx2);
2039             dy32             = _mm256_sub_pd(iy3,jy2);
2040             dz32             = _mm256_sub_pd(iz3,jz2);
2041             dx33             = _mm256_sub_pd(ix3,jx3);
2042             dy33             = _mm256_sub_pd(iy3,jy3);
2043             dz33             = _mm256_sub_pd(iz3,jz3);
2044
2045             /* Calculate squared distance and things based on it */
2046             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2047             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2048             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2049             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
2050             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2051             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2052             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
2053             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
2054             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
2055             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
2056
2057             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
2058             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
2059             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
2060             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
2061             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
2062             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
2063             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
2064             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
2065             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
2066
2067             rinvsq00         = gmx_mm256_inv_pd(rsq00);
2068
2069             fjx0             = _mm256_setzero_pd();
2070             fjy0             = _mm256_setzero_pd();
2071             fjz0             = _mm256_setzero_pd();
2072             fjx1             = _mm256_setzero_pd();
2073             fjy1             = _mm256_setzero_pd();
2074             fjz1             = _mm256_setzero_pd();
2075             fjx2             = _mm256_setzero_pd();
2076             fjy2             = _mm256_setzero_pd();
2077             fjz2             = _mm256_setzero_pd();
2078             fjx3             = _mm256_setzero_pd();
2079             fjy3             = _mm256_setzero_pd();
2080             fjz3             = _mm256_setzero_pd();
2081
2082             /**************************
2083              * CALCULATE INTERACTIONS *
2084              **************************/
2085
2086             /* LENNARD-JONES DISPERSION/REPULSION */
2087
2088             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2089             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2090
2091             fscal            = fvdw;
2092
2093             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2094
2095             /* Calculate temporary vectorial force */
2096             tx               = _mm256_mul_pd(fscal,dx00);
2097             ty               = _mm256_mul_pd(fscal,dy00);
2098             tz               = _mm256_mul_pd(fscal,dz00);
2099
2100             /* Update vectorial force */
2101             fix0             = _mm256_add_pd(fix0,tx);
2102             fiy0             = _mm256_add_pd(fiy0,ty);
2103             fiz0             = _mm256_add_pd(fiz0,tz);
2104
2105             fjx0             = _mm256_add_pd(fjx0,tx);
2106             fjy0             = _mm256_add_pd(fjy0,ty);
2107             fjz0             = _mm256_add_pd(fjz0,tz);
2108
2109             /**************************
2110              * CALCULATE INTERACTIONS *
2111              **************************/
2112
2113             r11              = _mm256_mul_pd(rsq11,rinv11);
2114             r11              = _mm256_andnot_pd(dummy_mask,r11);
2115
2116             /* Calculate table index by multiplying r with table scale and truncate to integer */
2117             rt               = _mm256_mul_pd(r11,vftabscale);
2118             vfitab           = _mm256_cvttpd_epi32(rt);
2119             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2120             vfitab           = _mm_slli_epi32(vfitab,2);
2121
2122             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2123             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2124             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2125             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2126             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2127             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2128             Heps             = _mm256_mul_pd(vfeps,H);
2129             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2130             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2131             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
2132
2133             fscal            = felec;
2134
2135             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2136
2137             /* Calculate temporary vectorial force */
2138             tx               = _mm256_mul_pd(fscal,dx11);
2139             ty               = _mm256_mul_pd(fscal,dy11);
2140             tz               = _mm256_mul_pd(fscal,dz11);
2141
2142             /* Update vectorial force */
2143             fix1             = _mm256_add_pd(fix1,tx);
2144             fiy1             = _mm256_add_pd(fiy1,ty);
2145             fiz1             = _mm256_add_pd(fiz1,tz);
2146
2147             fjx1             = _mm256_add_pd(fjx1,tx);
2148             fjy1             = _mm256_add_pd(fjy1,ty);
2149             fjz1             = _mm256_add_pd(fjz1,tz);
2150
2151             /**************************
2152              * CALCULATE INTERACTIONS *
2153              **************************/
2154
2155             r12              = _mm256_mul_pd(rsq12,rinv12);
2156             r12              = _mm256_andnot_pd(dummy_mask,r12);
