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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  95     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  97     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  99     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
 101     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 119     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 120     __m256d          dummy_mask,cutoff_mask;
 121     __m128           tmpmask0,tmpmask1;
 122     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 123     __m256d          one     = _mm256_set1_pd(1.0);
 124     __m256d          two     = _mm256_set1_pd(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm256_set1_pd(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     nvdwtype         = fr->ntype;
 139     vdwparam         = fr->nbfp;
 140     vdwtype          = mdatoms->typeA;
 141
 142     /* Setup water-specific parameters */
 143     inr              = nlist->iinr[0];
 144     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 145     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 146     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 147     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 148
 149     jq1              = _mm256_set1_pd(charge[inr+1]);
 150     jq2              = _mm256_set1_pd(charge[inr+2]);
 151     jq3              = _mm256_set1_pd(charge[inr+3]);
 152     vdwjidx0A        = 2*vdwtype[inr+0];
 153     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 154     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 155     qq11             = _mm256_mul_pd(iq1,jq1);
 156     qq12             = _mm256_mul_pd(iq1,jq2);
 157     qq13             = _mm256_mul_pd(iq1,jq3);
 158     qq21             = _mm256_mul_pd(iq2,jq1);
 159     qq22             = _mm256_mul_pd(iq2,jq2);
 160     qq23             = _mm256_mul_pd(iq2,jq3);
 161     qq31             = _mm256_mul_pd(iq3,jq1);
 162     qq32             = _mm256_mul_pd(iq3,jq2);
 163     qq33             = _mm256_mul_pd(iq3,jq3);
 164
 165     /* Avoid stupid compiler warnings */
 166     jnrA = jnrB = jnrC = jnrD = 0;
 167     j_coord_offsetA = 0;
 168     j_coord_offsetB = 0;
 169     j_coord_offsetC = 0;
 170     j_coord_offsetD = 0;
 171
 172     outeriter        = 0;
 173     inneriter        = 0;
 174
 175     for(iidx=0;iidx<4*DIM;iidx++)
 176     {
 177         scratch[iidx] = 0.0;
 178     }
 179
 180     /* Start outer loop over neighborlists */
 181     for(iidx=0; iidx<nri; iidx++)
 182     {
 183         /* Load shift vector for this list */
 184         i_shift_offset   = DIM*shiftidx[iidx];
 185
 186         /* Load limits for loop over neighbors */
 187         j_index_start    = jindex[iidx];
 188         j_index_end      = jindex[iidx+1];
 189
 190         /* Get outer coordinate index */
 191         inr              = iinr[iidx];
 192         i_coord_offset   = DIM*inr;
 193
 194         /* Load i particle coords and add shift vector */
 195         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 196                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 197
 198         fix0             = _mm256_setzero_pd();
 199         fiy0             = _mm256_setzero_pd();
 200         fiz0             = _mm256_setzero_pd();
 201         fix1             = _mm256_setzero_pd();
 202         fiy1             = _mm256_setzero_pd();
 203         fiz1             = _mm256_setzero_pd();
 204         fix2             = _mm256_setzero_pd();
 205         fiy2             = _mm256_setzero_pd();
 206         fiz2             = _mm256_setzero_pd();
 207         fix3             = _mm256_setzero_pd();
 208         fiy3             = _mm256_setzero_pd();
 209         fiz3             = _mm256_setzero_pd();
 210
 211         /* Reset potential sums */
 212         velecsum         = _mm256_setzero_pd();
 213         vvdwsum          = _mm256_setzero_pd();
 214
 215         /* Start inner kernel loop */
 216         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 217         {
 218
 219             /* Get j neighbor index, and coordinate index */
 220             jnrA             = jjnr[jidx];
 221             jnrB             = jjnr[jidx+1];
 222             jnrC             = jjnr[jidx+2];
 223             jnrD             = jjnr[jidx+3];
 224             j_coord_offsetA  = DIM*jnrA;
 225             j_coord_offsetB  = DIM*jnrB;
 226             j_coord_offsetC  = DIM*jnrC;
 227             j_coord_offsetD  = DIM*jnrD;
 228
 229             /* load j atom coordinates */
 230             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 231                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 232                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 233                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 234
 235             /* Calculate displacement vector */
 236             dx00             = _mm256_sub_pd(ix0,jx0);
