src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_avx_256_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_double.h"
  48 #include "kernelutil_x86_avx_256_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_avx_256_double
  52  * Electrostatics interaction: CubicSplineTable
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_avx_256_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  82     real             scratch[4*DIM];
  83     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  84     real *           vdwioffsetptr0;
  85     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  86     real *           vdwioffsetptr1;
  87     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  88     real *           vdwioffsetptr2;
  89     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  90     real *           vdwioffsetptr3;
  91     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  99     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 100     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 104     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 107     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 108     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 109     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 110     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 111     real             *charge;
 112     int              nvdwtype;
 113     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 114     int              *vdwtype;
 115     real             *vdwparam;
 116     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 117     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 118     __m128i          vfitab;
 119     __m128i          ifour       = _mm_set1_epi32(4);
 120     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 121     real             *vftab;
 122     __m256d          dummy_mask,cutoff_mask;
 123     __m128           tmpmask0,tmpmask1;
 124     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 125     __m256d          one     = _mm256_set1_pd(1.0);
 126     __m256d          two     = _mm256_set1_pd(2.0);
 127     x                = xx[0];
 128     f                = ff[0];
 129
 130     nri              = nlist->nri;
 131     iinr             = nlist->iinr;
 132     jindex           = nlist->jindex;
 133     jjnr             = nlist->jjnr;
 134     shiftidx         = nlist->shift;
 135     gid              = nlist->gid;
 136     shiftvec         = fr->shift_vec[0];
 137     fshift           = fr->fshift[0];
 138     facel            = _mm256_set1_pd(fr->epsfac);
 139     charge           = mdatoms->chargeA;
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     vftab            = kernel_data->table_elec->data;
 145     vftabscale       = _mm256_set1_pd(kernel_data->table_elec->scale);
 146
 147     /* Setup water-specific parameters */
 148     inr              = nlist->iinr[0];
 149     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 150     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 151     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 152     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 153
 154     jq1              = _mm256_set1_pd(charge[inr+1]);
 155     jq2              = _mm256_set1_pd(charge[inr+2]);
 156     jq3              = _mm256_set1_pd(charge[inr+3]);
 157     vdwjidx0A        = 2*vdwtype[inr+0];
 158     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 159     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 160     qq11             = _mm256_mul_pd(iq1,jq1);
 161     qq12             = _mm256_mul_pd(iq1,jq2);
 162     qq13             = _mm256_mul_pd(iq1,jq3);
 163     qq21             = _mm256_mul_pd(iq2,jq1);
 164     qq22             = _mm256_mul_pd(iq2,jq2);
 165     qq23             = _mm256_mul_pd(iq2,jq3);
 166     qq31             = _mm256_mul_pd(iq3,jq1);
 167     qq32             = _mm256_mul_pd(iq3,jq2);
 168     qq33             = _mm256_mul_pd(iq3,jq3);
 169
 170     /* Avoid stupid compiler warnings */
 171     jnrA = jnrB = jnrC = jnrD = 0;
 172     j_coord_offsetA = 0;
 173     j_coord_offsetB = 0;
 174     j_coord_offsetC = 0;
 175     j_coord_offsetD = 0;
 176
 177     outeriter        = 0;
 178     inneriter        = 0;
 179
 180     for(iidx=0;iidx<4*DIM;iidx++)
 181     {
 182         scratch[iidx] = 0.0;
 183     }
 184
 185     /* Start outer loop over neighborlists */
 186     for(iidx=0; iidx<nri; iidx++)
 187     {
 188         /* Load shift vector for this list */
 189         i_shift_offset   = DIM*shiftidx[iidx];
 190
 191         /* Load limits for loop over neighbors */
 192         j_index_start    = jindex[iidx];
 193         j_index_end      = jindex[iidx+1];
 194
 195         /* Get outer coordinate index */
 196         inr              = iinr[iidx];
 197         i_coord_offset   = DIM*inr;
 198
 199         /* Load i particle coords and add shift vector */
 200         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 201                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 202
 203         fix0             = _mm256_setzero_pd();
 204         fiy0             = _mm256_setzero_pd();
 205         fiz0             = _mm256_setzero_pd();
 206         fix1             = _mm256_setzero_pd();
 207         fiy1             = _mm256_setzero_pd();
 208         fiz1             = _mm256_setzero_pd();
 209         fix2             = _mm256_setzero_pd();
 210         fiy2             = _mm256_setzero_pd();
 211         fiz2             = _mm256_setzero_pd();
 212         fix3             = _mm256_setzero_pd();
 213         fiy3             = _mm256_setzero_pd();
 214         fiz3             = _mm256_setzero_pd();
 215
 216         /* Reset potential sums */
 217         velecsum         = _mm256_setzero_pd();
 218         vvdwsum          = _mm256_setzero_pd();
 219
 220         /* Start inner kernel loop */
 221         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 222         {
 223
 224             /* Get j neighbor index, and coordinate index */
 225             jnrA             = jjnr[jidx];
 226             jnrB             = jjnr[jidx+1];
 227             jnrC             = jjnr[jidx+2];
 228             jnrD             = jjnr[jidx+3];
 229             j_coord_offsetA  = DIM*jnrA;
 230             j_coord_offsetB  = DIM*jnrB;
 231             j_coord_offsetC  = DIM*jnrC;
 232             j_coord_offsetD  = DIM*jnrD;
 233
 234             /* load j atom coordinates */
 235             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 236                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 237                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 238                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 239
 240             /* Calculate displacement vector */
 241             dx00             = _mm256_sub_pd(ix0,jx0);
 242             dy00             = _mm256_sub_pd(iy0,jy0);
 243             dz00             = _mm256_sub_pd(iz0,jz0);
 244             dx11             = _mm256_sub_pd(ix1,jx1);
 245             dy11             = _mm256_sub_pd(iy1,jy1);
 246             dz11             = _mm256_sub_pd(iz1,jz1);
 247             dx12             = _mm256_sub_pd(ix1,jx2);
 248             dy12             = _mm256_sub_pd(iy1,jy2);
 249             dz12             = _mm256_sub_pd(iz1,jz2);
 250             dx13             = _mm256_sub_pd(ix1,jx3);
 251             dy13             = _mm256_sub_pd(iy1,jy3);
 252             dz13             = _mm256_sub_pd(iz1,jz3);
 253             dx21             = _mm256_sub_pd(ix2,jx1);
 254             dy21             = _mm256_sub_pd(iy2,jy1);
 255             dz21             = _mm256_sub_pd(iz2,jz1);
 256             dx22             = _mm256_sub_pd(ix2,jx2);
 257             dy22             = _mm256_sub_pd(iy2,jy2);
 258             dz22             = _mm256_sub_pd(iz2,jz2);
 259             dx23             = _mm256_sub_pd(ix2,jx3);
 260             dy23             = _mm256_sub_pd(iy2,jy3);
 261             dz23             = _mm256_sub_pd(iz2,jz3);
 262             dx31             = _mm256_sub_pd(ix3,jx1);
 263             dy31             = _mm256_sub_pd(iy3,jy1);
 264             dz31             = _mm256_sub_pd(iz3,jz1);
 265             dx32             = _mm256_sub_pd(ix3,jx2);
 266             dy32             = _mm256_sub_pd(iy3,jy2);
 267             dz32             = _mm256_sub_pd(iz3,jz2);
 268             dx33             = _mm256_sub_pd(ix3,jx3);
 269             dy33             = _mm256_sub_pd(iy3,jy3);
 270             dz33             = _mm256_sub_pd(iz3,jz3);
 271
 272             /* Calculate squared distance and things based on it */
 273             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 274             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 275             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 276             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 277             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 278             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 279             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 280             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 281             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 282             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 283
 284             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 285             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 286             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 287             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 288             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 289             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 290             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 291             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 292             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 293
 294             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 295
 296             fjx0             = _mm256_setzero_pd();
 297             fjy0             = _mm256_setzero_pd();
 298             fjz0             = _mm256_setzero_pd();
 299             fjx1             = _mm256_setzero_pd();
 300             fjy1             = _mm256_setzero_pd();
 301             fjz1             = _mm256_setzero_pd();
 302             fjx2             = _mm256_setzero_pd();
 303             fjy2             = _mm256_setzero_pd();
 304             fjz2             = _mm256_setzero_pd();
 305             fjx3             = _mm256_setzero_pd();
 306             fjy3             = _mm256_setzero_pd();
 307             fjz3             = _mm256_setzero_pd();
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             /* LENNARD-JONES DISPERSION/REPULSION */
 314
 315             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 316             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 317             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 318             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 319             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 320
 321             /* Update potential sum for this i atom from the interaction with this j atom. */
 322             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 323
 324             fscal            = fvdw;
 325
 326             /* Calculate temporary vectorial force */
 327             tx               = _mm256_mul_pd(fscal,dx00);
 328             ty               = _mm256_mul_pd(fscal,dy00);
 329             tz               = _mm256_mul_pd(fscal,dz00);
 330
 331             /* Update vectorial force */
 332             fix0             = _mm256_add_pd(fix0,tx);
 333             fiy0             = _mm256_add_pd(fiy0,ty);
 334             fiz0             = _mm256_add_pd(fiz0,tz);
 335
 336             fjx0             = _mm256_add_pd(fjx0,tx);
 337             fjy0             = _mm256_add_pd(fjy0,ty);
 338             fjz0             = _mm256_add_pd(fjz0,tz);
 339
 340             /**************************
 341              * CALCULATE INTERACTIONS *
 342              **************************/
 343
 344             r11              = _mm256_mul_pd(rsq11,rinv11);
 345
 346             /* Calculate table index by multiplying r with table scale and truncate to integer */
 347             rt               = _mm256_mul_pd(r11,vftabscale);
 348             vfitab           = _mm256_cvttpd_epi32(rt);
 349             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 350             vfitab           = _mm_slli_epi32(vfitab,2);
 351
 352             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 353             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 354             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 355             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 356             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 357             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 358             Heps             = _mm256_mul_pd(vfeps,H);
 359             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 360             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 361             velec            = _mm256_mul_pd(qq11,VV);
 362             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 363             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
 364
 365             /* Update potential sum for this i atom from the interaction with this j atom. */
 366             velecsum         = _mm256_add_pd(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             /* Calculate temporary vectorial force */
 371             tx               = _mm256_mul_pd(fscal,dx11);
 372             ty               = _mm256_mul_pd(fscal,dy11);
 373             tz               = _mm256_mul_pd(fscal,dz11);
 374
 375             /* Update vectorial force */
 376             fix1             = _mm256_add_pd(fix1,tx);
 377             fiy1             = _mm256_add_pd(fiy1,ty);
 378             fiz1             = _mm256_add_pd(fiz1,tz);
 379
 380             fjx1             = _mm256_add_pd(fjx1,tx);
 381             fjy1             = _mm256_add_pd(fjy1,ty);
 382             fjz1             = _mm256_add_pd(fjz1,tz);
 383
 384             /**************************
 385              * CALCULATE INTERACTIONS *
 386              **************************/
 387
 388             r12              = _mm256_mul_pd(rsq12,rinv12);
 389
 390             /* Calculate table index by multiplying r with table scale and truncate to integer */
 391             rt               = _mm256_mul_pd(r12,vftabscale);
 392             vfitab           = _mm256_cvttpd_epi32(rt);
 393             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 394             vfitab           = _mm_slli_epi32(vfitab,2);
 395
 396             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 397             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 398             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 399             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 400             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 401             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 402             Heps             = _mm256_mul_pd(vfeps,H);
 403             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 404             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 405             velec            = _mm256_mul_pd(qq12,VV);
 406             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 407             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
 408
 409             /* Update potential sum for this i atom from the interaction with this j atom. */
 410             velecsum         = _mm256_add_pd(velecsum,velec);
 411
 412             fscal            = felec;
 413
 414             /* Calculate temporary vectorial force */
 415             tx               = _mm256_mul_pd(fscal,dx12);
 416             ty               = _mm256_mul_pd(fscal,dy12);
 417             tz               = _mm256_mul_pd(fscal,dz12);
 418
 419             /* Update vectorial force */
 420             fix1             = _mm256_add_pd(fix1,tx);
 421             fiy1             = _mm256_add_pd(fiy1,ty);
 422             fiz1             = _mm256_add_pd(fiz1,tz);
 423
 424             fjx2             = _mm256_add_pd(fjx2,tx);
 425             fjy2             = _mm256_add_pd(fjy2,ty);
