src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecCoul_VdwLJ_GeomW4W4_avx_256_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_double.h"
  34 #include "kernelutil_x86_avx_256_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  63     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  64     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  65     real             rcutoff_scalar;
  66     real             *shiftvec,*fshift,*x,*f;
  67     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  68     real             scratch[4*DIM];
  69     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  70     real *           vdwioffsetptr0;
  71     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  72     real *           vdwioffsetptr1;
  73     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     real *           vdwioffsetptr3;
  77     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  78     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  79     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  81     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  82     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  83     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  84     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  85     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  86     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  87     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  88     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  89     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  90     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  91     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  92     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  93     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  94     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  95     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  96     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     int              nvdwtype;
  99     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 100     int              *vdwtype;
 101     real             *vdwparam;
 102     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 103     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 104     __m256d          dummy_mask,cutoff_mask;
 105     __m128           tmpmask0,tmpmask1;
 106     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 107     __m256d          one     = _mm256_set1_pd(1.0);
 108     __m256d          two     = _mm256_set1_pd(2.0);
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = _mm256_set1_pd(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 129     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 130     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 131     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 132
 133     jq1              = _mm256_set1_pd(charge[inr+1]);
 134     jq2              = _mm256_set1_pd(charge[inr+2]);
 135     jq3              = _mm256_set1_pd(charge[inr+3]);
 136     vdwjidx0A        = 2*vdwtype[inr+0];
 137     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 138     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 139     qq11             = _mm256_mul_pd(iq1,jq1);
 140     qq12             = _mm256_mul_pd(iq1,jq2);
 141     qq13             = _mm256_mul_pd(iq1,jq3);
 142     qq21             = _mm256_mul_pd(iq2,jq1);
 143     qq22             = _mm256_mul_pd(iq2,jq2);
 144     qq23             = _mm256_mul_pd(iq2,jq3);
 145     qq31             = _mm256_mul_pd(iq3,jq1);
 146     qq32             = _mm256_mul_pd(iq3,jq2);
 147     qq33             = _mm256_mul_pd(iq3,jq3);
 148
 149     /* Avoid stupid compiler warnings */
 150     jnrA = jnrB = jnrC = jnrD = 0;
 151     j_coord_offsetA = 0;
 152     j_coord_offsetB = 0;
 153     j_coord_offsetC = 0;
 154     j_coord_offsetD = 0;
 155
 156     outeriter        = 0;
 157     inneriter        = 0;
 158
 159     for(iidx=0;iidx<4*DIM;iidx++)
 160     {
 161         scratch[iidx] = 0.0;
 162     }
 163
 164     /* Start outer loop over neighborlists */
 165     for(iidx=0; iidx<nri; iidx++)
 166     {
 167         /* Load shift vector for this list */
 168         i_shift_offset   = DIM*shiftidx[iidx];
 169
 170         /* Load limits for loop over neighbors */
 171         j_index_start    = jindex[iidx];
 172         j_index_end      = jindex[iidx+1];
 173
 174         /* Get outer coordinate index */
 175         inr              = iinr[iidx];
 176         i_coord_offset   = DIM*inr;
 177
 178         /* Load i particle coords and add shift vector */
 179         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 180                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 181
 182         fix0             = _mm256_setzero_pd();
 183         fiy0             = _mm256_setzero_pd();
 184         fiz0             = _mm256_setzero_pd();
 185         fix1             = _mm256_setzero_pd();
 186         fiy1             = _mm256_setzero_pd();
 187         fiz1             = _mm256_setzero_pd();
 188         fix2             = _mm256_setzero_pd();
 189         fiy2             = _mm256_setzero_pd();
 190         fiz2             = _mm256_setzero_pd();
 191         fix3             = _mm256_setzero_pd();
 192         fiy3             = _mm256_setzero_pd();
 193         fiz3             = _mm256_setzero_pd();
 194
 195         /* Reset potential sums */
 196         velecsum         = _mm256_setzero_pd();
 197         vvdwsum          = _mm256_setzero_pd();
 198
 199         /* Start inner kernel loop */
 200         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 201         {
 202
 203             /* Get j neighbor index, and coordinate index */
 204             jnrA             = jjnr[jidx];
 205             jnrB             = jjnr[jidx+1];
 206             jnrC             = jjnr[jidx+2];
 207             jnrD             = jjnr[jidx+3];
 208             j_coord_offsetA  = DIM*jnrA;
 209             j_coord_offsetB  = DIM*jnrB;
 210             j_coord_offsetC  = DIM*jnrC;
 211             j_coord_offsetD  = DIM*jnrD;
 212
 213             /* load j atom coordinates */
 214             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 215                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 216                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 217                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 218
 219             /* Calculate displacement vector */
 220             dx00             = _mm256_sub_pd(ix0,jx0);
 221             dy00             = _mm256_sub_pd(iy0,jy0);
 222             dz00             = _mm256_sub_pd(iz0,jz0);
 223             dx11             = _mm256_sub_pd(ix1,jx1);
 224             dy11             = _mm256_sub_pd(iy1,jy1);
 225             dz11             = _mm256_sub_pd(iz1,jz1);
 226             dx12             = _mm256_sub_pd(ix1,jx2);
 227             dy12             = _mm256_sub_pd(iy1,jy2);
 228             dz12             = _mm256_sub_pd(iz1,jz2);
 229             dx13             = _mm256_sub_pd(ix1,jx3);
 230             dy13             = _mm256_sub_pd(iy1,jy3);
 231             dz13             = _mm256_sub_pd(iz1,jz3);
 232             dx21             = _mm256_sub_pd(ix2,jx1);
 233             dy21             = _mm256_sub_pd(iy2,jy1);
 234             dz21             = _mm256_sub_pd(iz2,jz1);
 235             dx22             = _mm256_sub_pd(ix2,jx2);
 236             dy22             = _mm256_sub_pd(iy2,jy2);
 237             dz22             = _mm256_sub_pd(iz2,jz2);
 238             dx23             = _mm256_sub_pd(ix2,jx3);
 239             dy23             = _mm256_sub_pd(iy2,jy3);
 240             dz23             = _mm256_sub_pd(iz2,jz3);
 241             dx31             = _mm256_sub_pd(ix3,jx1);
 242             dy31             = _mm256_sub_pd(iy3,jy1);
 243             dz31             = _mm256_sub_pd(iz3,jz1);
 244             dx32             = _mm256_sub_pd(ix3,jx2);
 245             dy32             = _mm256_sub_pd(iy3,jy2);
 246             dz32             = _mm256_sub_pd(iz3,jz2);
 247             dx33             = _mm256_sub_pd(ix3,jx3);
 248             dy33             = _mm256_sub_pd(iy3,jy3);
 249             dz33             = _mm256_sub_pd(iz3,jz3);
 250
 251             /* Calculate squared distance and things based on it */
 252             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 253             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 254             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 255             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 256             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 257             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 258             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 259             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 260             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 261             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 262
 263             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 264             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 265             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 266             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 267             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 268             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 269             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 270             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 271             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 272
 273             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 274             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 275             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 276             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 277             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 278             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 279             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 280             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 281             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 282             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 283
 284             fjx0             = _mm256_setzero_pd();
 285             fjy0             = _mm256_setzero_pd();
 286             fjz0             = _mm256_setzero_pd();
 287             fjx1             = _mm256_setzero_pd();
 288             fjy1             = _mm256_setzero_pd();
 289             fjz1             = _mm256_setzero_pd();
 290             fjx2             = _mm256_setzero_pd();
 291             fjy2             = _mm256_setzero_pd();
 292             fjz2             = _mm256_setzero_pd();
 293             fjx3             = _mm256_setzero_pd();
 294             fjy3             = _mm256_setzero_pd();
 295             fjz3             = _mm256_setzero_pd();
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             /* LENNARD-JONES DISPERSION/REPULSION */
 302
 303             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 304             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 305             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 306             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 307             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 308
 309             /* Update potential sum for this i atom from the interaction with this j atom. */