 237             dy00             = _mm256_sub_pd(iy0,jy0);
 238             dz00             = _mm256_sub_pd(iz0,jz0);
 239             dx11             = _mm256_sub_pd(ix1,jx1);
 240             dy11             = _mm256_sub_pd(iy1,jy1);
 241             dz11             = _mm256_sub_pd(iz1,jz1);
 242             dx12             = _mm256_sub_pd(ix1,jx2);
 243             dy12             = _mm256_sub_pd(iy1,jy2);
 244             dz12             = _mm256_sub_pd(iz1,jz2);
 245             dx13             = _mm256_sub_pd(ix1,jx3);
 246             dy13             = _mm256_sub_pd(iy1,jy3);
 247             dz13             = _mm256_sub_pd(iz1,jz3);
 248             dx21             = _mm256_sub_pd(ix2,jx1);
 249             dy21             = _mm256_sub_pd(iy2,jy1);
 250             dz21             = _mm256_sub_pd(iz2,jz1);
 251             dx22             = _mm256_sub_pd(ix2,jx2);
 252             dy22             = _mm256_sub_pd(iy2,jy2);
 253             dz22             = _mm256_sub_pd(iz2,jz2);
 254             dx23             = _mm256_sub_pd(ix2,jx3);
 255             dy23             = _mm256_sub_pd(iy2,jy3);
 256             dz23             = _mm256_sub_pd(iz2,jz3);
 257             dx31             = _mm256_sub_pd(ix3,jx1);
 258             dy31             = _mm256_sub_pd(iy3,jy1);
 259             dz31             = _mm256_sub_pd(iz3,jz1);
 260             dx32             = _mm256_sub_pd(ix3,jx2);
 261             dy32             = _mm256_sub_pd(iy3,jy2);
 262             dz32             = _mm256_sub_pd(iz3,jz2);
 263             dx33             = _mm256_sub_pd(ix3,jx3);
 264             dy33             = _mm256_sub_pd(iy3,jy3);
 265             dz33             = _mm256_sub_pd(iz3,jz3);
 266
 267             /* Calculate squared distance and things based on it */
 268             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 269             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 270             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 271             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 272             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 273             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 274             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 275             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 276             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 277             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 278
 279             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 280             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 281             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 282             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 283             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 284             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 285             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 286             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 287             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 288
 289             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 290             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 291             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 292             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 293             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 294             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 295             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 296             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 297             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 298             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 299
 300             fjx0             = _mm256_setzero_pd();
 301             fjy0             = _mm256_setzero_pd();
 302             fjz0             = _mm256_setzero_pd();
 303             fjx1             = _mm256_setzero_pd();
 304             fjy1             = _mm256_setzero_pd();
 305             fjz1             = _mm256_setzero_pd();
 306             fjx2             = _mm256_setzero_pd();
 307             fjy2             = _mm256_setzero_pd();
 308             fjz2             = _mm256_setzero_pd();
 309             fjx3             = _mm256_setzero_pd();
 310             fjy3             = _mm256_setzero_pd();
 311             fjz3             = _mm256_setzero_pd();
 312
 313             /**************************
 314              * CALCULATE INTERACTIONS *
 315              **************************/
 316
 317             /* LENNARD-JONES DISPERSION/REPULSION */
 318
 319             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 320             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 321             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 322             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 323             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 324