 426             fjz2             = _mm256_add_pd(fjz2,tz);
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             r13              = _mm256_mul_pd(rsq13,rinv13);
 433
 434             /* Calculate table index by multiplying r with table scale and truncate to integer */
 435             rt               = _mm256_mul_pd(r13,vftabscale);
 436             vfitab           = _mm256_cvttpd_epi32(rt);
 437             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 438             vfitab           = _mm_slli_epi32(vfitab,2);
 439
 440             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 441             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 442             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 443             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 444             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 445             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 446             Heps             = _mm256_mul_pd(vfeps,H);
 447             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 448             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 449             velec            = _mm256_mul_pd(qq13,VV);
 450             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 451             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
 452
 453             /* Update potential sum for this i atom from the interaction with this j atom. */
 454             velecsum         = _mm256_add_pd(velecsum,velec);
 455
 456             fscal            = felec;
 457
 458             /* Calculate temporary vectorial force */
 459             tx               = _mm256_mul_pd(fscal,dx13);
 460             ty               = _mm256_mul_pd(fscal,dy13);
 461             tz               = _mm256_mul_pd(fscal,dz13);
 462
 463             /* Update vectorial force */
 464             fix1             = _mm256_add_pd(fix1,tx);
 465             fiy1             = _mm256_add_pd(fiy1,ty);
 466             fiz1             = _mm256_add_pd(fiz1,tz);
 467
 468             fjx3             = _mm256_add_pd(fjx3,tx);
 469             fjy3             = _mm256_add_pd(fjy3,ty);
 470             fjz3             = _mm256_add_pd(fjz3,tz);
 471
 472             /**************************
 473              * CALCULATE INTERACTIONS *
 474              **************************/
 475
 476             r21              = _mm256_mul_pd(rsq21,rinv21);
 477
 478             /* Calculate table index by multiplying r with table scale and truncate to integer */
 479             rt               = _mm256_mul_pd(r21,vftabscale);
 480             vfitab           = _mm256_cvttpd_epi32(rt);
 481             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 482             vfitab           = _mm_slli_epi32(vfitab,2);
 483
 484             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 485             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 486             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 487             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 488             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 489             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 490             Heps             = _mm256_mul_pd(vfeps,H);
 491             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 492             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 493             velec            = _mm256_mul_pd(qq21,VV);
 494             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 495             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
 496
 497             /* Update potential sum for this i atom from the interaction with this j atom. */
 498             velecsum         = _mm256_add_pd(velecsum,velec);
 499
 500             fscal            = felec;
 501
 502             /* Calculate temporary vectorial force */
 503             tx               = _mm256_mul_pd(fscal,dx21);
 504             ty               = _mm256_mul_pd(fscal,dy21);
 505             tz               = _mm256_mul_pd(fscal,dz21);
 506
 507             /* Update vectorial force */
 508             fix2             = _mm256_add_pd(fix2,tx);
 509             fiy2             = _mm256_add_pd(fiy2,ty);
 510             fiz2             = _mm256_add_pd(fiz2,tz);
 511
 512             fjx1             = _mm256_add_pd(fjx1,tx);
 513             fjy1             = _mm256_add_pd(fjy1,ty);
 514             fjz1             = _mm256_add_pd(fjz1,tz);
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             r22              = _mm256_mul_pd(rsq22,rinv22);
 521
 522             /* Calculate table index by multiplying r with table scale and truncate to integer */
 523             rt               = _mm256_mul_pd(r22,vftabscale);
 524             vfitab           = _mm256_cvttpd_epi32(rt);
 525             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 526             vfitab           = _mm_slli_epi32(vfitab,2);
 527
 528             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 529             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 530             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 531             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 532             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 533             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 534             Heps             = _mm256_mul_pd(vfeps,H);
 535             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 536             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 537             velec            = _mm256_mul_pd(qq22,VV);
 538             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 539             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
 540
 541             /* Update potential sum for this i atom from the interaction with this j atom. */
 542             velecsum         = _mm256_add_pd(velecsum,velec);
 543
 544             fscal            = felec;
 545
 546             /* Calculate temporary vectorial force */
 547             tx               = _mm256_mul_pd(fscal,dx22);
 548             ty               = _mm256_mul_pd(fscal,dy22);
 549             tz               = _mm256_mul_pd(fscal,dz22);
 550
 551             /* Update vectorial force */
 552             fix2             = _mm256_add_pd(fix2,tx);
 553             fiy2             = _mm256_add_pd(fiy2,ty);
 554             fiz2             = _mm256_add_pd(fiz2,tz);
 555
 556             fjx2             = _mm256_add_pd(fjx2,tx);
 557             fjy2             = _mm256_add_pd(fjy2,ty);
 558             fjz2             = _mm256_add_pd(fjz2,tz);
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             r23              = _mm256_mul_pd(rsq23,rinv23);
 565
 566             /* Calculate table index by multiplying r with table scale and truncate to integer */
 567             rt               = _mm256_mul_pd(r23,vftabscale);
 568             vfitab           = _mm256_cvttpd_epi32(rt);
 569             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 570             vfitab           = _mm_slli_epi32(vfitab,2);
 571
 572             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 573             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 574             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 575             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 576             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 577             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 578             Heps             = _mm256_mul_pd(vfeps,H);
 579             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 580             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 581             velec            = _mm256_mul_pd(qq23,VV);
 582             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 583             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
 584
 585             /* Update potential sum for this i atom from the interaction with this j atom. */
 586             velecsum         = _mm256_add_pd(velecsum,velec);
 587
 588             fscal            = felec;
 589
 590             /* Calculate temporary vectorial force */
 591             tx               = _mm256_mul_pd(fscal,dx23);
 592             ty               = _mm256_mul_pd(fscal,dy23);
 593             tz               = _mm256_mul_pd(fscal,dz23);
 594
 595             /* Update vectorial force */
 596             fix2             = _mm256_add_pd(fix2,tx);
 597             fiy2             = _mm256_add_pd(fiy2,ty);
 598             fiz2             = _mm256_add_pd(fiz2,tz);
 599
 600             fjx3             = _mm256_add_pd(fjx3,tx);
 601             fjy3             = _mm256_add_pd(fjy3,ty);
 602             fjz3             = _mm256_add_pd(fjz3,tz);
 603
 604             /**************************
 605              * CALCULATE INTERACTIONS *
 606              **************************/
 607
 608             r31              = _mm256_mul_pd(rsq31,rinv31);
 609
 610             /* Calculate table index by multiplying r with table scale and truncate to integer */
 611             rt               = _mm256_mul_pd(r31,vftabscale);
 612             vfitab           = _mm256_cvttpd_epi32(rt);
 613             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 614             vfitab           = _mm_slli_epi32(vfitab,2);
 615
 616             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 617             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 618             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 619             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 620             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 621             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 622             Heps             = _mm256_mul_pd(vfeps,H);
 623             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 624             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 625             velec            = _mm256_mul_pd(qq31,VV);
 626             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 627             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
 628
 629             /* Update potential sum for this i atom from the interaction with this j atom. */
 630             velecsum         = _mm256_add_pd(velecsum,velec);
 631
 632             fscal            = felec;
 633
 634             /* Calculate temporary vectorial force */
 635             tx               = _mm256_mul_pd(fscal,dx31);
 636             ty               = _mm256_mul_pd(fscal,dy31);
 637             tz               = _mm256_mul_pd(fscal,dz31);
 638
 639             /* Update vectorial force */
 640             fix3             = _mm256_add_pd(fix3,tx);
 641             fiy3             = _mm256_add_pd(fiy3,ty);
 642             fiz3             = _mm256_add_pd(fiz3,tz);
 643
 644             fjx1             = _mm256_add_pd(fjx1,tx);
 645             fjy1             = _mm256_add_pd(fjy1,ty);
 646             fjz1             = _mm256_add_pd(fjz1,tz);
 647
 648             /**************************
 649              * CALCULATE INTERACTIONS *
 650              **************************/
 651
 652             r32              = _mm256_mul_pd(rsq32,rinv32);
 653
 654             /* Calculate table index by multiplying r with table scale and truncate to integer */
 655             rt               = _mm256_mul_pd(r32,vftabscale);
 656             vfitab           = _mm256_cvttpd_epi32(rt);
 657             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 658             vfitab           = _mm_slli_epi32(vfitab,2);
 659
 660             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 661             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 662             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 663             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 664             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 665             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 666             Heps             = _mm256_mul_pd(vfeps,H);
 667             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 668             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 669             velec            = _mm256_mul_pd(qq32,VV);
 670             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 671             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
 672
 673             /* Update potential sum for this i atom from the interaction with this j atom. */
 674             velecsum         = _mm256_add_pd(velecsum,velec);
 675
 676             fscal            = felec;
 677
 678             /* Calculate temporary vectorial force */
 679             tx               = _mm256_mul_pd(fscal,dx32);
 680             ty               = _mm256_mul_pd(fscal,dy32);
 681             tz               = _mm256_mul_pd(fscal,dz32);
 682
 683             /* Update vectorial force */
 684             fix3             = _mm256_add_pd(fix3,tx);
 685             fiy3             = _mm256_add_pd(fiy3,ty);
 686             fiz3             = _mm256_add_pd(fiz3,tz);
 687
 688             fjx2             = _mm256_add_pd(fjx2,tx);
 689             fjy2             = _mm256_add_pd(fjy2,ty);
 690             fjz2             = _mm256_add_pd(fjz2,tz);
 691
 692             /**************************
 693              * CALCULATE INTERACTIONS *
 694              **************************/
 695
 696             r33              = _mm256_mul_pd(rsq33,rinv33);
 697
 698             /* Calculate table index by multiplying r with table scale and truncate to integer */
 699             rt               = _mm256_mul_pd(r33,vftabscale);
 700             vfitab           = _mm256_cvttpd_epi32(rt);
 701             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 702             vfitab           = _mm_slli_epi32(vfitab,2);
 703
 704             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 705             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 706             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 707             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 708             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 709             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 710             Heps             = _mm256_mul_pd(vfeps,H);
 711             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 712             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 713             velec            = _mm256_mul_pd(qq33,VV);
 714             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 715             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
 716
 717             /* Update potential sum for this i atom from the interaction with this j atom. */
 718             velecsum         = _mm256_add_pd(velecsum,velec);
 719
 720             fscal            = felec;
 721
 722             /* Calculate temporary vectorial force */
 723             tx               = _mm256_mul_pd(fscal,dx33);
 724             ty               = _mm256_mul_pd(fscal,dy33);
 725             tz               = _mm256_mul_pd(fscal,dz33);
 726
 727             /* Update vectorial force */
 728             fix3             = _mm256_add_pd(fix3,tx);
 729             fiy3             = _mm256_add_pd(fiy3,ty);
 730             fiz3             = _mm256_add_pd(fiz3,tz);
 731
 732             fjx3             = _mm256_add_pd(fjx3,tx);
 733             fjy3             = _mm256_add_pd(fjy3,ty);
 734             fjz3             = _mm256_add_pd(fjz3,tz);
 735
 736             fjptrA             = f+j_coord_offsetA;
 737             fjptrB             = f+j_coord_offsetB;
 738             fjptrC             = f+j_coord_offsetC;
 739             fjptrD             = f+j_coord_offsetD;
 740
 741             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 742                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 743                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 744
 745             /* Inner loop uses 422 flops */
 746         }
 747
 748         if(jidx<j_index_end)
 749         {
 750
 751             /* Get j neighbor index, and coordinate index */
 752             jnrlistA         = jjnr[jidx];
 753             jnrlistB         = jjnr[jidx+1];
 754             jnrlistC         = jjnr[jidx+2];
 755             jnrlistD         = jjnr[jidx+3];
 756             /* Sign of each element will be negative for non-real atoms.