 310             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 311
 312             fscal            = fvdw;
 313
 314             /* Calculate temporary vectorial force */
 315             tx               = _mm256_mul_pd(fscal,dx00);
 316             ty               = _mm256_mul_pd(fscal,dy00);
 317             tz               = _mm256_mul_pd(fscal,dz00);
 318
 319             /* Update vectorial force */
 320             fix0             = _mm256_add_pd(fix0,tx);
 321             fiy0             = _mm256_add_pd(fiy0,ty);
 322             fiz0             = _mm256_add_pd(fiz0,tz);
 323
 324             fjx0             = _mm256_add_pd(fjx0,tx);
 325             fjy0             = _mm256_add_pd(fjy0,ty);
 326             fjz0             = _mm256_add_pd(fjz0,tz);
 327
 328             /**************************
 329              * CALCULATE INTERACTIONS *
 330              **************************/
 331
 332             /* COULOMB ELECTROSTATICS */
 333             velec            = _mm256_mul_pd(qq11,rinv11);
 334             felec            = _mm256_mul_pd(velec,rinvsq11);
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             velecsum         = _mm256_add_pd(velecsum,velec);
 338
 339             fscal            = felec;
 340
 341             /* Calculate temporary vectorial force */
 342             tx               = _mm256_mul_pd(fscal,dx11);
 343             ty               = _mm256_mul_pd(fscal,dy11);
 344             tz               = _mm256_mul_pd(fscal,dz11);
 345
 346             /* Update vectorial force */
 347             fix1             = _mm256_add_pd(fix1,tx);
 348             fiy1             = _mm256_add_pd(fiy1,ty);
 349             fiz1             = _mm256_add_pd(fiz1,tz);
 350
 351             fjx1             = _mm256_add_pd(fjx1,tx);
 352             fjy1             = _mm256_add_pd(fjy1,ty);
 353             fjz1             = _mm256_add_pd(fjz1,tz);
 354
 355             /**************************
 356              * CALCULATE INTERACTIONS *
 357              **************************/
 358
 359             /* COULOMB ELECTROSTATICS */
 360             velec            = _mm256_mul_pd(qq12,rinv12);
 361             felec            = _mm256_mul_pd(velec,rinvsq12);
 362
 363             /* Update potential sum for this i atom from the interaction with this j atom. */
 364             velecsum         = _mm256_add_pd(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             /* Calculate temporary vectorial force */
 369             tx               = _mm256_mul_pd(fscal,dx12);
 370             ty               = _mm256_mul_pd(fscal,dy12);
 371             tz               = _mm256_mul_pd(fscal,dz12);
 372
 373             /* Update vectorial force */
 374             fix1             = _mm256_add_pd(fix1,tx);
 375             fiy1             = _mm256_add_pd(fiy1,ty);
 376             fiz1             = _mm256_add_pd(fiz1,tz);
 377
 378             fjx2             = _mm256_add_pd(fjx2,tx);
 379             fjy2             = _mm256_add_pd(fjy2,ty);
 380             fjz2             = _mm256_add_pd(fjz2,tz);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             /* COULOMB ELECTROSTATICS */
 387             velec            = _mm256_mul_pd(qq13,rinv13);
 388             felec            = _mm256_mul_pd(velec,rinvsq13);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velecsum         = _mm256_add_pd(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395             /* Calculate temporary vectorial force */
 396             tx               = _mm256_mul_pd(fscal,dx13);
 397             ty               = _mm256_mul_pd(fscal,dy13);
 398             tz               = _mm256_mul_pd(fscal,dz13);
 399
 400             /* Update vectorial force */
 401             fix1             = _mm256_add_pd(fix1,tx);
 402             fiy1             = _mm256_add_pd(fiy1,ty);
 403             fiz1             = _mm256_add_pd(fiz1,tz);
 404
 405             fjx3             = _mm256_add_pd(fjx3,tx);
 406             fjy3             = _mm256_add_pd(fjy3,ty);
 407             fjz3             = _mm256_add_pd(fjz3,tz);
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             /* COULOMB ELECTROSTATICS */
 414             velec            = _mm256_mul_pd(qq21,rinv21);
 415             felec            = _mm256_mul_pd(velec,rinvsq21);
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velecsum         = _mm256_add_pd(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             /* Calculate temporary vectorial force */
 423             tx               = _mm256_mul_pd(fscal,dx21);
 424             ty               = _mm256_mul_pd(fscal,dy21);
 425             tz               = _mm256_mul_pd(fscal,dz21);
 426
 427             /* Update vectorial force */
 428             fix2             = _mm256_add_pd(fix2,tx);
 429             fiy2             = _mm256_add_pd(fiy2,ty);
 430             fiz2             = _mm256_add_pd(fiz2,tz);
 431
 432             fjx1             = _mm256_add_pd(fjx1,tx);
 433             fjy1             = _mm256_add_pd(fjy1,ty);
 434             fjz1             = _mm256_add_pd(fjz1,tz);
 435
 436             /**************************
 437              * CALCULATE INTERACTIONS *
 438              **************************/
 439
 440             /* COULOMB ELECTROSTATICS */
 441             velec            = _mm256_mul_pd(qq22,rinv22);
 442             felec            = _mm256_mul_pd(velec,rinvsq22);
 443
 444             /* Update potential sum for this i atom from the interaction with this j atom. */
 445             velecsum         = _mm256_add_pd(velecsum,velec);
 446
 447             fscal            = felec;
 448
 449             /* Calculate temporary vectorial force */
 450             tx               = _mm256_mul_pd(fscal,dx22);
 451             ty               = _mm256_mul_pd(fscal,dy22);
 452             tz               = _mm256_mul_pd(fscal,dz22);
 453
 454             /* Update vectorial force */
 455             fix2             = _mm256_add_pd(fix2,tx);
 456             fiy2             = _mm256_add_pd(fiy2,ty);
 457             fiz2             = _mm256_add_pd(fiz2,tz);
 458
 459             fjx2             = _mm256_add_pd(fjx2,tx);
 460             fjy2             = _mm256_add_pd(fjy2,ty);
 461             fjz2             = _mm256_add_pd(fjz2,tz);
 462
 463             /**************************
 464              * CALCULATE INTERACTIONS *
 465              **************************/
 466
 467             /* COULOMB ELECTROSTATICS */
 468             velec            = _mm256_mul_pd(qq23,rinv23);
 469             felec            = _mm256_mul_pd(velec,rinvsq23);
 470
 471             /* Update potential sum for this i atom from the interaction with this j atom. */
 472             velecsum         = _mm256_add_pd(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             /* Calculate temporary vectorial force */
 477             tx               = _mm256_mul_pd(fscal,dx23);
 478             ty               = _mm256_mul_pd(fscal,dy23);
 479             tz               = _mm256_mul_pd(fscal,dz23);
 480
 481             /* Update vectorial force */
 482             fix2             = _mm256_add_pd(fix2,tx);
 483             fiy2             = _mm256_add_pd(fiy2,ty);
 484             fiz2             = _mm256_add_pd(fiz2,tz);
 485
 486             fjx3             = _mm256_add_pd(fjx3,tx);
 487             fjy3             = _mm256_add_pd(fjy3,ty);
 488             fjz3             = _mm256_add_pd(fjz3,tz);
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             /* COULOMB ELECTROSTATICS */
 495             velec            = _mm256_mul_pd(qq31,rinv31);
 496             felec            = _mm256_mul_pd(velec,rinvsq31);
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velecsum         = _mm256_add_pd(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503             /* Calculate temporary vectorial force */
 504             tx               = _mm256_mul_pd(fscal,dx31);
 505             ty               = _mm256_mul_pd(fscal,dy31);
 506             tz               = _mm256_mul_pd(fscal,dz31);
 507
 508             /* Update vectorial force */
 509             fix3             = _mm256_add_pd(fix3,tx);
 510             fiy3             = _mm256_add_pd(fiy3,ty);
 511             fiz3             = _mm256_add_pd(fiz3,tz);
 512
 513             fjx1             = _mm256_add_pd(fjx1,tx);
 514             fjy1             = _mm256_add_pd(fjy1,ty);
 515             fjz1             = _mm256_add_pd(fjz1,tz);
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             /* COULOMB ELECTROSTATICS */
 522             velec            = _mm256_mul_pd(qq32,rinv32);
 523             felec            = _mm256_mul_pd(velec,rinvsq32);
 524
 525             /* Update potential sum for this i atom from the interaction with this j atom. */
 526             velecsum         = _mm256_add_pd(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             /* Calculate temporary vectorial force */
 531             tx               = _mm256_mul_pd(fscal,dx32);
 532             ty               = _mm256_mul_pd(fscal,dy32);
 533             tz               = _mm256_mul_pd(fscal,dz32);
 534
 535             /* Update vectorial force */
 536             fix3             = _mm256_add_pd(fix3,tx);
 537             fiy3             = _mm256_add_pd(fiy3,ty);
 538             fiz3             = _mm256_add_pd(fiz3,tz);
 539
 540             fjx2             = _mm256_add_pd(fjx2,tx);
 541             fjy2             = _mm256_add_pd(fjy2,ty);
 542             fjz2             = _mm256_add_pd(fjz2,tz);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             /* COULOMB ELECTROSTATICS */
 549             velec            = _mm256_mul_pd(qq33,rinv33);
 550             felec            = _mm256_mul_pd(velec,rinvsq33);
 551
 552             /* Update potential sum for this i atom from the interaction with this j atom. */
 553             velecsum         = _mm256_add_pd(velecsum,velec);
 554
 555             fscal            = felec;
 556
 557             /* Calculate temporary vectorial force */
 558             tx               = _mm256_mul_pd(fscal,dx33);
 559             ty               = _mm256_mul_pd(fscal,dy33);
 560             tz               = _mm256_mul_pd(fscal,dz33);
 561
 562             /* Update vectorial force */
 563             fix3             = _mm256_add_pd(fix3,tx);
 564             fiy3             = _mm256_add_pd(fiy3,ty);
 565             fiz3             = _mm256_add_pd(fiz3,tz);
 566
 567             fjx3             = _mm256_add_pd(fjx3,tx);
 568             fjy3             = _mm256_add_pd(fjy3,ty);
 569             fjz3             = _mm256_add_pd(fjz3,tz);
 570
 571             fjptrA             = f+j_coord_offsetA;
 572             fjptrB             = f+j_coord_offsetB;
 573             fjptrC             = f+j_coord_offsetC;
 574             fjptrD             = f+j_coord_offsetD;
 575
 576             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 577                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 578                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 579
 580             /* Inner loop uses 278 flops */
 581         }
 582
 583         if(jidx<j_index_end)
 584         {
 585
 586             /* Get j neighbor index, and coordinate index */
 587             jnrlistA         = jjnr[jidx];
 588             jnrlistB         = jjnr[jidx+1];
 589             jnrlistC         = jjnr[jidx+2];
 590             jnrlistD         = jjnr[jidx+3];
 591             /* Sign of each element will be negative for non-real atoms.