 325             /* Update potential sum for this i atom from the interaction with this j atom. */
 326             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 327
 328             fscal            = fvdw;
 329
 330             /* Calculate temporary vectorial force */
 331             tx               = _mm256_mul_pd(fscal,dx00);
 332             ty               = _mm256_mul_pd(fscal,dy00);
 333             tz               = _mm256_mul_pd(fscal,dz00);
 334
 335             /* Update vectorial force */
 336             fix0             = _mm256_add_pd(fix0,tx);
 337             fiy0             = _mm256_add_pd(fiy0,ty);
 338             fiz0             = _mm256_add_pd(fiz0,tz);
 339
 340             fjx0             = _mm256_add_pd(fjx0,tx);
 341             fjy0             = _mm256_add_pd(fjy0,ty);
 342             fjz0             = _mm256_add_pd(fjz0,tz);
 343
 344             /**************************
 345              * CALCULATE INTERACTIONS *
 346              **************************/
 347
 348             /* COULOMB ELECTROSTATICS */
 349             velec            = _mm256_mul_pd(qq11,rinv11);
 350             felec            = _mm256_mul_pd(velec,rinvsq11);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             velecsum         = _mm256_add_pd(velecsum,velec);
 354
 355             fscal            = felec;
 356
 357             /* Calculate temporary vectorial force */
 358             tx               = _mm256_mul_pd(fscal,dx11);
 359             ty               = _mm256_mul_pd(fscal,dy11);
 360             tz               = _mm256_mul_pd(fscal,dz11);
 361
 362             /* Update vectorial force */
 363             fix1             = _mm256_add_pd(fix1,tx);
 364             fiy1             = _mm256_add_pd(fiy1,ty);
 365             fiz1             = _mm256_add_pd(fiz1,tz);
 366
 367             fjx1             = _mm256_add_pd(fjx1,tx);
 368             fjy1             = _mm256_add_pd(fjy1,ty);
 369             fjz1             = _mm256_add_pd(fjz1,tz);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             /* COULOMB ELECTROSTATICS */
 376             velec            = _mm256_mul_pd(qq12,rinv12);
 377             felec            = _mm256_mul_pd(velec,rinvsq12);
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             velecsum         = _mm256_add_pd(velecsum,velec);
 381
 382             fscal            = felec;
 383
 384             /* Calculate temporary vectorial force */
 385             tx               = _mm256_mul_pd(fscal,dx12);
 386             ty               = _mm256_mul_pd(fscal,dy12);
 387             tz               = _mm256_mul_pd(fscal,dz12);
 388
 389             /* Update vectorial force */
 390             fix1             = _mm256_add_pd(fix1,tx);
 391             fiy1             = _mm256_add_pd(fiy1,ty);
 392             fiz1             = _mm256_add_pd(fiz1,tz);
 393
 394             fjx2             = _mm256_add_pd(fjx2,tx);
 395             fjy2             = _mm256_add_pd(fjy2,ty);
 396             fjz2             = _mm256_add_pd(fjz2,tz);
 397
 398             /**************************
 399              * CALCULATE INTERACTIONS *
 400              **************************/
 401
 402             /* COULOMB ELECTROSTATICS */
 403             velec            = _mm256_mul_pd(qq13,rinv13);
 404             felec            = _mm256_mul_pd(velec,rinvsq13);
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm256_add_pd(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm256_mul_pd(fscal,dx13);
 413             ty               = _mm256_mul_pd(fscal,dy13);
 414             tz               = _mm256_mul_pd(fscal,dz13);
 415
 416             /* Update vectorial force */
 417             fix1             = _mm256_add_pd(fix1,tx);
 418             fiy1             = _mm256_add_pd(fiy1,ty);
 419             fiz1             = _mm256_add_pd(fiz1,tz);
 420
 421             fjx3             = _mm256_add_pd(fjx3,tx);
 422             fjy3             = _mm256_add_pd(fjy3,ty);
 423             fjz3             = _mm256_add_pd(fjz3,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             /* COULOMB ELECTROSTATICS */
 430             velec            = _mm256_mul_pd(qq21,rinv21);
 431             felec            = _mm256_mul_pd(velec,rinvsq21);
 432
 433             /* Update potential sum for this i atom from the interaction with this j atom. */
 434             velecsum         = _mm256_add_pd(velecsum,velec);
 435
 436             fscal            = felec;
 437
 438             /* Calculate temporary vectorial force */
 439             tx               = _mm256_mul_pd(fscal,dx21);
 440             ty               = _mm256_mul_pd(fscal,dy21);
 441             tz               = _mm256_mul_pd(fscal,dz21);