 757              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 758              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 759              */
 760             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 761
 762             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 763             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 764             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 765
 766             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 767             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 768             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 769             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 770             j_coord_offsetA  = DIM*jnrA;
 771             j_coord_offsetB  = DIM*jnrB;
 772             j_coord_offsetC  = DIM*jnrC;
 773             j_coord_offsetD  = DIM*jnrD;
 774
 775             /* load j atom coordinates */
 776             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 777                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 778                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 779                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 780
 781             /* Calculate displacement vector */
 782             dx00             = _mm256_sub_pd(ix0,jx0);
 783             dy00             = _mm256_sub_pd(iy0,jy0);
 784             dz00             = _mm256_sub_pd(iz0,jz0);
 785             dx11             = _mm256_sub_pd(ix1,jx1);
 786             dy11             = _mm256_sub_pd(iy1,jy1);
 787             dz11             = _mm256_sub_pd(iz1,jz1);
 788             dx12             = _mm256_sub_pd(ix1,jx2);
 789             dy12             = _mm256_sub_pd(iy1,jy2);
 790             dz12             = _mm256_sub_pd(iz1,jz2);
 791             dx13             = _mm256_sub_pd(ix1,jx3);
 792             dy13             = _mm256_sub_pd(iy1,jy3);
 793             dz13             = _mm256_sub_pd(iz1,jz3);
 794             dx21             = _mm256_sub_pd(ix2,jx1);
 795             dy21             = _mm256_sub_pd(iy2,jy1);
 796             dz21             = _mm256_sub_pd(iz2,jz1);
 797             dx22             = _mm256_sub_pd(ix2,jx2);
 798             dy22             = _mm256_sub_pd(iy2,jy2);
 799             dz22             = _mm256_sub_pd(iz2,jz2);
 800             dx23             = _mm256_sub_pd(ix2,jx3);
 801             dy23             = _mm256_sub_pd(iy2,jy3);
 802             dz23             = _mm256_sub_pd(iz2,jz3);
 803             dx31             = _mm256_sub_pd(ix3,jx1);
 804             dy31             = _mm256_sub_pd(iy3,jy1);
 805             dz31             = _mm256_sub_pd(iz3,jz1);
 806             dx32             = _mm256_sub_pd(ix3,jx2);
 807             dy32             = _mm256_sub_pd(iy3,jy2);
 808             dz32             = _mm256_sub_pd(iz3,jz2);
 809             dx33             = _mm256_sub_pd(ix3,jx3);
 810             dy33             = _mm256_sub_pd(iy3,jy3);
 811             dz33             = _mm256_sub_pd(iz3,jz3);
 812
 813             /* Calculate squared distance and things based on it */
 814             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 815             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 816             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 817             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 818             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 819             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 820             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 821             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 822             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 823             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 824
 825             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 826             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 827             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 828             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 829             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 830             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 831             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 832             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 833             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 834
 835             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 836
 837             fjx0             = _mm256_setzero_pd();
 838             fjy0             = _mm256_setzero_pd();
 839             fjz0             = _mm256_setzero_pd();
 840             fjx1             = _mm256_setzero_pd();
 841             fjy1             = _mm256_setzero_pd();
 842             fjz1             = _mm256_setzero_pd();
 843             fjx2             = _mm256_setzero_pd();
 844             fjy2             = _mm256_setzero_pd();
 845             fjz2             = _mm256_setzero_pd();
 846             fjx3             = _mm256_setzero_pd();
 847             fjy3             = _mm256_setzero_pd();
 848             fjz3             = _mm256_setzero_pd();
 849
 850             /**************************
 851              * CALCULATE INTERACTIONS *
 852              **************************/
 853
 854             /* LENNARD-JONES DISPERSION/REPULSION */
 855
 856             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 857             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 858             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 859             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 860             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 861
 862             /* Update potential sum for this i atom from the interaction with this j atom. */
 863             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 864             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 865
 866             fscal            = fvdw;
 867
 868             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 869
 870             /* Calculate temporary vectorial force */
 871             tx               = _mm256_mul_pd(fscal,dx00);
 872             ty               = _mm256_mul_pd(fscal,dy00);
 873             tz               = _mm256_mul_pd(fscal,dz00);
 874
 875             /* Update vectorial force */
 876             fix0             = _mm256_add_pd(fix0,tx);
 877             fiy0             = _mm256_add_pd(fiy0,ty);
 878             fiz0             = _mm256_add_pd(fiz0,tz);
 879
 880             fjx0             = _mm256_add_pd(fjx0,tx);
 881             fjy0             = _mm256_add_pd(fjy0,ty);
 882             fjz0             = _mm256_add_pd(fjz0,tz);
 883
 884             /**************************
 885              * CALCULATE INTERACTIONS *
 886              **************************/
 887
 888             r11              = _mm256_mul_pd(rsq11,rinv11);
 889             r11              = _mm256_andnot_pd(dummy_mask,r11);
 890
 891             /* Calculate table index by multiplying r with table scale and truncate to integer */
 892             rt               = _mm256_mul_pd(r11,vftabscale);
 893             vfitab           = _mm256_cvttpd_epi32(rt);
 894             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 895             vfitab           = _mm_slli_epi32(vfitab,2);
 896
 897             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 898             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 899             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 900             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 901             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 902             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 903             Heps             = _mm256_mul_pd(vfeps,H);
 904             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 905             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 906             velec            = _mm256_mul_pd(qq11,VV);
 907             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 908             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
 909
 910             /* Update potential sum for this i atom from the interaction with this j atom. */
 911             velec            = _mm256_andnot_pd(dummy_mask,velec);
 912             velecsum         = _mm256_add_pd(velecsum,velec);
 913
 914             fscal            = felec;
 915
 916             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 917
 918             /* Calculate temporary vectorial force */
 919             tx               = _mm256_mul_pd(fscal,dx11);
 920             ty               = _mm256_mul_pd(fscal,dy11);
 921             tz               = _mm256_mul_pd(fscal,dz11);
 922
 923             /* Update vectorial force */
 924             fix1             = _mm256_add_pd(fix1,tx);
 925             fiy1             = _mm256_add_pd(fiy1,ty);
 926             fiz1             = _mm256_add_pd(fiz1,tz);
 927
 928             fjx1             = _mm256_add_pd(fjx1,tx);
 929             fjy1             = _mm256_add_pd(fjy1,ty);
 930             fjz1             = _mm256_add_pd(fjz1,tz);
 931
 932             /**************************
 933              * CALCULATE INTERACTIONS *
 934              **************************/
 935
 936             r12              = _mm256_mul_pd(rsq12,rinv12);
 937             r12              = _mm256_andnot_pd(dummy_mask,r12);
 938
 939             /* Calculate table index by multiplying r with table scale and truncate to integer */
 940             rt               = _mm256_mul_pd(r12,vftabscale);
 941             vfitab           = _mm256_cvttpd_epi32(rt);
 942             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 943             vfitab           = _mm_slli_epi32(vfitab,2);
 944
 945             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 946             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 947             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 948             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 949             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 950             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 951             Heps             = _mm256_mul_pd(vfeps,H);
 952             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 953             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 954             velec            = _mm256_mul_pd(qq12,VV);
 955             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 956             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
 957
 958             /* Update potential sum for this i atom from the interaction with this j atom. */
 959             velec            = _mm256_andnot_pd(dummy_mask,velec);
 960             velecsum         = _mm256_add_pd(velecsum,velec);
 961
 962             fscal            = felec;
 963
 964             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 965
 966             /* Calculate temporary vectorial force */
 967             tx               = _mm256_mul_pd(fscal,dx12);
 968             ty               = _mm256_mul_pd(fscal,dy12);
 969             tz               = _mm256_mul_pd(fscal,dz12);
 970
 971             /* Update vectorial force */
 972             fix1             = _mm256_add_pd(fix1,tx);
 973             fiy1             = _mm256_add_pd(fiy1,ty);
 974             fiz1             = _mm256_add_pd(fiz1,tz);
 975
 976             fjx2             = _mm256_add_pd(fjx2,tx);
 977             fjy2             = _mm256_add_pd(fjy2,ty);
 978             fjz2             = _mm256_add_pd(fjz2,tz);
 979
 980             /**************************
 981              * CALCULATE INTERACTIONS *
 982              **************************/
 983
 984             r13              = _mm256_mul_pd(rsq13,rinv13);
 985             r13              = _mm256_andnot_pd(dummy_mask,r13);
 986
 987             /* Calculate table index by multiplying r with table scale and truncate to integer */
 988             rt               = _mm256_mul_pd(r13,vftabscale);
 989             vfitab           = _mm256_cvttpd_epi32(rt);
 990             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 991             vfitab           = _mm_slli_epi32(vfitab,2);
 992
 993             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 994             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 995             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 996             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 997             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 998             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 999             Heps             = _mm256_mul_pd(vfeps,H);
1000             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1001             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1002             velec            = _mm256_mul_pd(qq13,VV);
1003             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1004             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
1005
1006             /* Update potential sum for this i atom from the interaction with this j atom. */
1007             velec            = _mm256_andnot_pd(dummy_mask,velec);
1008             velecsum         = _mm256_add_pd(velecsum,velec);
1009
1010             fscal            = felec;
1011
1012             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm256_mul_pd(fscal,dx13);
1016             ty               = _mm256_mul_pd(fscal,dy13);
1017             tz               = _mm256_mul_pd(fscal,dz13);
1018
1019             /* Update vectorial force */
1020             fix1             = _mm256_add_pd(fix1,tx);
1021             fiy1             = _mm256_add_pd(fiy1,ty);
1022             fiz1             = _mm256_add_pd(fiz1,tz);
1023
1024             fjx3             = _mm256_add_pd(fjx3,tx);
1025             fjy3             = _mm256_add_pd(fjy3,ty);
1026             fjz3             = _mm256_add_pd(fjz3,tz);
1027
1028             /**************************
1029              * CALCULATE INTERACTIONS *
1030              **************************/
1031
1032             r21              = _mm256_mul_pd(rsq21,rinv21);
1033             r21              = _mm256_andnot_pd(dummy_mask,r21);
1034
1035             /* Calculate table index by multiplying r with table scale and truncate to integer */
1036             rt               = _mm256_mul_pd(r21,vftabscale);
1037             vfitab           = _mm256_cvttpd_epi32(rt);
1038             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1039             vfitab           = _mm_slli_epi32(vfitab,2);
1040
1041             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1042             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1043             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1044             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1045             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1046             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1047             Heps             = _mm256_mul_pd(vfeps,H);
1048             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1049             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1050             velec            = _mm256_mul_pd(qq21,VV);
1051             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1052             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
1053
1054             /* Update potential sum for this i atom from the interaction with this j atom. */