 592              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 593              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 594              */
 595             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 596
 597             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 598             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 599             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 600
 601             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 602             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 603             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 604             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 605             j_coord_offsetA  = DIM*jnrA;
 606             j_coord_offsetB  = DIM*jnrB;
 607             j_coord_offsetC  = DIM*jnrC;
 608             j_coord_offsetD  = DIM*jnrD;
 609
 610             /* load j atom coordinates */
 611             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 612                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 613                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 614                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 615
 616             /* Calculate displacement vector */
 617             dx00             = _mm256_sub_pd(ix0,jx0);
 618             dy00             = _mm256_sub_pd(iy0,jy0);
 619             dz00             = _mm256_sub_pd(iz0,jz0);
 620             dx11             = _mm256_sub_pd(ix1,jx1);
 621             dy11             = _mm256_sub_pd(iy1,jy1);
 622             dz11             = _mm256_sub_pd(iz1,jz1);
 623             dx12             = _mm256_sub_pd(ix1,jx2);
 624             dy12             = _mm256_sub_pd(iy1,jy2);
 625             dz12             = _mm256_sub_pd(iz1,jz2);
 626             dx13             = _mm256_sub_pd(ix1,jx3);
 627             dy13             = _mm256_sub_pd(iy1,jy3);
 628             dz13             = _mm256_sub_pd(iz1,jz3);
 629             dx21             = _mm256_sub_pd(ix2,jx1);
 630             dy21             = _mm256_sub_pd(iy2,jy1);
 631             dz21             = _mm256_sub_pd(iz2,jz1);
 632             dx22             = _mm256_sub_pd(ix2,jx2);
 633             dy22             = _mm256_sub_pd(iy2,jy2);
 634             dz22             = _mm256_sub_pd(iz2,jz2);
 635             dx23             = _mm256_sub_pd(ix2,jx3);
 636             dy23             = _mm256_sub_pd(iy2,jy3);
 637             dz23             = _mm256_sub_pd(iz2,jz3);
 638             dx31             = _mm256_sub_pd(ix3,jx1);
 639             dy31             = _mm256_sub_pd(iy3,jy1);
 640             dz31             = _mm256_sub_pd(iz3,jz1);
 641             dx32             = _mm256_sub_pd(ix3,jx2);
 642             dy32             = _mm256_sub_pd(iy3,jy2);
 643             dz32             = _mm256_sub_pd(iz3,jz2);
 644             dx33             = _mm256_sub_pd(ix3,jx3);
 645             dy33             = _mm256_sub_pd(iy3,jy3);
 646             dz33             = _mm256_sub_pd(iz3,jz3);
 647
 648             /* Calculate squared distance and things based on it */
 649             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 650             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 651             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 652             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 653             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 654             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 655             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 656             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 657             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 658             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 659
 660             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 661             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 662             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 663             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 664             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 665             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 666             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 667             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 668             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 669
 670             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 671             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 672             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 673             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 674             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 675             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 676             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 677             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 678             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 679             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 680
 681             fjx0             = _mm256_setzero_pd();
 682             fjy0             = _mm256_setzero_pd();
 683             fjz0             = _mm256_setzero_pd();
 684             fjx1             = _mm256_setzero_pd();
 685             fjy1             = _mm256_setzero_pd();
 686             fjz1             = _mm256_setzero_pd();
 687             fjx2             = _mm256_setzero_pd();
 688             fjy2             = _mm256_setzero_pd();
 689             fjz2             = _mm256_setzero_pd();
 690             fjx3             = _mm256_setzero_pd();
 691             fjy3             = _mm256_setzero_pd();
 692             fjz3             = _mm256_setzero_pd();
 693
 694             /**************************
 695              * CALCULATE INTERACTIONS *
 696              **************************/
 697
 698             /* LENNARD-JONES DISPERSION/REPULSION */
 699
 700             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 701             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 702             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 703             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 704             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 705
 706             /* Update potential sum for this i atom from the interaction with this j atom. */
 707             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 708             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 709
 710             fscal            = fvdw;
 711
 712             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 713
 714             /* Calculate temporary vectorial force */
 715             tx               = _mm256_mul_pd(fscal,dx00);
 716             ty               = _mm256_mul_pd(fscal,dy00);
 717             tz               = _mm256_mul_pd(fscal,dz00);
 718
 719             /* Update vectorial force */
 720             fix0             = _mm256_add_pd(fix0,tx);
 721             fiy0             = _mm256_add_pd(fiy0,ty);
 722             fiz0             = _mm256_add_pd(fiz0,tz);
 723
 724             fjx0             = _mm256_add_pd(fjx0,tx);
 725             fjy0             = _mm256_add_pd(fjy0,ty);
 726             fjz0             = _mm256_add_pd(fjz0,tz);
 727
 728             /**************************
 729              * CALCULATE INTERACTIONS *
 730              **************************/
 731
 732             /* COULOMB ELECTROSTATICS */
 733             velec            = _mm256_mul_pd(qq11,rinv11);
 734             felec            = _mm256_mul_pd(velec,rinvsq11);
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             velec            = _mm256_andnot_pd(dummy_mask,velec);
 738             velecsum         = _mm256_add_pd(velecsum,velec);
 739
 740             fscal            = felec;
 741
 742             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 743
 744             /* Calculate temporary vectorial force */
 745             tx               = _mm256_mul_pd(fscal,dx11);
 746             ty               = _mm256_mul_pd(fscal,dy11);
 747             tz               = _mm256_mul_pd(fscal,dz11);
 748
 749             /* Update vectorial force */
 750             fix1             = _mm256_add_pd(fix1,tx);
 751             fiy1             = _mm256_add_pd(fiy1,ty);
 752             fiz1             = _mm256_add_pd(fiz1,tz);
 753
 754             fjx1             = _mm256_add_pd(fjx1,tx);
 755             fjy1             = _mm256_add_pd(fjy1,ty);
 756             fjz1             = _mm256_add_pd(fjz1,tz);
 757
 758             /**************************
 759              * CALCULATE INTERACTIONS *
 760              **************************/
 761
 762             /* COULOMB ELECTROSTATICS */
 763             velec            = _mm256_mul_pd(qq12,rinv12);
 764             felec            = _mm256_mul_pd(velec,rinvsq12);
 765
 766             /* Update potential sum for this i atom from the interaction with this j atom. */
 767             velec            = _mm256_andnot_pd(dummy_mask,velec);
 768             velecsum         = _mm256_add_pd(velecsum,velec);
 769
 770             fscal            = felec;
 771
 772             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm256_mul_pd(fscal,dx12);
 776             ty               = _mm256_mul_pd(fscal,dy12);
 777             tz               = _mm256_mul_pd(fscal,dz12);
 778
 779             /* Update vectorial force */
 780             fix1             = _mm256_add_pd(fix1,tx);
 781             fiy1             = _mm256_add_pd(fiy1,ty);
 782             fiz1             = _mm256_add_pd(fiz1,tz);
 783
 784             fjx2             = _mm256_add_pd(fjx2,tx);
 785             fjy2             = _mm256_add_pd(fjy2,ty);
 786             fjz2             = _mm256_add_pd(fjz2,tz);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm256_mul_pd(qq13,rinv13);
 794             felec            = _mm256_mul_pd(velec,rinvsq13);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm256_andnot_pd(dummy_mask,velec);
 798             velecsum         = _mm256_add_pd(velecsum,velec);
 799
 800             fscal            = felec;
 801
 802             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 803
 804             /* Calculate temporary vectorial force */
 805             tx               = _mm256_mul_pd(fscal,dx13);
 806             ty               = _mm256_mul_pd(fscal,dy13);
 807             tz               = _mm256_mul_pd(fscal,dz13);