 442
 443             /* Update vectorial force */
 444             fix2             = _mm256_add_pd(fix2,tx);
 445             fiy2             = _mm256_add_pd(fiy2,ty);
 446             fiz2             = _mm256_add_pd(fiz2,tz);
 447
 448             fjx1             = _mm256_add_pd(fjx1,tx);
 449             fjy1             = _mm256_add_pd(fjy1,ty);
 450             fjz1             = _mm256_add_pd(fjz1,tz);
 451
 452             /**************************
 453              * CALCULATE INTERACTIONS *
 454              **************************/
 455
 456             /* COULOMB ELECTROSTATICS */
 457             velec            = _mm256_mul_pd(qq22,rinv22);
 458             felec            = _mm256_mul_pd(velec,rinvsq22);
 459
 460             /* Update potential sum for this i atom from the interaction with this j atom. */
 461             velecsum         = _mm256_add_pd(velecsum,velec);
 462
 463             fscal            = felec;
 464
 465             /* Calculate temporary vectorial force */
 466             tx               = _mm256_mul_pd(fscal,dx22);
 467             ty               = _mm256_mul_pd(fscal,dy22);
 468             tz               = _mm256_mul_pd(fscal,dz22);
 469
 470             /* Update vectorial force */
 471             fix2             = _mm256_add_pd(fix2,tx);
 472             fiy2             = _mm256_add_pd(fiy2,ty);
 473             fiz2             = _mm256_add_pd(fiz2,tz);
 474
 475             fjx2             = _mm256_add_pd(fjx2,tx);
 476             fjy2             = _mm256_add_pd(fjy2,ty);
 477             fjz2             = _mm256_add_pd(fjz2,tz);
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             /* COULOMB ELECTROSTATICS */
 484             velec            = _mm256_mul_pd(qq23,rinv23);
 485             felec            = _mm256_mul_pd(velec,rinvsq23);
 486
 487             /* Update potential sum for this i atom from the interaction with this j atom. */
 488             velecsum         = _mm256_add_pd(velecsum,velec);
 489
 490             fscal            = felec;
 491
 492             /* Calculate temporary vectorial force */
 493             tx               = _mm256_mul_pd(fscal,dx23);
 494             ty               = _mm256_mul_pd(fscal,dy23);
 495             tz               = _mm256_mul_pd(fscal,dz23);
 496
 497             /* Update vectorial force */
 498             fix2             = _mm256_add_pd(fix2,tx);
 499             fiy2             = _mm256_add_pd(fiy2,ty);
 500             fiz2             = _mm256_add_pd(fiz2,tz);
 501
 502             fjx3             = _mm256_add_pd(fjx3,tx);
 503             fjy3             = _mm256_add_pd(fjy3,ty);
 504             fjz3             = _mm256_add_pd(fjz3,tz);
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             /* COULOMB ELECTROSTATICS */
 511             velec            = _mm256_mul_pd(qq31,rinv31);
 512             felec            = _mm256_mul_pd(velec,rinvsq31);
 513
 514             /* Update potential sum for this i atom from the interaction with this j atom. */
 515             velecsum         = _mm256_add_pd(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519             /* Calculate temporary vectorial force */
 520             tx               = _mm256_mul_pd(fscal,dx31);
 521             ty               = _mm256_mul_pd(fscal,dy31);
 522             tz               = _mm256_mul_pd(fscal,dz31);
 523
 524             /* Update vectorial force */
 525             fix3             = _mm256_add_pd(fix3,tx);
 526             fiy3             = _mm256_add_pd(fiy3,ty);
 527             fiz3             = _mm256_add_pd(fiz3,tz);
 528
 529             fjx1             = _mm256_add_pd(fjx1,tx);
 530             fjy1             = _mm256_add_pd(fjy1,ty);
 531             fjz1             = _mm256_add_pd(fjz1,tz);
 532
 533             /**************************
 534              * CALCULATE INTERACTIONS *
 535              **************************/
 536
 537             /* COULOMB ELECTROSTATICS */
 538             velec            = _mm256_mul_pd(qq32,rinv32);
 539             felec            = _mm256_mul_pd(velec,rinvsq32);
 540
 541             /* Update potential sum for this i atom from the interaction with this j atom. */
 542             velecsum         = _mm256_add_pd(velecsum,velec);
 543
 544             fscal            = felec;
 545
 546             /* Calculate temporary vectorial force */
 547             tx               = _mm256_mul_pd(fscal,dx32);
 548             ty               = _mm256_mul_pd(fscal,dy32);
 549             tz               = _mm256_mul_pd(fscal,dz32);
 550
 551             /* Update vectorial force */
 552             fix3             = _mm256_add_pd(fix3,tx);
 553             fiy3             = _mm256_add_pd(fiy3,ty);
 554             fiz3             = _mm256_add_pd(fiz3,tz);
 555
 556             fjx2             = _mm256_add_pd(fjx2,tx);
 557             fjy2             = _mm256_add_pd(fjy2,ty);