1055             velec            = _mm256_andnot_pd(dummy_mask,velec);
1056             velecsum         = _mm256_add_pd(velecsum,velec);
1057
1058             fscal            = felec;
1059
1060             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1061
1062             /* Calculate temporary vectorial force */
1063             tx               = _mm256_mul_pd(fscal,dx21);
1064             ty               = _mm256_mul_pd(fscal,dy21);
1065             tz               = _mm256_mul_pd(fscal,dz21);
1066
1067             /* Update vectorial force */
1068             fix2             = _mm256_add_pd(fix2,tx);
1069             fiy2             = _mm256_add_pd(fiy2,ty);
1070             fiz2             = _mm256_add_pd(fiz2,tz);
1071
1072             fjx1             = _mm256_add_pd(fjx1,tx);
1073             fjy1             = _mm256_add_pd(fjy1,ty);
1074             fjz1             = _mm256_add_pd(fjz1,tz);
1075
1076             /**************************
1077              * CALCULATE INTERACTIONS *
1078              **************************/
1079
1080             r22              = _mm256_mul_pd(rsq22,rinv22);
1081             r22              = _mm256_andnot_pd(dummy_mask,r22);
1082
1083             /* Calculate table index by multiplying r with table scale and truncate to integer */
1084             rt               = _mm256_mul_pd(r22,vftabscale);
1085             vfitab           = _mm256_cvttpd_epi32(rt);
1086             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1087             vfitab           = _mm_slli_epi32(vfitab,2);
1088
1089             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1090             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1091             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1092             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1093             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1094             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1095             Heps             = _mm256_mul_pd(vfeps,H);
1096             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1097             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1098             velec            = _mm256_mul_pd(qq22,VV);
1099             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1100             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
1101
1102             /* Update potential sum for this i atom from the interaction with this j atom. */
1103             velec            = _mm256_andnot_pd(dummy_mask,velec);
1104             velecsum         = _mm256_add_pd(velecsum,velec);
1105
1106             fscal            = felec;
1107
1108             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1109
1110             /* Calculate temporary vectorial force */
1111             tx               = _mm256_mul_pd(fscal,dx22);
1112             ty               = _mm256_mul_pd(fscal,dy22);
1113             tz               = _mm256_mul_pd(fscal,dz22);
1114
1115             /* Update vectorial force */
1116             fix2             = _mm256_add_pd(fix2,tx);
1117             fiy2             = _mm256_add_pd(fiy2,ty);
1118             fiz2             = _mm256_add_pd(fiz2,tz);
1119
1120             fjx2             = _mm256_add_pd(fjx2,tx);
1121             fjy2             = _mm256_add_pd(fjy2,ty);
1122             fjz2             = _mm256_add_pd(fjz2,tz);
1123
1124             /**************************
1125              * CALCULATE INTERACTIONS *
1126              **************************/
1127
1128             r23              = _mm256_mul_pd(rsq23,rinv23);
1129             r23              = _mm256_andnot_pd(dummy_mask,r23);
1130
1131             /* Calculate table index by multiplying r with table scale and truncate to integer */
1132             rt               = _mm256_mul_pd(r23,vftabscale);
1133             vfitab           = _mm256_cvttpd_epi32(rt);
1134             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1135             vfitab           = _mm_slli_epi32(vfitab,2);
1136
1137             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1138             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1139             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1140             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1141             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1142             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1143             Heps             = _mm256_mul_pd(vfeps,H);
1144             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1145             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1146             velec            = _mm256_mul_pd(qq23,VV);
1147             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1148             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
1149
1150             /* Update potential sum for this i atom from the interaction with this j atom. */
1151             velec            = _mm256_andnot_pd(dummy_mask,velec);
1152             velecsum         = _mm256_add_pd(velecsum,velec);
1153
1154             fscal            = felec;
1155
1156             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1157
1158             /* Calculate temporary vectorial force */
1159             tx               = _mm256_mul_pd(fscal,dx23);
1160             ty               = _mm256_mul_pd(fscal,dy23);
1161             tz               = _mm256_mul_pd(fscal,dz23);
1162
1163             /* Update vectorial force */
1164             fix2             = _mm256_add_pd(fix2,tx);
1165             fiy2             = _mm256_add_pd(fiy2,ty);
1166             fiz2             = _mm256_add_pd(fiz2,tz);
1167
1168             fjx3             = _mm256_add_pd(fjx3,tx);
1169             fjy3             = _mm256_add_pd(fjy3,ty);
1170             fjz3             = _mm256_add_pd(fjz3,tz);
1171
1172             /**************************
1173              * CALCULATE INTERACTIONS *
1174              **************************/
1175
1176             r31              = _mm256_mul_pd(rsq31,rinv31);
1177             r31              = _mm256_andnot_pd(dummy_mask,r31);
1178
1179             /* Calculate table index by multiplying r with table scale and truncate to integer */
1180             rt               = _mm256_mul_pd(r31,vftabscale);
1181             vfitab           = _mm256_cvttpd_epi32(rt);
1182             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1183             vfitab           = _mm_slli_epi32(vfitab,2);
1184
1185             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1186             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1187             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1188             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1189             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1190             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1191             Heps             = _mm256_mul_pd(vfeps,H);
1192             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1193             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1194             velec            = _mm256_mul_pd(qq31,VV);
1195             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1196             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
1197
1198             /* Update potential sum for this i atom from the interaction with this j atom. */
1199             velec            = _mm256_andnot_pd(dummy_mask,velec);
1200             velecsum         = _mm256_add_pd(velecsum,velec);
1201
1202             fscal            = felec;
1203
1204             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1205
1206             /* Calculate temporary vectorial force */
1207             tx               = _mm256_mul_pd(fscal,dx31);
1208             ty               = _mm256_mul_pd(fscal,dy31);
1209             tz               = _mm256_mul_pd(fscal,dz31);
1210
1211             /* Update vectorial force */
1212             fix3             = _mm256_add_pd(fix3,tx);
1213             fiy3             = _mm256_add_pd(fiy3,ty);
1214             fiz3             = _mm256_add_pd(fiz3,tz);
1215
1216             fjx1             = _mm256_add_pd(fjx1,tx);
1217             fjy1             = _mm256_add_pd(fjy1,ty);
1218             fjz1             = _mm256_add_pd(fjz1,tz);
1219
1220             /**************************
1221              * CALCULATE INTERACTIONS *
1222              **************************/
1223
1224             r32              = _mm256_mul_pd(rsq32,rinv32);
1225             r32              = _mm256_andnot_pd(dummy_mask,r32);
1226
1227             /* Calculate table index by multiplying r with table scale and truncate to integer */
1228             rt               = _mm256_mul_pd(r32,vftabscale);
1229             vfitab           = _mm256_cvttpd_epi32(rt);
1230             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1231             vfitab           = _mm_slli_epi32(vfitab,2);
1232
1233             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1234             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1235             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1236             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1237             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1238             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1239             Heps             = _mm256_mul_pd(vfeps,H);
1240             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1241             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1242             velec            = _mm256_mul_pd(qq32,VV);
1243             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1244             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
1245
1246             /* Update potential sum for this i atom from the interaction with this j atom. */
1247             velec            = _mm256_andnot_pd(dummy_mask,velec);
1248             velecsum         = _mm256_add_pd(velecsum,velec);
1249
1250             fscal            = felec;
1251
1252             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1253
1254             /* Calculate temporary vectorial force */
1255             tx               = _mm256_mul_pd(fscal,dx32);
1256             ty               = _mm256_mul_pd(fscal,dy32);
1257             tz               = _mm256_mul_pd(fscal,dz32);
1258
1259             /* Update vectorial force */
1260             fix3             = _mm256_add_pd(fix3,tx);
1261             fiy3             = _mm256_add_pd(fiy3,ty);
1262             fiz3             = _mm256_add_pd(fiz3,tz);
1263
1264             fjx2             = _mm256_add_pd(fjx2,tx);
1265             fjy2             = _mm256_add_pd(fjy2,ty);
1266             fjz2             = _mm256_add_pd(fjz2,tz);
1267
1268             /**************************
1269              * CALCULATE INTERACTIONS *
1270              **************************/
1271
1272             r33              = _mm256_mul_pd(rsq33,rinv33);
1273             r33              = _mm256_andnot_pd(dummy_mask,r33);
1274
1275             /* Calculate table index by multiplying r with table scale and truncate to integer */
1276             rt               = _mm256_mul_pd(r33,vftabscale);
1277             vfitab           = _mm256_cvttpd_epi32(rt);
1278             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1279             vfitab           = _mm_slli_epi32(vfitab,2);
1280
1281             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1282             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1283             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1284             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1285             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1286             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1287             Heps             = _mm256_mul_pd(vfeps,H);
1288             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1289             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
1290             velec            = _mm256_mul_pd(qq33,VV);
1291             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1292             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
1293
1294             /* Update potential sum for this i atom from the interaction with this j atom. */
1295             velec            = _mm256_andnot_pd(dummy_mask,velec);
1296             velecsum         = _mm256_add_pd(velecsum,velec);
1297
1298             fscal            = felec;
1299
1300             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1301
1302             /* Calculate temporary vectorial force */
1303             tx               = _mm256_mul_pd(fscal,dx33);
1304             ty               = _mm256_mul_pd(fscal,dy33);
1305             tz               = _mm256_mul_pd(fscal,dz33);
1306
1307             /* Update vectorial force */
1308             fix3             = _mm256_add_pd(fix3,tx);
1309             fiy3             = _mm256_add_pd(fiy3,ty);
1310             fiz3             = _mm256_add_pd(fiz3,tz);
1311
1312             fjx3             = _mm256_add_pd(fjx3,tx);
1313             fjy3             = _mm256_add_pd(fjy3,ty);
1314             fjz3             = _mm256_add_pd(fjz3,tz);
1315
1316             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1317             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1318             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1319             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1320
1321             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1322                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1323                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1324
1325             /* Inner loop uses 431 flops */
1326         }
1327
1328         /* End of innermost loop */
1329
1330         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1331                                                  f+i_coord_offset,fshift+i_shift_offset);
1332
1333         ggid                        = gid[iidx];
1334         /* Update potential energies */
1335         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1336         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1337
1338         /* Increment number of inner iterations */
1339         inneriter                  += j_index_end - j_index_start;
1340
1341         /* Outer loop uses 26 flops */
1342     }
1343
1344     /* Increment number of outer iterations */
1345     outeriter        += nri;
1346
1347     /* Update outer/inner flops */
1348
1349     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*431);
1350 }
1351 /*
1352  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_avx_256_double
1353  * Electrostatics interaction: CubicSplineTable
1354  * VdW interaction:            LennardJones
1355  * Geometry:                   Water4-Water4
1356  * Calculate force/pot:        Force
1357  */
1358 void
1359 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_avx_256_double
1360                     (t_nblist                    * gmx_restrict       nlist,
1361                      rvec                        * gmx_restrict          xx,
1362                      rvec                        * gmx_restrict          ff,
1363                      t_forcerec                  * gmx_restrict          fr,
1364                      t_mdatoms                   * gmx_restrict     mdatoms,
1365                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1366                      t_nrnb                      * gmx_restrict        nrnb)
1367 {
1368     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1369      * just 0 for non-waters.