 808
 809             /* Update vectorial force */
 810             fix1             = _mm256_add_pd(fix1,tx);
 811             fiy1             = _mm256_add_pd(fiy1,ty);
 812             fiz1             = _mm256_add_pd(fiz1,tz);
 813
 814             fjx3             = _mm256_add_pd(fjx3,tx);
 815             fjy3             = _mm256_add_pd(fjy3,ty);
 816             fjz3             = _mm256_add_pd(fjz3,tz);
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             /* COULOMB ELECTROSTATICS */
 823             velec            = _mm256_mul_pd(qq21,rinv21);
 824             felec            = _mm256_mul_pd(velec,rinvsq21);
 825
 826             /* Update potential sum for this i atom from the interaction with this j atom. */
 827             velec            = _mm256_andnot_pd(dummy_mask,velec);
 828             velecsum         = _mm256_add_pd(velecsum,velec);
 829
 830             fscal            = felec;
 831
 832             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 833
 834             /* Calculate temporary vectorial force */
 835             tx               = _mm256_mul_pd(fscal,dx21);
 836             ty               = _mm256_mul_pd(fscal,dy21);
 837             tz               = _mm256_mul_pd(fscal,dz21);
 838
 839             /* Update vectorial force */
 840             fix2             = _mm256_add_pd(fix2,tx);
 841             fiy2             = _mm256_add_pd(fiy2,ty);
 842             fiz2             = _mm256_add_pd(fiz2,tz);
 843
 844             fjx1             = _mm256_add_pd(fjx1,tx);
 845             fjy1             = _mm256_add_pd(fjy1,ty);
 846             fjz1             = _mm256_add_pd(fjz1,tz);
 847
 848             /**************************
 849              * CALCULATE INTERACTIONS *
 850              **************************/
 851
 852             /* COULOMB ELECTROSTATICS */
 853             velec            = _mm256_mul_pd(qq22,rinv22);
 854             felec            = _mm256_mul_pd(velec,rinvsq22);
 855
 856             /* Update potential sum for this i atom from the interaction with this j atom. */
 857             velec            = _mm256_andnot_pd(dummy_mask,velec);
 858             velecsum         = _mm256_add_pd(velecsum,velec);
 859
 860             fscal            = felec;
 861
 862             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 863
 864             /* Calculate temporary vectorial force */
 865             tx               = _mm256_mul_pd(fscal,dx22);
 866             ty               = _mm256_mul_pd(fscal,dy22);
 867             tz               = _mm256_mul_pd(fscal,dz22);
 868
 869             /* Update vectorial force */
 870             fix2             = _mm256_add_pd(fix2,tx);
 871             fiy2             = _mm256_add_pd(fiy2,ty);
 872             fiz2             = _mm256_add_pd(fiz2,tz);
 873
 874             fjx2             = _mm256_add_pd(fjx2,tx);
 875             fjy2             = _mm256_add_pd(fjy2,ty);
 876             fjz2             = _mm256_add_pd(fjz2,tz);
 877
 878             /**************************
 879              * CALCULATE INTERACTIONS *
 880              **************************/
 881
 882             /* COULOMB ELECTROSTATICS */
 883             velec            = _mm256_mul_pd(qq23,rinv23);
 884             felec            = _mm256_mul_pd(velec,rinvsq23);
 885
 886             /* Update potential sum for this i atom from the interaction with this j atom. */
 887             velec            = _mm256_andnot_pd(dummy_mask,velec);
 888             velecsum         = _mm256_add_pd(velecsum,velec);
 889
 890             fscal            = felec;
 891
 892             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 893
 894             /* Calculate temporary vectorial force */
 895             tx               = _mm256_mul_pd(fscal,dx23);
 896             ty               = _mm256_mul_pd(fscal,dy23);
 897             tz               = _mm256_mul_pd(fscal,dz23);
 898
 899             /* Update vectorial force */
 900             fix2             = _mm256_add_pd(fix2,tx);
 901             fiy2             = _mm256_add_pd(fiy2,ty);
 902             fiz2             = _mm256_add_pd(fiz2,tz);
 903
 904             fjx3             = _mm256_add_pd(fjx3,tx);
 905             fjy3             = _mm256_add_pd(fjy3,ty);
 906             fjz3             = _mm256_add_pd(fjz3,tz);
 907
 908             /**************************
 909              * CALCULATE INTERACTIONS *
 910              **************************/
 911
 912             /* COULOMB ELECTROSTATICS */
 913             velec            = _mm256_mul_pd(qq31,rinv31);
 914             felec            = _mm256_mul_pd(velec,rinvsq31);
 915
 916             /* Update potential sum for this i atom from the interaction with this j atom. */
 917             velec            = _mm256_andnot_pd(dummy_mask,velec);
 918             velecsum         = _mm256_add_pd(velecsum,velec);
 919
 920             fscal            = felec;
 921
 922             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 923
 924             /* Calculate temporary vectorial force */
 925             tx               = _mm256_mul_pd(fscal,dx31);
 926             ty               = _mm256_mul_pd(fscal,dy31);
 927             tz               = _mm256_mul_pd(fscal,dz31);
 928
 929             /* Update vectorial force */
 930             fix3             = _mm256_add_pd(fix3,tx);
 931             fiy3             = _mm256_add_pd(fiy3,ty);
 932             fiz3             = _mm256_add_pd(fiz3,tz);
 933
 934             fjx1             = _mm256_add_pd(fjx1,tx);
 935             fjy1             = _mm256_add_pd(fjy1,ty);
 936             fjz1             = _mm256_add_pd(fjz1,tz);
 937
 938             /**************************
 939              * CALCULATE INTERACTIONS *
 940              **************************/
 941
 942             /* COULOMB ELECTROSTATICS */
 943             velec            = _mm256_mul_pd(qq32,rinv32);
 944             felec            = _mm256_mul_pd(velec,rinvsq32);
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm256_andnot_pd(dummy_mask,velec);
 948             velecsum         = _mm256_add_pd(velecsum,velec);
 949
 950             fscal            = felec;
 951
 952             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 953
 954             /* Calculate temporary vectorial force */
 955             tx               = _mm256_mul_pd(fscal,dx32);
 956             ty               = _mm256_mul_pd(fscal,dy32);
 957             tz               = _mm256_mul_pd(fscal,dz32);
 958
 959             /* Update vectorial force */
 960             fix3             = _mm256_add_pd(fix3,tx);
 961             fiy3             = _mm256_add_pd(fiy3,ty);
 962             fiz3             = _mm256_add_pd(fiz3,tz);
 963
 964             fjx2             = _mm256_add_pd(fjx2,tx);
 965             fjy2             = _mm256_add_pd(fjy2,ty);
 966             fjz2             = _mm256_add_pd(fjz2,tz);
 967
 968             /**************************
 969              * CALCULATE INTERACTIONS *
 970              **************************/
 971
 972             /* COULOMB ELECTROSTATICS */
 973             velec            = _mm256_mul_pd(qq33,rinv33);
 974             felec            = _mm256_mul_pd(velec,rinvsq33);
 975
 976             /* Update potential sum for this i atom from the interaction with this j atom. */
 977             velec            = _mm256_andnot_pd(dummy_mask,velec);
 978             velecsum         = _mm256_add_pd(velecsum,velec);
 979
 980             fscal            = felec;
 981
 982             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 983
 984             /* Calculate temporary vectorial force */
 985             tx               = _mm256_mul_pd(fscal,dx33);
 986             ty               = _mm256_mul_pd(fscal,dy33);
 987             tz               = _mm256_mul_pd(fscal,dz33);
 988
 989             /* Update vectorial force */
 990             fix3             = _mm256_add_pd(fix3,tx);
 991             fiy3             = _mm256_add_pd(fiy3,ty);
 992             fiz3             = _mm256_add_pd(fiz3,tz);
 993
 994             fjx3             = _mm256_add_pd(fjx3,tx);
 995             fjy3             = _mm256_add_pd(fjy3,ty);
 996             fjz3             = _mm256_add_pd(fjz3,tz);
 997
 998             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 999             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1000             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1001             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1002
1003             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1004                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1005                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1006
1007             /* Inner loop uses 278 flops */
1008         }
1009
1010         /* End of innermost loop */
1011
1012         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1013                                                  f+i_coord_offset,fshift+i_shift_offset);
1014
1015         ggid                        = gid[iidx];
1016         /* Update potential energies */
1017         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1018         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1019
1020         /* Increment number of inner iterations */
1021         inneriter                  += j_index_end - j_index_start;
1022
1023         /* Outer loop uses 26 flops */
1024     }
1025
1026     /* Increment number of outer iterations */
1027     outeriter        += nri;
1028
1029     /* Update outer/inner flops */
1030
1031     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1032 }
1033 /*
1034  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1035  * Electrostatics interaction: Coulomb
1036  * VdW interaction:            LennardJones
1037  * Geometry:                   Water4-Water4
1038  * Calculate force/pot:        Force
1039  */
1040 void
1041 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1042                     (t_nblist * gmx_restrict                nlist,
1043                      rvec * gmx_restrict                    xx,
1044                      rvec * gmx_restrict                    ff,
1045                      t_forcerec * gmx_restrict              fr,
1046                      t_mdatoms * gmx_restrict               mdatoms,
1047                      nb_kernel_data_t * gmx_restrict        kernel_data,
1048                      t_nrnb * gmx_restrict                  nrnb)
1049 {
1050     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1051      * just 0 for non-waters.