2157
2158             /* Calculate table index by multiplying r with table scale and truncate to integer */
2159             rt               = _mm256_mul_pd(r12,vftabscale);
2160             vfitab           = _mm256_cvttpd_epi32(rt);
2161             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2162             vfitab           = _mm_slli_epi32(vfitab,2);
2163
2164             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2165             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2166             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2167             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2168             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2169             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2170             Heps             = _mm256_mul_pd(vfeps,H);
2171             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2172             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2173             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
2174
2175             fscal            = felec;
2176
2177             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2178
2179             /* Calculate temporary vectorial force */
2180             tx               = _mm256_mul_pd(fscal,dx12);
2181             ty               = _mm256_mul_pd(fscal,dy12);
2182             tz               = _mm256_mul_pd(fscal,dz12);
2183
2184             /* Update vectorial force */
2185             fix1             = _mm256_add_pd(fix1,tx);
2186             fiy1             = _mm256_add_pd(fiy1,ty);
2187             fiz1             = _mm256_add_pd(fiz1,tz);
2188
2189             fjx2             = _mm256_add_pd(fjx2,tx);
2190             fjy2             = _mm256_add_pd(fjy2,ty);
2191             fjz2             = _mm256_add_pd(fjz2,tz);
2192
2193             /**************************
2194              * CALCULATE INTERACTIONS *
2195              **************************/
2196
2197             r13              = _mm256_mul_pd(rsq13,rinv13);
2198             r13              = _mm256_andnot_pd(dummy_mask,r13);
2199
2200             /* Calculate table index by multiplying r with table scale and truncate to integer */
2201             rt               = _mm256_mul_pd(r13,vftabscale);
2202             vfitab           = _mm256_cvttpd_epi32(rt);
2203             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2204             vfitab           = _mm_slli_epi32(vfitab,2);
2205
2206             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2207             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2208             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2209             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2210             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2211             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2212             Heps             = _mm256_mul_pd(vfeps,H);
2213             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2214             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2215             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
2216
2217             fscal            = felec;
2218
2219             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2220
2221             /* Calculate temporary vectorial force */
2222             tx               = _mm256_mul_pd(fscal,dx13);
2223             ty               = _mm256_mul_pd(fscal,dy13);
2224             tz               = _mm256_mul_pd(fscal,dz13);
2225
2226             /* Update vectorial force */
2227             fix1             = _mm256_add_pd(fix1,tx);
2228             fiy1             = _mm256_add_pd(fiy1,ty);
2229             fiz1             = _mm256_add_pd(fiz1,tz);
2230
2231             fjx3             = _mm256_add_pd(fjx3,tx);
2232             fjy3             = _mm256_add_pd(fjy3,ty);
2233             fjz3             = _mm256_add_pd(fjz3,tz);
2234
2235             /**************************
2236              * CALCULATE INTERACTIONS *
2237              **************************/
2238
2239             r21              = _mm256_mul_pd(rsq21,rinv21);
2240             r21              = _mm256_andnot_pd(dummy_mask,r21);
2241
2242             /* Calculate table index by multiplying r with table scale and truncate to integer */
2243             rt               = _mm256_mul_pd(r21,vftabscale);
2244             vfitab           = _mm256_cvttpd_epi32(rt);
2245             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2246             vfitab           = _mm_slli_epi32(vfitab,2);
2247
2248             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2249             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2250             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2251             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2252             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2253             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2254             Heps             = _mm256_mul_pd(vfeps,H);
2255             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2256             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2257             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
2258
2259             fscal            = felec;
2260
2261             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2262
2263             /* Calculate temporary vectorial force */
2264             tx               = _mm256_mul_pd(fscal,dx21);
2265             ty               = _mm256_mul_pd(fscal,dy21);
2266             tz               = _mm256_mul_pd(fscal,dz21);
2267
2268             /* Update vectorial force */
2269             fix2             = _mm256_add_pd(fix2,tx);
2270             fiy2             = _mm256_add_pd(fiy2,ty);
2271             fiz2             = _mm256_add_pd(fiz2,tz);
2272
2273             fjx1             = _mm256_add_pd(fjx1,tx);