1370      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1371      * jnr indices corresponding to data put in the four positions in the SIMD register.
1372      */
1373     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1374     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1375     int              jnrA,jnrB,jnrC,jnrD;
1376     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1377     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1378     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1379     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1380     real             rcutoff_scalar;
1381     real             *shiftvec,*fshift,*x,*f;
1382     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1383     real             scratch[4*DIM];
1384     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1385     real *           vdwioffsetptr0;
1386     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1387     real *           vdwioffsetptr1;
1388     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1389     real *           vdwioffsetptr2;
1390     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1391     real *           vdwioffsetptr3;
1392     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1393     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1394     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1395     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1396     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1397     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1398     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1399     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1400     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1401     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1402     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1403     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1404     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1405     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1406     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1407     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1408     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1409     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1410     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1411     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1412     real             *charge;
1413     int              nvdwtype;
1414     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1415     int              *vdwtype;
1416     real             *vdwparam;
1417     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1418     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1419     __m128i          vfitab;
1420     __m128i          ifour       = _mm_set1_epi32(4);
1421     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1422     real             *vftab;
1423     __m256d          dummy_mask,cutoff_mask;
1424     __m128           tmpmask0,tmpmask1;
1425     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1426     __m256d          one     = _mm256_set1_pd(1.0);
1427     __m256d          two     = _mm256_set1_pd(2.0);
1428     x                = xx[0];
1429     f                = ff[0];
1430
1431     nri              = nlist->nri;
1432     iinr             = nlist->iinr;
1433     jindex           = nlist->jindex;
1434     jjnr             = nlist->jjnr;
1435     shiftidx         = nlist->shift;
1436     gid              = nlist->gid;
1437     shiftvec         = fr->shift_vec[0];
1438     fshift           = fr->fshift[0];
1439     facel            = _mm256_set1_pd(fr->epsfac);
1440     charge           = mdatoms->chargeA;
1441     nvdwtype         = fr->ntype;
1442     vdwparam         = fr->nbfp;
1443     vdwtype          = mdatoms->typeA;
1444
1445     vftab            = kernel_data->table_elec->data;
1446     vftabscale       = _mm256_set1_pd(kernel_data->table_elec->scale);
1447
1448     /* Setup water-specific parameters */
1449     inr              = nlist->iinr[0];
1450     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1451     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1452     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1453     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1454
1455     jq1              = _mm256_set1_pd(charge[inr+1]);
1456     jq2              = _mm256_set1_pd(charge[inr+2]);
1457     jq3              = _mm256_set1_pd(charge[inr+3]);
1458     vdwjidx0A        = 2*vdwtype[inr+0];
1459     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1460     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1461     qq11             = _mm256_mul_pd(iq1,jq1);
1462     qq12             = _mm256_mul_pd(iq1,jq2);
1463     qq13             = _mm256_mul_pd(iq1,jq3);
1464     qq21             = _mm256_mul_pd(iq2,jq1);
1465     qq22             = _mm256_mul_pd(iq2,jq2);
1466     qq23             = _mm256_mul_pd(iq2,jq3);
1467     qq31             = _mm256_mul_pd(iq3,jq1);
1468     qq32             = _mm256_mul_pd(iq3,jq2);
1469     qq33             = _mm256_mul_pd(iq3,jq3);
1470
1471     /* Avoid stupid compiler warnings */
1472     jnrA = jnrB = jnrC = jnrD = 0;
1473     j_coord_offsetA = 0;
1474     j_coord_offsetB = 0;
1475     j_coord_offsetC = 0;
1476     j_coord_offsetD = 0;
1477
1478     outeriter        = 0;
1479     inneriter        = 0;
1480
1481     for(iidx=0;iidx<4*DIM;iidx++)
1482     {
1483         scratch[iidx] = 0.0;
1484     }
1485
1486     /* Start outer loop over neighborlists */
1487     for(iidx=0; iidx<nri; iidx++)
1488     {
1489         /* Load shift vector for this list */
1490         i_shift_offset   = DIM*shiftidx[iidx];
1491
1492         /* Load limits for loop over neighbors */
1493         j_index_start    = jindex[iidx];
1494         j_index_end      = jindex[iidx+1];
1495
1496         /* Get outer coordinate index */
1497         inr              = iinr[iidx];
1498         i_coord_offset   = DIM*inr;
1499
1500         /* Load i particle coords and add shift vector */
1501         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1502                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1503
1504         fix0             = _mm256_setzero_pd();
1505         fiy0             = _mm256_setzero_pd();
1506         fiz0             = _mm256_setzero_pd();
1507         fix1             = _mm256_setzero_pd();
1508         fiy1             = _mm256_setzero_pd();
1509         fiz1             = _mm256_setzero_pd();
1510         fix2             = _mm256_setzero_pd();
1511         fiy2             = _mm256_setzero_pd();
1512         fiz2             = _mm256_setzero_pd();
1513         fix3             = _mm256_setzero_pd();
1514         fiy3             = _mm256_setzero_pd();
1515         fiz3             = _mm256_setzero_pd();
1516
1517         /* Start inner kernel loop */
1518         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1519         {
1520
1521             /* Get j neighbor index, and coordinate index */
1522             jnrA             = jjnr[jidx];
1523             jnrB             = jjnr[jidx+1];
1524             jnrC             = jjnr[jidx+2];
1525             jnrD             = jjnr[jidx+3];
1526             j_coord_offsetA  = DIM*jnrA;
1527             j_coord_offsetB  = DIM*jnrB;
1528             j_coord_offsetC  = DIM*jnrC;
1529             j_coord_offsetD  = DIM*jnrD;
1530
1531             /* load j atom coordinates */
1532             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1533                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1534                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1535                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1536
1537             /* Calculate displacement vector */
1538             dx00             = _mm256_sub_pd(ix0,jx0);
1539             dy00             = _mm256_sub_pd(iy0,jy0);
1540             dz00             = _mm256_sub_pd(iz0,jz0);
1541             dx11             = _mm256_sub_pd(ix1,jx1);
1542             dy11             = _mm256_sub_pd(iy1,jy1);
1543             dz11             = _mm256_sub_pd(iz1,jz1);
1544             dx12             = _mm256_sub_pd(ix1,jx2);
1545             dy12             = _mm256_sub_pd(iy1,jy2);
1546             dz12             = _mm256_sub_pd(iz1,jz2);
1547             dx13             = _mm256_sub_pd(ix1,jx3);
1548             dy13             = _mm256_sub_pd(iy1,jy3);
1549             dz13             = _mm256_sub_pd(iz1,jz3);
1550             dx21             = _mm256_sub_pd(ix2,jx1);
1551             dy21             = _mm256_sub_pd(iy2,jy1);
1552             dz21             = _mm256_sub_pd(iz2,jz1);
1553             dx22             = _mm256_sub_pd(ix2,jx2);
1554             dy22             = _mm256_sub_pd(iy2,jy2);
1555             dz22             = _mm256_sub_pd(iz2,jz2);
1556             dx23             = _mm256_sub_pd(ix2,jx3);
1557             dy23             = _mm256_sub_pd(iy2,jy3);
1558             dz23             = _mm256_sub_pd(iz2,jz3);
1559             dx31             = _mm256_sub_pd(ix3,jx1);
1560             dy31             = _mm256_sub_pd(iy3,jy1);
1561             dz31             = _mm256_sub_pd(iz3,jz1);
1562             dx32             = _mm256_sub_pd(ix3,jx2);
1563             dy32             = _mm256_sub_pd(iy3,jy2);
1564             dz32             = _mm256_sub_pd(iz3,jz2);
1565             dx33             = _mm256_sub_pd(ix3,jx3);
1566             dy33             = _mm256_sub_pd(iy3,jy3);
1567             dz33             = _mm256_sub_pd(iz3,jz3);
1568
1569             /* Calculate squared distance and things based on it */
1570             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1571             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1572             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1573             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1574             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1575             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1576             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1577             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1578             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1579             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1580
1581             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1582             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1583             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1584             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1585             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1586             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1587             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1588             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1589             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1590
1591             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1592
1593             fjx0             = _mm256_setzero_pd();
1594             fjy0             = _mm256_setzero_pd();
1595             fjz0             = _mm256_setzero_pd();
1596             fjx1             = _mm256_setzero_pd();
1597             fjy1             = _mm256_setzero_pd();
1598             fjz1             = _mm256_setzero_pd();
1599             fjx2             = _mm256_setzero_pd();
1600             fjy2             = _mm256_setzero_pd();
1601             fjz2             = _mm256_setzero_pd();
1602             fjx3             = _mm256_setzero_pd();
1603             fjy3             = _mm256_setzero_pd();
1604             fjz3             = _mm256_setzero_pd();
1605
1606             /**************************
1607              * CALCULATE INTERACTIONS *
1608              **************************/
1609
1610             /* LENNARD-JONES DISPERSION/REPULSION */
1611
1612             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1613             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1614
1615             fscal            = fvdw;
1616
1617             /* Calculate temporary vectorial force */
1618             tx               = _mm256_mul_pd(fscal,dx00);
1619             ty               = _mm256_mul_pd(fscal,dy00);
1620             tz               = _mm256_mul_pd(fscal,dz00);
1621
1622             /* Update vectorial force */
1623             fix0             = _mm256_add_pd(fix0,tx);
1624             fiy0             = _mm256_add_pd(fiy0,ty);
1625             fiz0             = _mm256_add_pd(fiz0,tz);
1626
1627             fjx0             = _mm256_add_pd(fjx0,tx);
1628             fjy0             = _mm256_add_pd(fjy0,ty);
1629             fjz0             = _mm256_add_pd(fjz0,tz);
1630
1631             /**************************
1632              * CALCULATE INTERACTIONS *
1633              **************************/
1634
1635             r11              = _mm256_mul_pd(rsq11,rinv11);
1636
1637             /* Calculate table index by multiplying r with table scale and truncate to integer */
1638             rt               = _mm256_mul_pd(r11,vftabscale);
1639             vfitab           = _mm256_cvttpd_epi32(rt);
1640             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1641             vfitab           = _mm_slli_epi32(vfitab,2);
1642
1643             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1644             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1645             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1646             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1647             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1648             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1649             Heps             = _mm256_mul_pd(vfeps,H);
1650             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1651             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1652             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
1653
1654             fscal            = felec;
1655
1656             /* Calculate temporary vectorial force */
1657             tx               = _mm256_mul_pd(fscal,dx11);
1658             ty               = _mm256_mul_pd(fscal,dy11);
1659             tz               = _mm256_mul_pd(fscal,dz11);
1660
1661             /* Update vectorial force */
1662             fix1             = _mm256_add_pd(fix1,tx);
1663             fiy1             = _mm256_add_pd(fiy1,ty);
1664             fiz1             = _mm256_add_pd(fiz1,tz);
1665
1666             fjx1             = _mm256_add_pd(fjx1,tx);
1667             fjy1             = _mm256_add_pd(fjy1,ty);
1668             fjz1             = _mm256_add_pd(fjz1,tz);
1669
1670             /**************************
1671              * CALCULATE INTERACTIONS *
1672              **************************/
1673
1674             r12              = _mm256_mul_pd(rsq12,rinv12);
1675
1676             /* Calculate table index by multiplying r with table scale and truncate to integer */
1677             rt               = _mm256_mul_pd(r12,vftabscale);
1678             vfitab           = _mm256_cvttpd_epi32(rt);
1679             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1680             vfitab           = _mm_slli_epi32(vfitab,2);
1681
1682             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1683             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1684             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1685             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1686             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1687             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1688             Heps             = _mm256_mul_pd(vfeps,H);