1052      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1053      * jnr indices corresponding to data put in the four positions in the SIMD register.
1054      */
1055     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1056     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1057     int              jnrA,jnrB,jnrC,jnrD;
1058     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1059     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1060     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1061     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1062     real             rcutoff_scalar;
1063     real             *shiftvec,*fshift,*x,*f;
1064     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1065     real             scratch[4*DIM];
1066     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1067     real *           vdwioffsetptr0;
1068     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1069     real *           vdwioffsetptr1;
1070     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1071     real *           vdwioffsetptr2;
1072     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1073     real *           vdwioffsetptr3;
1074     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1075     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1076     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1077     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1078     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1079     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1080     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1081     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1082     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1083     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1084     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1085     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1086     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1087     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1088     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1089     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1090     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1091     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1092     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1093     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1094     real             *charge;
1095     int              nvdwtype;
1096     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1097     int              *vdwtype;
1098     real             *vdwparam;
1099     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1100     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1101     __m256d          dummy_mask,cutoff_mask;
1102     __m128           tmpmask0,tmpmask1;
1103     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1104     __m256d          one     = _mm256_set1_pd(1.0);
1105     __m256d          two     = _mm256_set1_pd(2.0);
1106     x                = xx[0];
1107     f                = ff[0];
1108
1109     nri              = nlist->nri;
1110     iinr             = nlist->iinr;
1111     jindex           = nlist->jindex;
1112     jjnr             = nlist->jjnr;
1113     shiftidx         = nlist->shift;
1114     gid              = nlist->gid;
1115     shiftvec         = fr->shift_vec[0];
1116     fshift           = fr->fshift[0];
1117     facel            = _mm256_set1_pd(fr->epsfac);
1118     charge           = mdatoms->chargeA;
1119     nvdwtype         = fr->ntype;
1120     vdwparam         = fr->nbfp;
1121     vdwtype          = mdatoms->typeA;
1122
1123     /* Setup water-specific parameters */
1124     inr              = nlist->iinr[0];
1125     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1126     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1127     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1128     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1129
1130     jq1              = _mm256_set1_pd(charge[inr+1]);
1131     jq2              = _mm256_set1_pd(charge[inr+2]);
1132     jq3              = _mm256_set1_pd(charge[inr+3]);
1133     vdwjidx0A        = 2*vdwtype[inr+0];
1134     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1135     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1136     qq11             = _mm256_mul_pd(iq1,jq1);
1137     qq12             = _mm256_mul_pd(iq1,jq2);
1138     qq13             = _mm256_mul_pd(iq1,jq3);
1139     qq21             = _mm256_mul_pd(iq2,jq1);
1140     qq22             = _mm256_mul_pd(iq2,jq2);
1141     qq23             = _mm256_mul_pd(iq2,jq3);
1142     qq31             = _mm256_mul_pd(iq3,jq1);
1143     qq32             = _mm256_mul_pd(iq3,jq2);
1144     qq33             = _mm256_mul_pd(iq3,jq3);
1145
1146     /* Avoid stupid compiler warnings */
1147     jnrA = jnrB = jnrC = jnrD = 0;
1148     j_coord_offsetA = 0;
1149     j_coord_offsetB = 0;
1150     j_coord_offsetC = 0;
1151     j_coord_offsetD = 0;
1152
1153     outeriter        = 0;
1154     inneriter        = 0;
1155
1156     for(iidx=0;iidx<4*DIM;iidx++)
1157     {
1158         scratch[iidx] = 0.0;
1159     }
1160
1161     /* Start outer loop over neighborlists */
1162     for(iidx=0; iidx<nri; iidx++)
1163     {
1164         /* Load shift vector for this list */
1165         i_shift_offset   = DIM*shiftidx[iidx];
1166
1167         /* Load limits for loop over neighbors */
1168         j_index_start    = jindex[iidx];
1169         j_index_end      = jindex[iidx+1];
1170
1171         /* Get outer coordinate index */
1172         inr              = iinr[iidx];
1173         i_coord_offset   = DIM*inr;
1174
1175         /* Load i particle coords and add shift vector */
1176         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1177                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1178
1179         fix0             = _mm256_setzero_pd();
1180         fiy0             = _mm256_setzero_pd();
1181         fiz0             = _mm256_setzero_pd();
1182         fix1             = _mm256_setzero_pd();
1183         fiy1             = _mm256_setzero_pd();
1184         fiz1             = _mm256_setzero_pd();
1185         fix2             = _mm256_setzero_pd();
1186         fiy2             = _mm256_setzero_pd();
1187         fiz2             = _mm256_setzero_pd();
1188         fix3             = _mm256_setzero_pd();
1189         fiy3             = _mm256_setzero_pd();
1190         fiz3             = _mm256_setzero_pd();
1191
1192         /* Start inner kernel loop */
1193         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1194         {
1195
1196             /* Get j neighbor index, and coordinate index */
1197             jnrA             = jjnr[jidx];
1198             jnrB             = jjnr[jidx+1];
1199             jnrC             = jjnr[jidx+2];
1200             jnrD             = jjnr[jidx+3];
1201             j_coord_offsetA  = DIM*jnrA;
1202             j_coord_offsetB  = DIM*jnrB;
1203             j_coord_offsetC  = DIM*jnrC;
1204             j_coord_offsetD  = DIM*jnrD;
1205
1206             /* load j atom coordinates */
1207             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1208                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1209                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1210                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1211
1212             /* Calculate displacement vector */
1213             dx00             = _mm256_sub_pd(ix0,jx0);
1214             dy00             = _mm256_sub_pd(iy0,jy0);
1215             dz00             = _mm256_sub_pd(iz0,jz0);
1216             dx11             = _mm256_sub_pd(ix1,jx1);
1217             dy11             = _mm256_sub_pd(iy1,jy1);
1218             dz11             = _mm256_sub_pd(iz1,jz1);
1219             dx12             = _mm256_sub_pd(ix1,jx2);
1220             dy12             = _mm256_sub_pd(iy1,jy2);
1221             dz12             = _mm256_sub_pd(iz1,jz2);
1222             dx13             = _mm256_sub_pd(ix1,jx3);
1223             dy13             = _mm256_sub_pd(iy1,jy3);
1224             dz13             = _mm256_sub_pd(iz1,jz3);
1225             dx21             = _mm256_sub_pd(ix2,jx1);
1226             dy21             = _mm256_sub_pd(iy2,jy1);
1227             dz21             = _mm256_sub_pd(iz2,jz1);
1228             dx22             = _mm256_sub_pd(ix2,jx2);
1229             dy22             = _mm256_sub_pd(iy2,jy2);
1230             dz22             = _mm256_sub_pd(iz2,jz2);
1231             dx23             = _mm256_sub_pd(ix2,jx3);
1232             dy23             = _mm256_sub_pd(iy2,jy3);
1233             dz23             = _mm256_sub_pd(iz2,jz3);
1234             dx31             = _mm256_sub_pd(ix3,jx1);
1235             dy31             = _mm256_sub_pd(iy3,jy1);
1236             dz31             = _mm256_sub_pd(iz3,jz1);
1237             dx32             = _mm256_sub_pd(ix3,jx2);
1238             dy32             = _mm256_sub_pd(iy3,jy2);
1239             dz32             = _mm256_sub_pd(iz3,jz2);
1240             dx33             = _mm256_sub_pd(ix3,jx3);
1241             dy33             = _mm256_sub_pd(iy3,jy3);
1242             dz33             = _mm256_sub_pd(iz3,jz3);
1243
1244             /* Calculate squared distance and things based on it */
1245             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1246             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1247             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1248             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1249             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1250             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1251             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1252             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1253             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1254             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1255
1256             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1257             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1258             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1259             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1260             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1261             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1262             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1263             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1264             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1265