 558             fjz2             = _mm256_add_pd(fjz2,tz);
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             /* COULOMB ELECTROSTATICS */
 565             velec            = _mm256_mul_pd(qq33,rinv33);
 566             felec            = _mm256_mul_pd(velec,rinvsq33);
 567
 568             /* Update potential sum for this i atom from the interaction with this j atom. */
 569             velecsum         = _mm256_add_pd(velecsum,velec);
 570
 571             fscal            = felec;
 572
 573             /* Calculate temporary vectorial force */
 574             tx               = _mm256_mul_pd(fscal,dx33);
 575             ty               = _mm256_mul_pd(fscal,dy33);
 576             tz               = _mm256_mul_pd(fscal,dz33);
 577
 578             /* Update vectorial force */
 579             fix3             = _mm256_add_pd(fix3,tx);
 580             fiy3             = _mm256_add_pd(fiy3,ty);
 581             fiz3             = _mm256_add_pd(fiz3,tz);
 582
 583             fjx3             = _mm256_add_pd(fjx3,tx);
 584             fjy3             = _mm256_add_pd(fjy3,ty);
 585             fjz3             = _mm256_add_pd(fjz3,tz);
 586
 587             fjptrA             = f+j_coord_offsetA;
 588             fjptrB             = f+j_coord_offsetB;
 589             fjptrC             = f+j_coord_offsetC;
 590             fjptrD             = f+j_coord_offsetD;
 591
 592             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 593                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 594                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 595
 596             /* Inner loop uses 278 flops */
 597         }
 598
 599         if(jidx<j_index_end)
 600         {
 601
 602             /* Get j neighbor index, and coordinate index */
 603             jnrlistA         = jjnr[jidx];
 604             jnrlistB         = jjnr[jidx+1];
 605             jnrlistC         = jjnr[jidx+2];
 606             jnrlistD         = jjnr[jidx+3];
 607             /* Sign of each element will be negative for non-real atoms.
 608              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 609              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 610              */
 611             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 612
 613             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 614             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 615             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 616
 617             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 618             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 619             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 620             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 621             j_coord_offsetA  = DIM*jnrA;
 622             j_coord_offsetB  = DIM*jnrB;
 623             j_coord_offsetC  = DIM*jnrC;
 624             j_coord_offsetD  = DIM*jnrD;
 625
 626             /* load j atom coordinates */
 627             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 628                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 629                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 630                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 631
 632             /* Calculate displacement vector */
 633             dx00             = _mm256_sub_pd(ix0,jx0);
 634             dy00             = _mm256_sub_pd(iy0,jy0);
 635             dz00             = _mm256_sub_pd(iz0,jz0);
 636             dx11             = _mm256_sub_pd(ix1,jx1);
 637             dy11             = _mm256_sub_pd(iy1,jy1);
 638             dz11             = _mm256_sub_pd(iz1,jz1);
 639             dx12             = _mm256_sub_pd(ix1,jx2);
 640             dy12             = _mm256_sub_pd(iy1,jy2);
 641             dz12             = _mm256_sub_pd(iz1,jz2);
 642             dx13             = _mm256_sub_pd(ix1,jx3);
 643             dy13             = _mm256_sub_pd(iy1,jy3);
 644             dz13             = _mm256_sub_pd(iz1,jz3);
 645             dx21             = _mm256_sub_pd(ix2,jx1);
 646             dy21             = _mm256_sub_pd(iy2,jy1);
 647             dz21             = _mm256_sub_pd(iz2,jz1);
 648             dx22             = _mm256_sub_pd(ix2,jx2);
 649             dy22             = _mm256_sub_pd(iy2,jy2);
 650             dz22             = _mm256_sub_pd(iz2,jz2);
 651             dx23             = _mm256_sub_pd(ix2,jx3);
 652             dy23             = _mm256_sub_pd(iy2,jy3);
 653             dz23             = _mm256_sub_pd(iz2,jz3);
 654             dx31             = _mm256_sub_pd(ix3,jx1);
 655             dy31             = _mm256_sub_pd(iy3,jy1);
 656             dz31             = _mm256_sub_pd(iz3,jz1);
 657             dx32             = _mm256_sub_pd(ix3,jx2);
 658             dy32             = _mm256_sub_pd(iy3,jy2);