2274             fjy1             = _mm256_add_pd(fjy1,ty);
2275             fjz1             = _mm256_add_pd(fjz1,tz);
2276
2277             /**************************
2278              * CALCULATE INTERACTIONS *
2279              **************************/
2280
2281             r22              = _mm256_mul_pd(rsq22,rinv22);
2282             r22              = _mm256_andnot_pd(dummy_mask,r22);
2283
2284             /* Calculate table index by multiplying r with table scale and truncate to integer */
2285             rt               = _mm256_mul_pd(r22,vftabscale);
2286             vfitab           = _mm256_cvttpd_epi32(rt);
2287             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2288             vfitab           = _mm_slli_epi32(vfitab,2);
2289
2290             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2291             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2292             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2293             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2294             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2295             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2296             Heps             = _mm256_mul_pd(vfeps,H);
2297             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2298             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2299             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
2300
2301             fscal            = felec;
2302
2303             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2304
2305             /* Calculate temporary vectorial force */
2306             tx               = _mm256_mul_pd(fscal,dx22);
2307             ty               = _mm256_mul_pd(fscal,dy22);
2308             tz               = _mm256_mul_pd(fscal,dz22);
2309
2310             /* Update vectorial force */
2311             fix2             = _mm256_add_pd(fix2,tx);
2312             fiy2             = _mm256_add_pd(fiy2,ty);
2313             fiz2             = _mm256_add_pd(fiz2,tz);
2314
2315             fjx2             = _mm256_add_pd(fjx2,tx);
2316             fjy2             = _mm256_add_pd(fjy2,ty);
2317             fjz2             = _mm256_add_pd(fjz2,tz);
2318
2319             /**************************
2320              * CALCULATE INTERACTIONS *
2321              **************************/
2322
2323             r23              = _mm256_mul_pd(rsq23,rinv23);
2324             r23              = _mm256_andnot_pd(dummy_mask,r23);
2325
2326             /* Calculate table index by multiplying r with table scale and truncate to integer */
2327             rt               = _mm256_mul_pd(r23,vftabscale);
2328             vfitab           = _mm256_cvttpd_epi32(rt);
2329             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2330             vfitab           = _mm_slli_epi32(vfitab,2);
2331
2332             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2333             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2334             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2335             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2336             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2337             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2338             Heps             = _mm256_mul_pd(vfeps,H);
2339             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2340             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2341             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
2342
2343             fscal            = felec;
2344
2345             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2346
2347             /* Calculate temporary vectorial force */
2348             tx               = _mm256_mul_pd(fscal,dx23);
2349             ty               = _mm256_mul_pd(fscal,dy23);
2350             tz               = _mm256_mul_pd(fscal,dz23);
2351
2352             /* Update vectorial force */
2353             fix2             = _mm256_add_pd(fix2,tx);
2354             fiy2             = _mm256_add_pd(fiy2,ty);
2355             fiz2             = _mm256_add_pd(fiz2,tz);
2356
2357             fjx3             = _mm256_add_pd(fjx3,tx);
2358             fjy3             = _mm256_add_pd(fjy3,ty);
2359             fjz3             = _mm256_add_pd(fjz3,tz);
2360
2361             /**************************
2362              * CALCULATE INTERACTIONS *
2363              **************************/
2364
2365             r31              = _mm256_mul_pd(rsq31,rinv31);
2366             r31              = _mm256_andnot_pd(dummy_mask,r31);
2367
2368             /* Calculate table index by multiplying r with table scale and truncate to integer */
2369             rt               = _mm256_mul_pd(r31,vftabscale);
2370             vfitab           = _mm256_cvttpd_epi32(rt);
2371             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2372             vfitab           = _mm_slli_epi32(vfitab,2);
2373
2374             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2375             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2376             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2377             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2378             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2379             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2380             Heps             = _mm256_mul_pd(vfeps,H);
2381             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2382             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2383             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
2384
2385             fscal            = felec;
2386
2387             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2388
2389             /* Calculate temporary vectorial force */
2390             tx               = _mm256_mul_pd(fscal,dx31);