1266             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1267             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1268             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1269             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1270             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1271             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1272             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1273             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1274             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1275             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1276
1277             fjx0             = _mm256_setzero_pd();
1278             fjy0             = _mm256_setzero_pd();
1279             fjz0             = _mm256_setzero_pd();
1280             fjx1             = _mm256_setzero_pd();
1281             fjy1             = _mm256_setzero_pd();
1282             fjz1             = _mm256_setzero_pd();
1283             fjx2             = _mm256_setzero_pd();
1284             fjy2             = _mm256_setzero_pd();
1285             fjz2             = _mm256_setzero_pd();
1286             fjx3             = _mm256_setzero_pd();
1287             fjy3             = _mm256_setzero_pd();
1288             fjz3             = _mm256_setzero_pd();
1289
1290             /**************************
1291              * CALCULATE INTERACTIONS *
1292              **************************/
1293
1294             /* LENNARD-JONES DISPERSION/REPULSION */
1295
1296             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1297             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1298
1299             fscal            = fvdw;
1300
1301             /* Calculate temporary vectorial force */
1302             tx               = _mm256_mul_pd(fscal,dx00);
1303             ty               = _mm256_mul_pd(fscal,dy00);
1304             tz               = _mm256_mul_pd(fscal,dz00);
1305
1306             /* Update vectorial force */
1307             fix0             = _mm256_add_pd(fix0,tx);
1308             fiy0             = _mm256_add_pd(fiy0,ty);
1309             fiz0             = _mm256_add_pd(fiz0,tz);
1310
1311             fjx0             = _mm256_add_pd(fjx0,tx);
1312             fjy0             = _mm256_add_pd(fjy0,ty);
1313             fjz0             = _mm256_add_pd(fjz0,tz);
1314
1315             /**************************
1316              * CALCULATE INTERACTIONS *
1317              **************************/
1318
1319             /* COULOMB ELECTROSTATICS */
1320             velec            = _mm256_mul_pd(qq11,rinv11);
1321             felec            = _mm256_mul_pd(velec,rinvsq11);
1322
1323             fscal            = felec;
1324
1325             /* Calculate temporary vectorial force */
1326             tx               = _mm256_mul_pd(fscal,dx11);
1327             ty               = _mm256_mul_pd(fscal,dy11);
1328             tz               = _mm256_mul_pd(fscal,dz11);
1329
1330             /* Update vectorial force */
1331             fix1             = _mm256_add_pd(fix1,tx);
1332             fiy1             = _mm256_add_pd(fiy1,ty);
1333             fiz1             = _mm256_add_pd(fiz1,tz);
1334
1335             fjx1             = _mm256_add_pd(fjx1,tx);
1336             fjy1             = _mm256_add_pd(fjy1,ty);
1337             fjz1             = _mm256_add_pd(fjz1,tz);
1338
1339             /**************************
1340              * CALCULATE INTERACTIONS *
1341              **************************/
1342
1343             /* COULOMB ELECTROSTATICS */
1344             velec            = _mm256_mul_pd(qq12,rinv12);
1345             felec            = _mm256_mul_pd(velec,rinvsq12);
1346
1347             fscal            = felec;
1348
1349             /* Calculate temporary vectorial force */
1350             tx               = _mm256_mul_pd(fscal,dx12);
1351             ty               = _mm256_mul_pd(fscal,dy12);
1352             tz               = _mm256_mul_pd(fscal,dz12);
1353
1354             /* Update vectorial force */
1355             fix1             = _mm256_add_pd(fix1,tx);
1356             fiy1             = _mm256_add_pd(fiy1,ty);
1357             fiz1             = _mm256_add_pd(fiz1,tz);
1358
1359             fjx2             = _mm256_add_pd(fjx2,tx);
1360             fjy2             = _mm256_add_pd(fjy2,ty);
1361             fjz2             = _mm256_add_pd(fjz2,tz);
1362
1363             /**************************
1364              * CALCULATE INTERACTIONS *
1365              **************************/
1366
1367             /* COULOMB ELECTROSTATICS */
1368             velec            = _mm256_mul_pd(qq13,rinv13);
1369             felec            = _mm256_mul_pd(velec,rinvsq13);
1370
1371             fscal            = felec;
1372
1373             /* Calculate temporary vectorial force */
1374             tx               = _mm256_mul_pd(fscal,dx13);
1375             ty               = _mm256_mul_pd(fscal,dy13);
1376             tz               = _mm256_mul_pd(fscal,dz13);
1377
1378             /* Update vectorial force */
1379             fix1             = _mm256_add_pd(fix1,tx);
1380             fiy1             = _mm256_add_pd(fiy1,ty);
1381             fiz1             = _mm256_add_pd(fiz1,tz);
1382
1383             fjx3             = _mm256_add_pd(fjx3,tx);
1384             fjy3             = _mm256_add_pd(fjy3,ty);
1385             fjz3             = _mm256_add_pd(fjz3,tz);
1386
1387             /**************************
1388              * CALCULATE INTERACTIONS *
1389              **************************/
1390
1391             /* COULOMB ELECTROSTATICS */
1392             velec            = _mm256_mul_pd(qq21,rinv21);
1393             felec            = _mm256_mul_pd(velec,rinvsq21);
1394
1395             fscal            = felec;
1396
1397             /* Calculate temporary vectorial force */
1398             tx               = _mm256_mul_pd(fscal,dx21);
1399             ty               = _mm256_mul_pd(fscal,dy21);
1400             tz               = _mm256_mul_pd(fscal,dz21);
1401
1402             /* Update vectorial force */
1403             fix2             = _mm256_add_pd(fix2,tx);
1404             fiy2             = _mm256_add_pd(fiy2,ty);
1405             fiz2             = _mm256_add_pd(fiz2,tz);
1406
1407             fjx1             = _mm256_add_pd(fjx1,tx);
1408             fjy1             = _mm256_add_pd(fjy1,ty);
1409             fjz1             = _mm256_add_pd(fjz1,tz);
1410
1411             /**************************
1412              * CALCULATE INTERACTIONS *
1413              **************************/
1414
1415             /* COULOMB ELECTROSTATICS */
1416             velec            = _mm256_mul_pd(qq22,rinv22);
1417             felec            = _mm256_mul_pd(velec,rinvsq22);
1418
1419             fscal            = felec;
1420
1421             /* Calculate temporary vectorial force */
1422             tx               = _mm256_mul_pd(fscal,dx22);
1423             ty               = _mm256_mul_pd(fscal,dy22);
1424             tz               = _mm256_mul_pd(fscal,dz22);
1425
1426             /* Update vectorial force */
1427             fix2             = _mm256_add_pd(fix2,tx);
1428             fiy2             = _mm256_add_pd(fiy2,ty);
1429             fiz2             = _mm256_add_pd(fiz2,tz);
1430
1431             fjx2             = _mm256_add_pd(fjx2,tx);
1432             fjy2             = _mm256_add_pd(fjy2,ty);
1433             fjz2             = _mm256_add_pd(fjz2,tz);
1434
1435             /**************************
1436              * CALCULATE INTERACTIONS *
1437              **************************/
1438
1439             /* COULOMB ELECTROSTATICS */
1440             velec            = _mm256_mul_pd(qq23,rinv23);
1441             felec            = _mm256_mul_pd(velec,rinvsq23);
1442
1443             fscal            = felec;
1444
1445             /* Calculate temporary vectorial force */
1446             tx               = _mm256_mul_pd(fscal,dx23);
1447             ty               = _mm256_mul_pd(fscal,dy23);
1448             tz               = _mm256_mul_pd(fscal,dz23);
1449
1450             /* Update vectorial force */
1451             fix2             = _mm256_add_pd(fix2,tx);
1452             fiy2             = _mm256_add_pd(fiy2,ty);
1453             fiz2             = _mm256_add_pd(fiz2,tz);
1454
1455             fjx3             = _mm256_add_pd(fjx3,tx);
1456             fjy3             = _mm256_add_pd(fjy3,ty);
1457             fjz3             = _mm256_add_pd(fjz3,tz);
1458
1459             /**************************
1460              * CALCULATE INTERACTIONS *
1461              **************************/
1462
1463             /* COULOMB ELECTROSTATICS */
1464             velec            = _mm256_mul_pd(qq31,rinv31);
1465             felec            = _mm256_mul_pd(velec,rinvsq31);
1466
1467             fscal            = felec;
1468
1469             /* Calculate temporary vectorial force */
1470             tx               = _mm256_mul_pd(fscal,dx31);
1471             ty               = _mm256_mul_pd(fscal,dy31);
1472             tz               = _mm256_mul_pd(fscal,dz31);
1473
1474             /* Update vectorial force */
1475             fix3             = _mm256_add_pd(fix3,tx);
1476             fiy3             = _mm256_add_pd(fiy3,ty);
1477             fiz3             = _mm256_add_pd(fiz3,tz);
1478
1479             fjx1             = _mm256_add_pd(fjx1,tx);
1480             fjy1             = _mm256_add_pd(fjy1,ty);
1481             fjz1             = _mm256_add_pd(fjz1,tz);
1482
1483             /**************************
1484              * CALCULATE INTERACTIONS *
1485              **************************/
1486
1487             /* COULOMB ELECTROSTATICS */
1488             velec            = _mm256_mul_pd(qq32,rinv32);
1489             felec            = _mm256_mul_pd(velec,rinvsq32);
1490
1491             fscal            = felec;
1492
1493             /* Calculate temporary vectorial force */