 659             dz32             = _mm256_sub_pd(iz3,jz2);
 660             dx33             = _mm256_sub_pd(ix3,jx3);
 661             dy33             = _mm256_sub_pd(iy3,jy3);
 662             dz33             = _mm256_sub_pd(iz3,jz3);
 663
 664             /* Calculate squared distance and things based on it */
 665             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 666             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 667             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 668             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 669             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 670             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 671             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 672             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 673             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 674             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 675
 676             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 677             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 678             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 679             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 680             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 681             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 682             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 683             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 684             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 685
 686             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 687             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 688             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 689             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 690             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 691             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 692             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 693             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 694             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 695             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 696
 697             fjx0             = _mm256_setzero_pd();
 698             fjy0             = _mm256_setzero_pd();
 699             fjz0             = _mm256_setzero_pd();
 700             fjx1             = _mm256_setzero_pd();
 701             fjy1             = _mm256_setzero_pd();
 702             fjz1             = _mm256_setzero_pd();
 703             fjx2             = _mm256_setzero_pd();
 704             fjy2             = _mm256_setzero_pd();
 705             fjz2             = _mm256_setzero_pd();
 706             fjx3             = _mm256_setzero_pd();
 707             fjy3             = _mm256_setzero_pd();
 708             fjz3             = _mm256_setzero_pd();
 709
 710             /**************************
 711              * CALCULATE INTERACTIONS *
 712              **************************/
 713
 714             /* LENNARD-JONES DISPERSION/REPULSION */
 715
 716             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 717             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 718             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 719             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 720             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 721
 722             /* Update potential sum for this i atom from the interaction with this j atom. */
 723             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 724             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 725
 726             fscal            = fvdw;
 727
 728             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 729
 730             /* Calculate temporary vectorial force */
 731             tx               = _mm256_mul_pd(fscal,dx00);
 732             ty               = _mm256_mul_pd(fscal,dy00);
 733             tz               = _mm256_mul_pd(fscal,dz00);
 734
 735             /* Update vectorial force */
 736             fix0             = _mm256_add_pd(fix0,tx);
 737             fiy0             = _mm256_add_pd(fiy0,ty);
 738             fiz0             = _mm256_add_pd(fiz0,tz);
 739
 740             fjx0             = _mm256_add_pd(fjx0,tx);
 741             fjy0             = _mm256_add_pd(fjy0,ty);
 742             fjz0             = _mm256_add_pd(fjz0,tz);
 743
 744             /**************************
 745              * CALCULATE INTERACTIONS *
 746              **************************/
 747
 748             /* COULOMB ELECTROSTATICS */
 749             velec            = _mm256_mul_pd(qq11,rinv11);
 750             felec            = _mm256_mul_pd(velec,rinvsq11);