1689             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1690             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1691             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
1692
1693             fscal            = felec;
1694
1695             /* Calculate temporary vectorial force */
1696             tx               = _mm256_mul_pd(fscal,dx12);
1697             ty               = _mm256_mul_pd(fscal,dy12);
1698             tz               = _mm256_mul_pd(fscal,dz12);
1699
1700             /* Update vectorial force */
1701             fix1             = _mm256_add_pd(fix1,tx);
1702             fiy1             = _mm256_add_pd(fiy1,ty);
1703             fiz1             = _mm256_add_pd(fiz1,tz);
1704
1705             fjx2             = _mm256_add_pd(fjx2,tx);
1706             fjy2             = _mm256_add_pd(fjy2,ty);
1707             fjz2             = _mm256_add_pd(fjz2,tz);
1708
1709             /**************************
1710              * CALCULATE INTERACTIONS *
1711              **************************/
1712
1713             r13              = _mm256_mul_pd(rsq13,rinv13);
1714
1715             /* Calculate table index by multiplying r with table scale and truncate to integer */
1716             rt               = _mm256_mul_pd(r13,vftabscale);
1717             vfitab           = _mm256_cvttpd_epi32(rt);
1718             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1719             vfitab           = _mm_slli_epi32(vfitab,2);
1720
1721             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1722             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1723             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1724             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1725             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1726             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1727             Heps             = _mm256_mul_pd(vfeps,H);
1728             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1729             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1730             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
1731
1732             fscal            = felec;
1733
1734             /* Calculate temporary vectorial force */
1735             tx               = _mm256_mul_pd(fscal,dx13);
1736             ty               = _mm256_mul_pd(fscal,dy13);
1737             tz               = _mm256_mul_pd(fscal,dz13);
1738
1739             /* Update vectorial force */
1740             fix1             = _mm256_add_pd(fix1,tx);
1741             fiy1             = _mm256_add_pd(fiy1,ty);
1742             fiz1             = _mm256_add_pd(fiz1,tz);
1743
1744             fjx3             = _mm256_add_pd(fjx3,tx);
1745             fjy3             = _mm256_add_pd(fjy3,ty);
1746             fjz3             = _mm256_add_pd(fjz3,tz);
1747
1748             /**************************
1749              * CALCULATE INTERACTIONS *
1750              **************************/
1751
1752             r21              = _mm256_mul_pd(rsq21,rinv21);
1753
1754             /* Calculate table index by multiplying r with table scale and truncate to integer */
1755             rt               = _mm256_mul_pd(r21,vftabscale);
1756             vfitab           = _mm256_cvttpd_epi32(rt);
1757             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1758             vfitab           = _mm_slli_epi32(vfitab,2);
1759
1760             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1761             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1762             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1763             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1764             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1765             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1766             Heps             = _mm256_mul_pd(vfeps,H);
1767             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1768             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1769             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
1770
1771             fscal            = felec;
1772
1773             /* Calculate temporary vectorial force */
1774             tx               = _mm256_mul_pd(fscal,dx21);
1775             ty               = _mm256_mul_pd(fscal,dy21);
1776             tz               = _mm256_mul_pd(fscal,dz21);
1777
1778             /* Update vectorial force */
1779             fix2             = _mm256_add_pd(fix2,tx);
1780             fiy2             = _mm256_add_pd(fiy2,ty);
1781             fiz2             = _mm256_add_pd(fiz2,tz);
1782
1783             fjx1             = _mm256_add_pd(fjx1,tx);
1784             fjy1             = _mm256_add_pd(fjy1,ty);
1785             fjz1             = _mm256_add_pd(fjz1,tz);
1786
1787             /**************************
1788              * CALCULATE INTERACTIONS *
1789              **************************/
1790
1791             r22              = _mm256_mul_pd(rsq22,rinv22);
1792
1793             /* Calculate table index by multiplying r with table scale and truncate to integer */
1794             rt               = _mm256_mul_pd(r22,vftabscale);
1795             vfitab           = _mm256_cvttpd_epi32(rt);
1796             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1797             vfitab           = _mm_slli_epi32(vfitab,2);
1798
1799             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1800             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1801             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1802             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1803             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1804             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1805             Heps             = _mm256_mul_pd(vfeps,H);
1806             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1807             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1808             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
1809
1810             fscal            = felec;
1811
1812             /* Calculate temporary vectorial force */
1813             tx               = _mm256_mul_pd(fscal,dx22);
1814             ty               = _mm256_mul_pd(fscal,dy22);
1815             tz               = _mm256_mul_pd(fscal,dz22);
1816
1817             /* Update vectorial force */
1818             fix2             = _mm256_add_pd(fix2,tx);
1819             fiy2             = _mm256_add_pd(fiy2,ty);
1820             fiz2             = _mm256_add_pd(fiz2,tz);
1821
1822             fjx2             = _mm256_add_pd(fjx2,tx);
1823             fjy2             = _mm256_add_pd(fjy2,ty);
1824             fjz2             = _mm256_add_pd(fjz2,tz);
1825
1826             /**************************
1827              * CALCULATE INTERACTIONS *
1828              **************************/
1829
1830             r23              = _mm256_mul_pd(rsq23,rinv23);
1831
1832             /* Calculate table index by multiplying r with table scale and truncate to integer */
1833             rt               = _mm256_mul_pd(r23,vftabscale);
1834             vfitab           = _mm256_cvttpd_epi32(rt);
1835             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1836             vfitab           = _mm_slli_epi32(vfitab,2);
1837
1838             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1839             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1840             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1841             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1842             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1843             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1844             Heps             = _mm256_mul_pd(vfeps,H);
1845             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1846             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1847             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
1848
1849             fscal            = felec;
1850
1851             /* Calculate temporary vectorial force */
1852             tx               = _mm256_mul_pd(fscal,dx23);
1853             ty               = _mm256_mul_pd(fscal,dy23);
1854             tz               = _mm256_mul_pd(fscal,dz23);
1855
1856             /* Update vectorial force */
1857             fix2             = _mm256_add_pd(fix2,tx);
1858             fiy2             = _mm256_add_pd(fiy2,ty);
1859             fiz2             = _mm256_add_pd(fiz2,tz);
1860
1861             fjx3             = _mm256_add_pd(fjx3,tx);
1862             fjy3             = _mm256_add_pd(fjy3,ty);
1863             fjz3             = _mm256_add_pd(fjz3,tz);
1864
1865             /**************************
1866              * CALCULATE INTERACTIONS *
1867              **************************/
1868
1869             r31              = _mm256_mul_pd(rsq31,rinv31);
1870
1871             /* Calculate table index by multiplying r with table scale and truncate to integer */
1872             rt               = _mm256_mul_pd(r31,vftabscale);
1873             vfitab           = _mm256_cvttpd_epi32(rt);
1874             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1875             vfitab           = _mm_slli_epi32(vfitab,2);
1876
1877             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1878             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1879             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1880             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1881             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1882             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1883             Heps             = _mm256_mul_pd(vfeps,H);
1884             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1885             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1886             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
1887
1888             fscal            = felec;
1889
1890             /* Calculate temporary vectorial force */
1891             tx               = _mm256_mul_pd(fscal,dx31);
1892             ty               = _mm256_mul_pd(fscal,dy31);
1893             tz               = _mm256_mul_pd(fscal,dz31);
1894
1895             /* Update vectorial force */
1896             fix3             = _mm256_add_pd(fix3,tx);
1897             fiy3             = _mm256_add_pd(fiy3,ty);
1898             fiz3             = _mm256_add_pd(fiz3,tz);
1899
1900             fjx1             = _mm256_add_pd(fjx1,tx);
1901             fjy1             = _mm256_add_pd(fjy1,ty);
1902             fjz1             = _mm256_add_pd(fjz1,tz);
1903
1904             /**************************
1905              * CALCULATE INTERACTIONS *
1906              **************************/
1907
1908             r32              = _mm256_mul_pd(rsq32,rinv32);
1909
1910             /* Calculate table index by multiplying r with table scale and truncate to integer */
1911             rt               = _mm256_mul_pd(r32,vftabscale);
1912             vfitab           = _mm256_cvttpd_epi32(rt);
1913             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1914             vfitab           = _mm_slli_epi32(vfitab,2);
1915
1916             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1917             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1918             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1919             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1920             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1921             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1922             Heps             = _mm256_mul_pd(vfeps,H);
1923             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1924             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1925             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
1926
1927             fscal            = felec;
1928
1929             /* Calculate temporary vectorial force */
1930             tx               = _mm256_mul_pd(fscal,dx32);
1931             ty               = _mm256_mul_pd(fscal,dy32);
1932             tz               = _mm256_mul_pd(fscal,dz32);
1933
1934             /* Update vectorial force */
1935             fix3             = _mm256_add_pd(fix3,tx);
1936             fiy3             = _mm256_add_pd(fiy3,ty);
1937             fiz3             = _mm256_add_pd(fiz3,tz);
1938
1939             fjx2             = _mm256_add_pd(fjx2,tx);
1940             fjy2             = _mm256_add_pd(fjy2,ty);
1941             fjz2             = _mm256_add_pd(fjz2,tz);
1942
1943             /**************************
1944              * CALCULATE INTERACTIONS *
1945              **************************/
1946
1947             r33              = _mm256_mul_pd(rsq33,rinv33);
1948
1949             /* Calculate table index by multiplying r with table scale and truncate to integer */
1950             rt               = _mm256_mul_pd(r33,vftabscale);
1951             vfitab           = _mm256_cvttpd_epi32(rt);
1952             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1953             vfitab           = _mm_slli_epi32(vfitab,2);
1954
1955             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1956             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1957             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1958             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1959             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1960             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1961             Heps             = _mm256_mul_pd(vfeps,H);
1962             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1963             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1964             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
1965
1966             fscal            = felec;
1967
1968             /* Calculate temporary vectorial force */
1969             tx               = _mm256_mul_pd(fscal,dx33);
1970             ty               = _mm256_mul_pd(fscal,dy33);
1971             tz               = _mm256_mul_pd(fscal,dz33);
1972
1973             /* Update vectorial force */
1974             fix3             = _mm256_add_pd(fix3,tx);
1975             fiy3             = _mm256_add_pd(fiy3,ty);
1976             fiz3             = _mm256_add_pd(fiz3,tz);
1977
1978             fjx3             = _mm256_add_pd(fjx3,tx);
1979             fjy3             = _mm256_add_pd(fjy3,ty);
1980             fjz3             = _mm256_add_pd(fjz3,tz);
1981
1982             fjptrA             = f+j_coord_offsetA;
1983             fjptrB             = f+j_coord_offsetB;
1984             fjptrC             = f+j_coord_offsetC;
1985             fjptrD             = f+j_coord_offsetD;
1986
1987             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1988                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1989                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1990
1991             /* Inner loop uses 381 flops */
1992         }
1993
1994         if(jidx<j_index_end)
1995         {
1996
1997             /* Get j neighbor index, and coordinate index */
1998             jnrlistA         = jjnr[jidx];
1999             jnrlistB         = jjnr[jidx+1];
2000             jnrlistC         = jjnr[jidx+2];
2001             jnrlistD         = jjnr[jidx+3];
2002             /* Sign of each element will be negative for non-real atoms.