1494             tx               = _mm256_mul_pd(fscal,dx32);
1495             ty               = _mm256_mul_pd(fscal,dy32);
1496             tz               = _mm256_mul_pd(fscal,dz32);
1497
1498             /* Update vectorial force */
1499             fix3             = _mm256_add_pd(fix3,tx);
1500             fiy3             = _mm256_add_pd(fiy3,ty);
1501             fiz3             = _mm256_add_pd(fiz3,tz);
1502
1503             fjx2             = _mm256_add_pd(fjx2,tx);
1504             fjy2             = _mm256_add_pd(fjy2,ty);
1505             fjz2             = _mm256_add_pd(fjz2,tz);
1506
1507             /**************************
1508              * CALCULATE INTERACTIONS *
1509              **************************/
1510
1511             /* COULOMB ELECTROSTATICS */
1512             velec            = _mm256_mul_pd(qq33,rinv33);
1513             felec            = _mm256_mul_pd(velec,rinvsq33);
1514
1515             fscal            = felec;
1516
1517             /* Calculate temporary vectorial force */
1518             tx               = _mm256_mul_pd(fscal,dx33);
1519             ty               = _mm256_mul_pd(fscal,dy33);
1520             tz               = _mm256_mul_pd(fscal,dz33);
1521
1522             /* Update vectorial force */
1523             fix3             = _mm256_add_pd(fix3,tx);
1524             fiy3             = _mm256_add_pd(fiy3,ty);
1525             fiz3             = _mm256_add_pd(fiz3,tz);
1526
1527             fjx3             = _mm256_add_pd(fjx3,tx);
1528             fjy3             = _mm256_add_pd(fjy3,ty);
1529             fjz3             = _mm256_add_pd(fjz3,tz);
1530
1531             fjptrA             = f+j_coord_offsetA;
1532             fjptrB             = f+j_coord_offsetB;
1533             fjptrC             = f+j_coord_offsetC;
1534             fjptrD             = f+j_coord_offsetD;
1535
1536             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1537                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1538                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1539
1540             /* Inner loop uses 264 flops */
1541         }
1542
1543         if(jidx<j_index_end)
1544         {
1545
1546             /* Get j neighbor index, and coordinate index */
1547             jnrlistA         = jjnr[jidx];
1548             jnrlistB         = jjnr[jidx+1];
1549             jnrlistC         = jjnr[jidx+2];
1550             jnrlistD         = jjnr[jidx+3];
1551             /* Sign of each element will be negative for non-real atoms.
1552              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1553              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1554              */
1555             tmpmask0 = gmx_mm_castsi128_pd(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1556
1557             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1558             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1559             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1560
1561             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1562             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1563             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1564             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1565             j_coord_offsetA  = DIM*jnrA;
1566             j_coord_offsetB  = DIM*jnrB;
1567             j_coord_offsetC  = DIM*jnrC;
1568             j_coord_offsetD  = DIM*jnrD;
1569
1570             /* load j atom coordinates */
1571             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1572                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1573                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1574                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1575
1576             /* Calculate displacement vector */
1577             dx00             = _mm256_sub_pd(ix0,jx0);
1578             dy00             = _mm256_sub_pd(iy0,jy0);
1579             dz00             = _mm256_sub_pd(iz0,jz0);
1580             dx11             = _mm256_sub_pd(ix1,jx1);
1581             dy11             = _mm256_sub_pd(iy1,jy1);
1582             dz11             = _mm256_sub_pd(iz1,jz1);
1583             dx12             = _mm256_sub_pd(ix1,jx2);
1584             dy12             = _mm256_sub_pd(iy1,jy2);
1585             dz12             = _mm256_sub_pd(iz1,jz2);
1586             dx13             = _mm256_sub_pd(ix1,jx3);
1587             dy13             = _mm256_sub_pd(iy1,jy3);
1588             dz13             = _mm256_sub_pd(iz1,jz3);
1589             dx21             = _mm256_sub_pd(ix2,jx1);
1590             dy21             = _mm256_sub_pd(iy2,jy1);
1591             dz21             = _mm256_sub_pd(iz2,jz1);
1592             dx22             = _mm256_sub_pd(ix2,jx2);
1593             dy22             = _mm256_sub_pd(iy2,jy2);
1594             dz22             = _mm256_sub_pd(iz2,jz2);
1595             dx23             = _mm256_sub_pd(ix2,jx3);
1596             dy23             = _mm256_sub_pd(iy2,jy3);
1597             dz23             = _mm256_sub_pd(iz2,jz3);
1598             dx31             = _mm256_sub_pd(ix3,jx1);
1599             dy31             = _mm256_sub_pd(iy3,jy1);
1600             dz31             = _mm256_sub_pd(iz3,jz1);
1601             dx32             = _mm256_sub_pd(ix3,jx2);
1602             dy32             = _mm256_sub_pd(iy3,jy2);
1603             dz32             = _mm256_sub_pd(iz3,jz2);
1604             dx33             = _mm256_sub_pd(ix3,jx3);
1605             dy33             = _mm256_sub_pd(iy3,jy3);
1606             dz33             = _mm256_sub_pd(iz3,jz3);
1607
1608             /* Calculate squared distance and things based on it */
1609             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1610             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1611             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1612             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1613             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1614             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1615             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1616             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1617             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1618             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1619
1620             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1621             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1622             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1623             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1624             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1625             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1626             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1627             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1628             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1629
1630             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1631             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1632             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1633             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1634             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1635             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1636             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1637             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1638             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1639             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1640
1641             fjx0             = _mm256_setzero_pd();
1642             fjy0             = _mm256_setzero_pd();
1643             fjz0             = _mm256_setzero_pd();
1644             fjx1             = _mm256_setzero_pd();
1645             fjy1             = _mm256_setzero_pd();
1646             fjz1             = _mm256_setzero_pd();
1647             fjx2             = _mm256_setzero_pd();
1648             fjy2             = _mm256_setzero_pd();
1649             fjz2             = _mm256_setzero_pd();
1650             fjx3             = _mm256_setzero_pd();
1651             fjy3             = _mm256_setzero_pd();
1652             fjz3             = _mm256_setzero_pd();
1653
1654             /**************************
1655              * CALCULATE INTERACTIONS *
1656              **************************/
1657
1658             /* LENNARD-JONES DISPERSION/REPULSION */
1659
1660             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1661             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1662
1663             fscal            = fvdw;
1664
1665             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1666
1667             /* Calculate temporary vectorial force */
1668             tx               = _mm256_mul_pd(fscal,dx00);
1669             ty               = _mm256_mul_pd(fscal,dy00);
1670             tz               = _mm256_mul_pd(fscal,dz00);
1671
1672             /* Update vectorial force */
1673             fix0             = _mm256_add_pd(fix0,tx);
1674             fiy0             = _mm256_add_pd(fiy0,ty);
1675             fiz0             = _mm256_add_pd(fiz0,tz);
1676
1677             fjx0             = _mm256_add_pd(fjx0,tx);
1678             fjy0             = _mm256_add_pd(fjy0,ty);
1679             fjz0             = _mm256_add_pd(fjz0,tz);
1680
1681             /**************************
1682              * CALCULATE INTERACTIONS *
1683              **************************/
1684
1685             /* COULOMB ELECTROSTATICS */
1686             velec            = _mm256_mul_pd(qq11,rinv11);
1687             felec            = _mm256_mul_pd(velec,rinvsq11);
1688
1689             fscal            = felec;
1690
1691             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1692
1693             /* Calculate temporary vectorial force */
1694             tx               = _mm256_mul_pd(fscal,dx11);
1695             ty               = _mm256_mul_pd(fscal,dy11);
1696             tz               = _mm256_mul_pd(fscal,dz11);
1697
1698             /* Update vectorial force */