 751
 752             /* Update potential sum for this i atom from the interaction with this j atom. */
 753             velec            = _mm256_andnot_pd(dummy_mask,velec);
 754             velecsum         = _mm256_add_pd(velecsum,velec);
 755
 756             fscal            = felec;
 757
 758             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 759
 760             /* Calculate temporary vectorial force */
 761             tx               = _mm256_mul_pd(fscal,dx11);
 762             ty               = _mm256_mul_pd(fscal,dy11);
 763             tz               = _mm256_mul_pd(fscal,dz11);
 764
 765             /* Update vectorial force */
 766             fix1             = _mm256_add_pd(fix1,tx);
 767             fiy1             = _mm256_add_pd(fiy1,ty);
 768             fiz1             = _mm256_add_pd(fiz1,tz);
 769
 770             fjx1             = _mm256_add_pd(fjx1,tx);
 771             fjy1             = _mm256_add_pd(fjy1,ty);
 772             fjz1             = _mm256_add_pd(fjz1,tz);
 773
 774             /**************************
 775              * CALCULATE INTERACTIONS *
 776              **************************/
 777
 778             /* COULOMB ELECTROSTATICS */
 779             velec            = _mm256_mul_pd(qq12,rinv12);
 780             felec            = _mm256_mul_pd(velec,rinvsq12);
 781
 782             /* Update potential sum for this i atom from the interaction with this j atom. */
 783             velec            = _mm256_andnot_pd(dummy_mask,velec);
 784             velecsum         = _mm256_add_pd(velecsum,velec);
 785
 786             fscal            = felec;
 787
 788             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 789
 790             /* Calculate temporary vectorial force */
 791             tx               = _mm256_mul_pd(fscal,dx12);
 792             ty               = _mm256_mul_pd(fscal,dy12);
 793             tz               = _mm256_mul_pd(fscal,dz12);
 794
 795             /* Update vectorial force */
 796             fix1             = _mm256_add_pd(fix1,tx);
 797             fiy1             = _mm256_add_pd(fiy1,ty);
 798             fiz1             = _mm256_add_pd(fiz1,tz);
 799
 800             fjx2             = _mm256_add_pd(fjx2,tx);
 801             fjy2             = _mm256_add_pd(fjy2,ty);
 802             fjz2             = _mm256_add_pd(fjz2,tz);
 803
 804             /**************************
 805              * CALCULATE INTERACTIONS *
 806              **************************/
 807
 808             /* COULOMB ELECTROSTATICS */
 809             velec            = _mm256_mul_pd(qq13,rinv13);
 810             felec            = _mm256_mul_pd(velec,rinvsq13);
 811
 812             /* Update potential sum for this i atom from the interaction with this j atom. */
 813             velec            = _mm256_andnot_pd(dummy_mask,velec);
 814             velecsum         = _mm256_add_pd(velecsum,velec);
 815
 816             fscal            = felec;
 817
 818             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 819
 820             /* Calculate temporary vectorial force */
 821             tx               = _mm256_mul_pd(fscal,dx13);
 822             ty               = _mm256_mul_pd(fscal,dy13);
 823             tz               = _mm256_mul_pd(fscal,dz13);
 824
 825             /* Update vectorial force */
 826             fix1             = _mm256_add_pd(fix1,tx);
 827             fiy1             = _mm256_add_pd(fiy1,ty);
 828             fiz1             = _mm256_add_pd(fiz1,tz);
 829
 830             fjx3             = _mm256_add_pd(fjx3,tx);
 831             fjy3             = _mm256_add_pd(fjy3,ty);
 832             fjz3             = _mm256_add_pd(fjz3,tz);
 833
 834             /**************************
 835              * CALCULATE INTERACTIONS *
 836              **************************/
 837
 838             /* COULOMB ELECTROSTATICS */
 839             velec            = _mm256_mul_pd(qq21,rinv21);
 840             felec            = _mm256_mul_pd(velec,rinvsq21);
 841
 842             /* Update potential sum for this i atom from the interaction with this j atom. */
 843             velec            = _mm256_andnot_pd(dummy_mask,velec);
 844             velecsum         = _mm256_add_pd(velecsum,velec);
 845
 846             fscal            = felec;
 847
 848             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 849
 850             /* Calculate temporary vectorial force */
 851             tx               = _mm256_mul_pd(fscal,dx21);
 852             ty               = _mm256_mul_pd(fscal,dy21);
 853             tz               = _mm256_mul_pd(fscal,dz21);
 854
 855             /* Update vectorial force */
 856             fix2             = _mm256_add_pd(fix2,tx);
 857             fiy2             = _mm256_add_pd(fiy2,ty);
 858             fiz2             = _mm256_add_pd(fiz2,tz);
 859