2003              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2004              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
2005              */
2006             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2007
2008             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
2009             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
2010             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
2011
2012             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2013             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2014             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2015             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2016             j_coord_offsetA  = DIM*jnrA;
2017             j_coord_offsetB  = DIM*jnrB;
2018             j_coord_offsetC  = DIM*jnrC;
2019             j_coord_offsetD  = DIM*jnrD;
2020
2021             /* load j atom coordinates */
2022             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
2023                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2024                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2025                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
2026
2027             /* Calculate displacement vector */
2028             dx00             = _mm256_sub_pd(ix0,jx0);
2029             dy00             = _mm256_sub_pd(iy0,jy0);
2030             dz00             = _mm256_sub_pd(iz0,jz0);
2031             dx11             = _mm256_sub_pd(ix1,jx1);
2032             dy11             = _mm256_sub_pd(iy1,jy1);
2033             dz11             = _mm256_sub_pd(iz1,jz1);
2034             dx12             = _mm256_sub_pd(ix1,jx2);
2035             dy12             = _mm256_sub_pd(iy1,jy2);
2036             dz12             = _mm256_sub_pd(iz1,jz2);
2037             dx13             = _mm256_sub_pd(ix1,jx3);
2038             dy13             = _mm256_sub_pd(iy1,jy3);
2039             dz13             = _mm256_sub_pd(iz1,jz3);
2040             dx21             = _mm256_sub_pd(ix2,jx1);
2041             dy21             = _mm256_sub_pd(iy2,jy1);
2042             dz21             = _mm256_sub_pd(iz2,jz1);
2043             dx22             = _mm256_sub_pd(ix2,jx2);
2044             dy22             = _mm256_sub_pd(iy2,jy2);
2045             dz22             = _mm256_sub_pd(iz2,jz2);
2046             dx23             = _mm256_sub_pd(ix2,jx3);
2047             dy23             = _mm256_sub_pd(iy2,jy3);
2048             dz23             = _mm256_sub_pd(iz2,jz3);
2049             dx31             = _mm256_sub_pd(ix3,jx1);
2050             dy31             = _mm256_sub_pd(iy3,jy1);
2051             dz31             = _mm256_sub_pd(iz3,jz1);
2052             dx32             = _mm256_sub_pd(ix3,jx2);
2053             dy32             = _mm256_sub_pd(iy3,jy2);
2054             dz32             = _mm256_sub_pd(iz3,jz2);
2055             dx33             = _mm256_sub_pd(ix3,jx3);
2056             dy33             = _mm256_sub_pd(iy3,jy3);
2057             dz33             = _mm256_sub_pd(iz3,jz3);
2058
2059             /* Calculate squared distance and things based on it */
2060             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2061             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2062             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2063             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
2064             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2065             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2066             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
2067             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
2068             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
2069             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
2070
2071             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
2072             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
2073             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
2074             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
2075             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
2076             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
2077             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
2078             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
2079             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
2080
2081             rinvsq00         = gmx_mm256_inv_pd(rsq00);
2082
2083             fjx0             = _mm256_setzero_pd();
2084             fjy0             = _mm256_setzero_pd();
2085             fjz0             = _mm256_setzero_pd();
2086             fjx1             = _mm256_setzero_pd();
2087             fjy1             = _mm256_setzero_pd();
2088             fjz1             = _mm256_setzero_pd();
2089             fjx2             = _mm256_setzero_pd();
2090             fjy2             = _mm256_setzero_pd();
2091             fjz2             = _mm256_setzero_pd();
2092             fjx3             = _mm256_setzero_pd();
2093             fjy3             = _mm256_setzero_pd();
2094             fjz3             = _mm256_setzero_pd();
2095
2096             /**************************
2097              * CALCULATE INTERACTIONS *
2098              **************************/
2099
2100             /* LENNARD-JONES DISPERSION/REPULSION */
2101
2102             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2103             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2104
2105             fscal            = fvdw;
2106
2107             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2108
2109             /* Calculate temporary vectorial force */
2110             tx               = _mm256_mul_pd(fscal,dx00);
2111             ty               = _mm256_mul_pd(fscal,dy00);
2112             tz               = _mm256_mul_pd(fscal,dz00);
2113
2114             /* Update vectorial force */
2115             fix0             = _mm256_add_pd(fix0,tx);
2116             fiy0             = _mm256_add_pd(fiy0,ty);
2117             fiz0             = _mm256_add_pd(fiz0,tz);
2118
2119             fjx0             = _mm256_add_pd(fjx0,tx);
2120             fjy0             = _mm256_add_pd(fjy0,ty);
2121             fjz0             = _mm256_add_pd(fjz0,tz);
2122
2123             /**************************
2124              * CALCULATE INTERACTIONS *
2125              **************************/
2126
2127             r11              = _mm256_mul_pd(rsq11,rinv11);
2128             r11              = _mm256_andnot_pd(dummy_mask,r11);
2129
2130             /* Calculate table index by multiplying r with table scale and truncate to integer */
2131             rt               = _mm256_mul_pd(r11,vftabscale);
2132             vfitab           = _mm256_cvttpd_epi32(rt);
2133             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2134             vfitab           = _mm_slli_epi32(vfitab,2);
2135
2136             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2137             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2138             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2139             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2140             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2141             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2142             Heps             = _mm256_mul_pd(vfeps,H);
2143             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2144             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2145             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq11,FF),_mm256_mul_pd(vftabscale,rinv11)));
2146
2147             fscal            = felec;
2148
2149             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2150
2151             /* Calculate temporary vectorial force */
2152             tx               = _mm256_mul_pd(fscal,dx11);
2153             ty               = _mm256_mul_pd(fscal,dy11);
2154             tz               = _mm256_mul_pd(fscal,dz11);
2155
2156             /* Update vectorial force */
2157             fix1             = _mm256_add_pd(fix1,tx);
2158             fiy1             = _mm256_add_pd(fiy1,ty);
2159             fiz1             = _mm256_add_pd(fiz1,tz);
2160
2161             fjx1             = _mm256_add_pd(fjx1,tx);
2162             fjy1             = _mm256_add_pd(fjy1,ty);
2163             fjz1             = _mm256_add_pd(fjz1,tz);
2164
2165             /**************************
2166              * CALCULATE INTERACTIONS *
2167              **************************/
2168
2169             r12              = _mm256_mul_pd(rsq12,rinv12);
2170             r12              = _mm256_andnot_pd(dummy_mask,r12);
2171
2172             /* Calculate table index by multiplying r with table scale and truncate to integer */
2173             rt               = _mm256_mul_pd(r12,vftabscale);
2174             vfitab           = _mm256_cvttpd_epi32(rt);
2175             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2176             vfitab           = _mm_slli_epi32(vfitab,2);
2177
2178             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2179             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2180             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2181             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2182             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2183             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2184             Heps             = _mm256_mul_pd(vfeps,H);
2185             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2186             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2187             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq12,FF),_mm256_mul_pd(vftabscale,rinv12)));
2188
2189             fscal            = felec;
2190
2191             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2192
2193             /* Calculate temporary vectorial force */
2194             tx               = _mm256_mul_pd(fscal,dx12);
2195             ty               = _mm256_mul_pd(fscal,dy12);
2196             tz               = _mm256_mul_pd(fscal,dz12);
2197
2198             /* Update vectorial force */
2199             fix1             = _mm256_add_pd(fix1,tx);
2200             fiy1             = _mm256_add_pd(fiy1,ty);
2201             fiz1             = _mm256_add_pd(fiz1,tz);
2202
2203             fjx2             = _mm256_add_pd(fjx2,tx);
2204             fjy2             = _mm256_add_pd(fjy2,ty);
2205             fjz2             = _mm256_add_pd(fjz2,tz);
2206
2207             /**************************
2208              * CALCULATE INTERACTIONS *
2209              **************************/
2210
2211             r13              = _mm256_mul_pd(rsq13,rinv13);
2212             r13              = _mm256_andnot_pd(dummy_mask,r13);
2213
2214             /* Calculate table index by multiplying r with table scale and truncate to integer */
2215             rt               = _mm256_mul_pd(r13,vftabscale);
2216             vfitab           = _mm256_cvttpd_epi32(rt);
2217             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2218             vfitab           = _mm_slli_epi32(vfitab,2);
2219
2220             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2221             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2222             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2223             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2224             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2225             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2226             Heps             = _mm256_mul_pd(vfeps,H);
2227             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2228             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2229             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq13,FF),_mm256_mul_pd(vftabscale,rinv13)));
2230
2231             fscal            = felec;
2232
2233             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2234
2235             /* Calculate temporary vectorial force */
2236             tx               = _mm256_mul_pd(fscal,dx13);
2237             ty               = _mm256_mul_pd(fscal,dy13);
2238             tz               = _mm256_mul_pd(fscal,dz13);
2239
2240             /* Update vectorial force */
2241             fix1             = _mm256_add_pd(fix1,tx);
2242             fiy1             = _mm256_add_pd(fiy1,ty);
2243             fiz1             = _mm256_add_pd(fiz1,tz);
2244
2245             fjx3             = _mm256_add_pd(fjx3,tx);
2246             fjy3             = _mm256_add_pd(fjy3,ty);
2247             fjz3             = _mm256_add_pd(fjz3,tz);
2248
2249             /**************************
2250              * CALCULATE INTERACTIONS *
2251              **************************/
2252
2253             r21              = _mm256_mul_pd(rsq21,rinv21);
2254             r21              = _mm256_andnot_pd(dummy_mask,r21);
2255
2256             /* Calculate table index by multiplying r with table scale and truncate to integer */
2257             rt               = _mm256_mul_pd(r21,vftabscale);
2258             vfitab           = _mm256_cvttpd_epi32(rt);
2259             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2260             vfitab           = _mm_slli_epi32(vfitab,2);