1699             fix1             = _mm256_add_pd(fix1,tx);
1700             fiy1             = _mm256_add_pd(fiy1,ty);
1701             fiz1             = _mm256_add_pd(fiz1,tz);
1702
1703             fjx1             = _mm256_add_pd(fjx1,tx);
1704             fjy1             = _mm256_add_pd(fjy1,ty);
1705             fjz1             = _mm256_add_pd(fjz1,tz);
1706
1707             /**************************
1708              * CALCULATE INTERACTIONS *
1709              **************************/
1710
1711             /* COULOMB ELECTROSTATICS */
1712             velec            = _mm256_mul_pd(qq12,rinv12);
1713             felec            = _mm256_mul_pd(velec,rinvsq12);
1714
1715             fscal            = felec;
1716
1717             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1718
1719             /* Calculate temporary vectorial force */
1720             tx               = _mm256_mul_pd(fscal,dx12);
1721             ty               = _mm256_mul_pd(fscal,dy12);
1722             tz               = _mm256_mul_pd(fscal,dz12);
1723
1724             /* Update vectorial force */
1725             fix1             = _mm256_add_pd(fix1,tx);
1726             fiy1             = _mm256_add_pd(fiy1,ty);
1727             fiz1             = _mm256_add_pd(fiz1,tz);
1728
1729             fjx2             = _mm256_add_pd(fjx2,tx);
1730             fjy2             = _mm256_add_pd(fjy2,ty);
1731             fjz2             = _mm256_add_pd(fjz2,tz);
1732
1733             /**************************
1734              * CALCULATE INTERACTIONS *
1735              **************************/
1736
1737             /* COULOMB ELECTROSTATICS */
1738             velec            = _mm256_mul_pd(qq13,rinv13);
1739             felec            = _mm256_mul_pd(velec,rinvsq13);
1740
1741             fscal            = felec;
1742
1743             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1744
1745             /* Calculate temporary vectorial force */
1746             tx               = _mm256_mul_pd(fscal,dx13);
1747             ty               = _mm256_mul_pd(fscal,dy13);
1748             tz               = _mm256_mul_pd(fscal,dz13);
1749
1750             /* Update vectorial force */
1751             fix1             = _mm256_add_pd(fix1,tx);
1752             fiy1             = _mm256_add_pd(fiy1,ty);
1753             fiz1             = _mm256_add_pd(fiz1,tz);
1754
1755             fjx3             = _mm256_add_pd(fjx3,tx);
1756             fjy3             = _mm256_add_pd(fjy3,ty);
1757             fjz3             = _mm256_add_pd(fjz3,tz);
1758
1759             /**************************
1760              * CALCULATE INTERACTIONS *
1761              **************************/
1762
1763             /* COULOMB ELECTROSTATICS */
1764             velec            = _mm256_mul_pd(qq21,rinv21);
1765             felec            = _mm256_mul_pd(velec,rinvsq21);
1766
1767             fscal            = felec;
1768
1769             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1770
1771             /* Calculate temporary vectorial force */
1772             tx               = _mm256_mul_pd(fscal,dx21);
1773             ty               = _mm256_mul_pd(fscal,dy21);
1774             tz               = _mm256_mul_pd(fscal,dz21);
1775
1776             /* Update vectorial force */
1777             fix2             = _mm256_add_pd(fix2,tx);
1778             fiy2             = _mm256_add_pd(fiy2,ty);
1779             fiz2             = _mm256_add_pd(fiz2,tz);
1780
1781             fjx1             = _mm256_add_pd(fjx1,tx);
1782             fjy1             = _mm256_add_pd(fjy1,ty);
1783             fjz1             = _mm256_add_pd(fjz1,tz);
1784
1785             /**************************
1786              * CALCULATE INTERACTIONS *
1787              **************************/
1788
1789             /* COULOMB ELECTROSTATICS */
1790             velec            = _mm256_mul_pd(qq22,rinv22);
1791             felec            = _mm256_mul_pd(velec,rinvsq22);
1792
1793             fscal            = felec;
1794
1795             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1796
1797             /* Calculate temporary vectorial force */
1798             tx               = _mm256_mul_pd(fscal,dx22);
1799             ty               = _mm256_mul_pd(fscal,dy22);
1800             tz               = _mm256_mul_pd(fscal,dz22);
1801
1802             /* Update vectorial force */
1803             fix2             = _mm256_add_pd(fix2,tx);
1804             fiy2             = _mm256_add_pd(fiy2,ty);
1805             fiz2             = _mm256_add_pd(fiz2,tz);
1806
1807             fjx2             = _mm256_add_pd(fjx2,tx);
1808             fjy2             = _mm256_add_pd(fjy2,ty);
1809             fjz2             = _mm256_add_pd(fjz2,tz);
1810
1811             /**************************
1812              * CALCULATE INTERACTIONS *
1813              **************************/
1814
1815             /* COULOMB ELECTROSTATICS */
1816             velec            = _mm256_mul_pd(qq23,rinv23);
1817             felec            = _mm256_mul_pd(velec,rinvsq23);
1818
1819             fscal            = felec;
1820
1821             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1822
1823             /* Calculate temporary vectorial force */
1824             tx               = _mm256_mul_pd(fscal,dx23);
1825             ty               = _mm256_mul_pd(fscal,dy23);
1826             tz               = _mm256_mul_pd(fscal,dz23);
1827
1828             /* Update vectorial force */
1829             fix2             = _mm256_add_pd(fix2,tx);
1830             fiy2             = _mm256_add_pd(fiy2,ty);
1831             fiz2             = _mm256_add_pd(fiz2,tz);
1832
1833             fjx3             = _mm256_add_pd(fjx3,tx);
1834             fjy3             = _mm256_add_pd(fjy3,ty);
1835             fjz3             = _mm256_add_pd(fjz3,tz);
1836
1837             /**************************
1838              * CALCULATE INTERACTIONS *
1839              **************************/
1840
1841             /* COULOMB ELECTROSTATICS */
1842             velec            = _mm256_mul_pd(qq31,rinv31);
1843             felec            = _mm256_mul_pd(velec,rinvsq31);
1844
1845             fscal            = felec;
1846
1847             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1848
1849             /* Calculate temporary vectorial force */
1850             tx               = _mm256_mul_pd(fscal,dx31);
1851             ty               = _mm256_mul_pd(fscal,dy31);
1852             tz               = _mm256_mul_pd(fscal,dz31);
1853
1854             /* Update vectorial force */
1855             fix3             = _mm256_add_pd(fix3,tx);
1856             fiy3             = _mm256_add_pd(fiy3,ty);
1857             fiz3             = _mm256_add_pd(fiz3,tz);
1858
1859             fjx1             = _mm256_add_pd(fjx1,tx);
1860             fjy1             = _mm256_add_pd(fjy1,ty);
1861             fjz1             = _mm256_add_pd(fjz1,tz);
1862
1863             /**************************
1864              * CALCULATE INTERACTIONS *
1865              **************************/
1866
1867             /* COULOMB ELECTROSTATICS */
1868             velec            = _mm256_mul_pd(qq32,rinv32);
1869             felec            = _mm256_mul_pd(velec,rinvsq32);
1870
1871             fscal            = felec;
1872
1873             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1874
1875             /* Calculate temporary vectorial force */
1876             tx               = _mm256_mul_pd(fscal,dx32);
1877             ty               = _mm256_mul_pd(fscal,dy32);
1878             tz               = _mm256_mul_pd(fscal,dz32);
1879
1880             /* Update vectorial force */
1881             fix3             = _mm256_add_pd(fix3,tx);
1882             fiy3             = _mm256_add_pd(fiy3,ty);
1883             fiz3             = _mm256_add_pd(fiz3,tz);
1884
1885             fjx2             = _mm256_add_pd(fjx2,tx);
1886             fjy2             = _mm256_add_pd(fjy2,ty);
1887             fjz2             = _mm256_add_pd(fjz2,tz);
1888
1889             /**************************
1890              * CALCULATE INTERACTIONS *
1891              **************************/
1892
1893             /* COULOMB ELECTROSTATICS */
1894             velec            = _mm256_mul_pd(qq33,rinv33);
1895             felec            = _mm256_mul_pd(velec,rinvsq33);
1896
1897             fscal            = felec;
1898
1899             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1900
1901             /* Calculate temporary vectorial force */
1902             tx               = _mm256_mul_pd(fscal,dx33);
1903             ty               = _mm256_mul_pd(fscal,dy33);
1904             tz               = _mm256_mul_pd(fscal,dz33);
1905
1906             /* Update vectorial force */
1907             fix3             = _mm256_add_pd(fix3,tx);
1908             fiy3             = _mm256_add_pd(fiy3,ty);
1909             fiz3             = _mm256_add_pd(fiz3,tz);
1910
1911             fjx3             = _mm256_add_pd(fjx3,tx);
1912             fjy3             = _mm256_add_pd(fjy3,ty);
1913             fjz3             = _mm256_add_pd(fjz3,tz);
1914
1915             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1916             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1917             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1918             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1919
1920             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1921                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1922                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1923
1924             /* Inner loop uses 264 flops */
1925         }
1926
1927         /* End of innermost loop */
1928
1929         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1930                                                  f+i_coord_offset,fshift+i_shift_offset);
1931
1932         /* Increment number of inner iterations */
1933         inneriter                  += j_index_end - j_index_start;
1934
1935         /* Outer loop uses 24 flops */
1936     }
1937
1938     /* Increment number of outer iterations */
1939     outeriter        += nri;
1940
1941     /* Update outer/inner flops */
1942
1943     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
1944 }