2391             ty               = _mm256_mul_pd(fscal,dy31);
2392             tz               = _mm256_mul_pd(fscal,dz31);
2393
2394             /* Update vectorial force */
2395             fix3             = _mm256_add_pd(fix3,tx);
2396             fiy3             = _mm256_add_pd(fiy3,ty);
2397             fiz3             = _mm256_add_pd(fiz3,tz);
2398
2399             fjx1             = _mm256_add_pd(fjx1,tx);
2400             fjy1             = _mm256_add_pd(fjy1,ty);
2401             fjz1             = _mm256_add_pd(fjz1,tz);
2402
2403             /**************************
2404              * CALCULATE INTERACTIONS *
2405              **************************/
2406
2407             r32              = _mm256_mul_pd(rsq32,rinv32);
2408             r32              = _mm256_andnot_pd(dummy_mask,r32);
2409
2410             /* Calculate table index by multiplying r with table scale and truncate to integer */
2411             rt               = _mm256_mul_pd(r32,vftabscale);
2412             vfitab           = _mm256_cvttpd_epi32(rt);
2413             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2414             vfitab           = _mm_slli_epi32(vfitab,2);
2415
2416             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2417             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2418             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2419             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2420             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2421             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2422             Heps             = _mm256_mul_pd(vfeps,H);
2423             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2424             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2425             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
2426
2427             fscal            = felec;
2428
2429             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2430
2431             /* Calculate temporary vectorial force */
2432             tx               = _mm256_mul_pd(fscal,dx32);
2433             ty               = _mm256_mul_pd(fscal,dy32);
2434             tz               = _mm256_mul_pd(fscal,dz32);
2435
2436             /* Update vectorial force */
2437             fix3             = _mm256_add_pd(fix3,tx);
2438             fiy3             = _mm256_add_pd(fiy3,ty);
2439             fiz3             = _mm256_add_pd(fiz3,tz);
2440
2441             fjx2             = _mm256_add_pd(fjx2,tx);
2442             fjy2             = _mm256_add_pd(fjy2,ty);
2443             fjz2             = _mm256_add_pd(fjz2,tz);
2444
2445             /**************************
2446              * CALCULATE INTERACTIONS *
2447              **************************/
2448
2449             r33              = _mm256_mul_pd(rsq33,rinv33);
2450             r33              = _mm256_andnot_pd(dummy_mask,r33);
2451
2452             /* Calculate table index by multiplying r with table scale and truncate to integer */
2453             rt               = _mm256_mul_pd(r33,vftabscale);
2454             vfitab           = _mm256_cvttpd_epi32(rt);
2455             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2456             vfitab           = _mm_slli_epi32(vfitab,2);
2457
2458             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2459             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2460             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2461             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2462             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2463             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2464             Heps             = _mm256_mul_pd(vfeps,H);
2465             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2466             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2467             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
2468
2469             fscal            = felec;
2470
2471             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2472
2473             /* Calculate temporary vectorial force */
2474             tx               = _mm256_mul_pd(fscal,dx33);
2475             ty               = _mm256_mul_pd(fscal,dy33);
2476             tz               = _mm256_mul_pd(fscal,dz33);
2477
2478             /* Update vectorial force */
2479             fix3             = _mm256_add_pd(fix3,tx);
2480             fiy3             = _mm256_add_pd(fiy3,ty);
2481             fiz3             = _mm256_add_pd(fiz3,tz);
2482
2483             fjx3             = _mm256_add_pd(fjx3,tx);
2484             fjy3             = _mm256_add_pd(fjy3,ty);
2485             fjz3             = _mm256_add_pd(fjz3,tz);
2486
2487             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2488             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2489             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2490             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2491
2492             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2493                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2494                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2495
2496             /* Inner loop uses 390 flops */
2497         }
2498
2499         /* End of innermost loop */
2500
2501         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2502                                                  f+i_coord_offset,fshift+i_shift_offset);
2503
2504         /* Increment number of inner iterations */
2505         inneriter                  += j_index_end - j_index_start;
2506
2507         /* Outer loop uses 24 flops */
2508     }
2509
2510     /* Increment number of outer iterations */
2511     outeriter        += nri;
2512
2513     /* Update outer/inner flops */
2514
2515     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*390);
2516 }