2261
2262             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2263             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2264             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2265             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2266             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2267             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2268             Heps             = _mm256_mul_pd(vfeps,H);
2269             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2270             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2271             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq21,FF),_mm256_mul_pd(vftabscale,rinv21)));
2272
2273             fscal            = felec;
2274
2275             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2276
2277             /* Calculate temporary vectorial force */
2278             tx               = _mm256_mul_pd(fscal,dx21);
2279             ty               = _mm256_mul_pd(fscal,dy21);
2280             tz               = _mm256_mul_pd(fscal,dz21);
2281
2282             /* Update vectorial force */
2283             fix2             = _mm256_add_pd(fix2,tx);
2284             fiy2             = _mm256_add_pd(fiy2,ty);
2285             fiz2             = _mm256_add_pd(fiz2,tz);
2286
2287             fjx1             = _mm256_add_pd(fjx1,tx);
2288             fjy1             = _mm256_add_pd(fjy1,ty);
2289             fjz1             = _mm256_add_pd(fjz1,tz);
2290
2291             /**************************
2292              * CALCULATE INTERACTIONS *
2293              **************************/
2294
2295             r22              = _mm256_mul_pd(rsq22,rinv22);
2296             r22              = _mm256_andnot_pd(dummy_mask,r22);
2297
2298             /* Calculate table index by multiplying r with table scale and truncate to integer */
2299             rt               = _mm256_mul_pd(r22,vftabscale);
2300             vfitab           = _mm256_cvttpd_epi32(rt);
2301             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2302             vfitab           = _mm_slli_epi32(vfitab,2);
2303
2304             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2305             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2306             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2307             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2308             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2309             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2310             Heps             = _mm256_mul_pd(vfeps,H);
2311             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2312             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2313             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq22,FF),_mm256_mul_pd(vftabscale,rinv22)));
2314
2315             fscal            = felec;
2316
2317             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2318
2319             /* Calculate temporary vectorial force */
2320             tx               = _mm256_mul_pd(fscal,dx22);
2321             ty               = _mm256_mul_pd(fscal,dy22);
2322             tz               = _mm256_mul_pd(fscal,dz22);
2323
2324             /* Update vectorial force */
2325             fix2             = _mm256_add_pd(fix2,tx);
2326             fiy2             = _mm256_add_pd(fiy2,ty);
2327             fiz2             = _mm256_add_pd(fiz2,tz);
2328
2329             fjx2             = _mm256_add_pd(fjx2,tx);
2330             fjy2             = _mm256_add_pd(fjy2,ty);
2331             fjz2             = _mm256_add_pd(fjz2,tz);
2332
2333             /**************************
2334              * CALCULATE INTERACTIONS *
2335              **************************/
2336
2337             r23              = _mm256_mul_pd(rsq23,rinv23);
2338             r23              = _mm256_andnot_pd(dummy_mask,r23);
2339
2340             /* Calculate table index by multiplying r with table scale and truncate to integer */
2341             rt               = _mm256_mul_pd(r23,vftabscale);
2342             vfitab           = _mm256_cvttpd_epi32(rt);
2343             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2344             vfitab           = _mm_slli_epi32(vfitab,2);
2345
2346             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2347             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2348             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2349             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2350             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2351             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2352             Heps             = _mm256_mul_pd(vfeps,H);
2353             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2354             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2355             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq23,FF),_mm256_mul_pd(vftabscale,rinv23)));
2356
2357             fscal            = felec;
2358
2359             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2360
2361             /* Calculate temporary vectorial force */
2362             tx               = _mm256_mul_pd(fscal,dx23);
2363             ty               = _mm256_mul_pd(fscal,dy23);
2364             tz               = _mm256_mul_pd(fscal,dz23);
2365
2366             /* Update vectorial force */
2367             fix2             = _mm256_add_pd(fix2,tx);
2368             fiy2             = _mm256_add_pd(fiy2,ty);
2369             fiz2             = _mm256_add_pd(fiz2,tz);
2370
2371             fjx3             = _mm256_add_pd(fjx3,tx);
2372             fjy3             = _mm256_add_pd(fjy3,ty);
2373             fjz3             = _mm256_add_pd(fjz3,tz);
2374
2375             /**************************
2376              * CALCULATE INTERACTIONS *
2377              **************************/
2378
2379             r31              = _mm256_mul_pd(rsq31,rinv31);
2380             r31              = _mm256_andnot_pd(dummy_mask,r31);
2381
2382             /* Calculate table index by multiplying r with table scale and truncate to integer */
2383             rt               = _mm256_mul_pd(r31,vftabscale);
2384             vfitab           = _mm256_cvttpd_epi32(rt);
2385             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2386             vfitab           = _mm_slli_epi32(vfitab,2);
2387
2388             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2389             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2390             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2391             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2392             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2393             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2394             Heps             = _mm256_mul_pd(vfeps,H);
2395             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2396             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2397             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq31,FF),_mm256_mul_pd(vftabscale,rinv31)));
2398
2399             fscal            = felec;
2400
2401             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2402
2403             /* Calculate temporary vectorial force */
2404             tx               = _mm256_mul_pd(fscal,dx31);
2405             ty               = _mm256_mul_pd(fscal,dy31);
2406             tz               = _mm256_mul_pd(fscal,dz31);
2407
2408             /* Update vectorial force */
2409             fix3             = _mm256_add_pd(fix3,tx);
2410             fiy3             = _mm256_add_pd(fiy3,ty);
2411             fiz3             = _mm256_add_pd(fiz3,tz);
2412
2413             fjx1             = _mm256_add_pd(fjx1,tx);
2414             fjy1             = _mm256_add_pd(fjy1,ty);
2415             fjz1             = _mm256_add_pd(fjz1,tz);
2416
2417             /**************************
2418              * CALCULATE INTERACTIONS *
2419              **************************/
2420
2421             r32              = _mm256_mul_pd(rsq32,rinv32);
2422             r32              = _mm256_andnot_pd(dummy_mask,r32);
2423
2424             /* Calculate table index by multiplying r with table scale and truncate to integer */
2425             rt               = _mm256_mul_pd(r32,vftabscale);
2426             vfitab           = _mm256_cvttpd_epi32(rt);
2427             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2428             vfitab           = _mm_slli_epi32(vfitab,2);
2429
2430             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2431             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2432             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2433             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2434             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2435             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2436             Heps             = _mm256_mul_pd(vfeps,H);
2437             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2438             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2439             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq32,FF),_mm256_mul_pd(vftabscale,rinv32)));
2440
2441             fscal            = felec;
2442
2443             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2444
2445             /* Calculate temporary vectorial force */
2446             tx               = _mm256_mul_pd(fscal,dx32);
2447             ty               = _mm256_mul_pd(fscal,dy32);
2448             tz               = _mm256_mul_pd(fscal,dz32);
2449
2450             /* Update vectorial force */
2451             fix3             = _mm256_add_pd(fix3,tx);
2452             fiy3             = _mm256_add_pd(fiy3,ty);
2453             fiz3             = _mm256_add_pd(fiz3,tz);
2454
2455             fjx2             = _mm256_add_pd(fjx2,tx);
2456             fjy2             = _mm256_add_pd(fjy2,ty);
2457             fjz2             = _mm256_add_pd(fjz2,tz);
2458
2459             /**************************
2460              * CALCULATE INTERACTIONS *
2461              **************************/
2462
2463             r33              = _mm256_mul_pd(rsq33,rinv33);
2464             r33              = _mm256_andnot_pd(dummy_mask,r33);
2465
2466             /* Calculate table index by multiplying r with table scale and truncate to integer */
2467             rt               = _mm256_mul_pd(r33,vftabscale);
2468             vfitab           = _mm256_cvttpd_epi32(rt);
2469             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
2470             vfitab           = _mm_slli_epi32(vfitab,2);
2471
2472             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2473             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2474             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
2475             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
2476             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
2477             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
2478             Heps             = _mm256_mul_pd(vfeps,H);
2479             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
2480             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
2481             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq33,FF),_mm256_mul_pd(vftabscale,rinv33)));
2482
2483             fscal            = felec;
2484
2485             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2486
2487             /* Calculate temporary vectorial force */
2488             tx               = _mm256_mul_pd(fscal,dx33);
2489             ty               = _mm256_mul_pd(fscal,dy33);
2490             tz               = _mm256_mul_pd(fscal,dz33);
2491
2492             /* Update vectorial force */
2493             fix3             = _mm256_add_pd(fix3,tx);
2494             fiy3             = _mm256_add_pd(fiy3,ty);
2495             fiz3             = _mm256_add_pd(fiz3,tz);
2496
2497             fjx3             = _mm256_add_pd(fjx3,tx);
2498             fjy3             = _mm256_add_pd(fjy3,ty);
2499             fjz3             = _mm256_add_pd(fjz3,tz);
2500
2501             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2502             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2503             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2504             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2505
2506             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2507                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2508                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2509
2510             /* Inner loop uses 390 flops */
2511         }
2512
2513         /* End of innermost loop */
2514
2515         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2516                                                  f+i_coord_offset,fshift+i_shift_offset);
2517
2518         /* Increment number of inner iterations */
2519         inneriter                  += j_index_end - j_index_start;
2520
2521         /* Outer loop uses 24 flops */
2522     }
2523
2524     /* Increment number of outer iterations */
2525     outeriter        += nri;
2526
2527     /* Update outer/inner flops */
2528
2529     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*390);
2530 }