 860             fjx1             = _mm256_add_pd(fjx1,tx);
 861             fjy1             = _mm256_add_pd(fjy1,ty);
 862             fjz1             = _mm256_add_pd(fjz1,tz);
 863
 864             /**************************
 865              * CALCULATE INTERACTIONS *
 866              **************************/
 867
 868             /* COULOMB ELECTROSTATICS */
 869             velec            = _mm256_mul_pd(qq22,rinv22);
 870             felec            = _mm256_mul_pd(velec,rinvsq22);
 871
 872             /* Update potential sum for this i atom from the interaction with this j atom. */
 873             velec            = _mm256_andnot_pd(dummy_mask,velec);
 874             velecsum         = _mm256_add_pd(velecsum,velec);
 875
 876             fscal            = felec;
 877
 878             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 879
 880             /* Calculate temporary vectorial force */
 881             tx               = _mm256_mul_pd(fscal,dx22);
 882             ty               = _mm256_mul_pd(fscal,dy22);
 883             tz               = _mm256_mul_pd(fscal,dz22);
 884
 885             /* Update vectorial force */
 886             fix2             = _mm256_add_pd(fix2,tx);
 887             fiy2             = _mm256_add_pd(fiy2,ty);
 888             fiz2             = _mm256_add_pd(fiz2,tz);
 889
 890             fjx2             = _mm256_add_pd(fjx2,tx);
 891             fjy2             = _mm256_add_pd(fjy2,ty);
 892             fjz2             = _mm256_add_pd(fjz2,tz);
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             /* COULOMB ELECTROSTATICS */
 899             velec            = _mm256_mul_pd(qq23,rinv23);
 900             felec            = _mm256_mul_pd(velec,rinvsq23);
 901
 902             /* Update potential sum for this i atom from the interaction with this j atom. */
 903             velec            = _mm256_andnot_pd(dummy_mask,velec);
 904             velecsum         = _mm256_add_pd(velecsum,velec);
 905
 906             fscal            = felec;
 907
 908             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 909
 910             /* Calculate temporary vectorial force */
 911             tx               = _mm256_mul_pd(fscal,dx23);
 912             ty               = _mm256_mul_pd(fscal,dy23);
 913             tz               = _mm256_mul_pd(fscal,dz23);
 914
 915             /* Update vectorial force */
 916             fix2             = _mm256_add_pd(fix2,tx);
 917             fiy2             = _mm256_add_pd(fiy2,ty);
 918             fiz2             = _mm256_add_pd(fiz2,tz);
 919
 920             fjx3             = _mm256_add_pd(fjx3,tx);
 921             fjy3             = _mm256_add_pd(fjy3,ty);
 922             fjz3             = _mm256_add_pd(fjz3,tz);
 923
 924             /**************************
 925              * CALCULATE INTERACTIONS *
 926              **************************/
 927
 928             /* COULOMB ELECTROSTATICS */
 929             velec            = _mm256_mul_pd(qq31,rinv31);
 930             felec            = _mm256_mul_pd(velec,rinvsq31);
 931
 932             /* Update potential sum for this i atom from the interaction with this j atom. */
 933             velec            = _mm256_andnot_pd(dummy_mask,velec);
 934             velecsum         = _mm256_add_pd(velecsum,velec);
 935
 936             fscal            = felec;
 937
 938             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 939
 940             /* Calculate temporary vectorial force */
 941             tx               = _mm256_mul_pd(fscal,dx31);
 942             ty               = _mm256_mul_pd(fscal,dy31);
 943             tz               = _mm256_mul_pd(fscal,dz31);
 944
 945             /* Update vectorial force */
 946             fix3             = _mm256_add_pd(fix3,tx);
 947             fiy3             = _mm256_add_pd(fiy3,ty);
 948             fiz3             = _mm256_add_pd(fiz3,tz);
 949
 950             fjx1             = _mm256_add_pd(fjx1,tx);
 951             fjy1             = _mm256_add_pd(fjy1,ty);
 952             fjz1             = _mm256_add_pd(fjz1,tz);
 953
 954             /**************************
 955              * CALCULATE INTERACTIONS *
 956              **************************/
 957
 958             /* COULOMB ELECTROSTATICS */
 959             velec            = _mm256_mul_pd(qq32,rinv32);
 960             felec            = _mm256_mul_pd(velec,rinvsq32);
 961
 962             /* Update potential sum for this i atom from the interaction with this j atom. */
 963             velec            = _mm256_andnot_pd(dummy_mask,velec);
 964             velecsum         = _mm256_add_pd(velecsum,velec);
 965
 966             fscal            = felec;
 967
 968             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 969
 970             /* Calculate temporary vectorial force */
 971             tx               = _mm256_mul_pd(fscal,dx32);
 972             ty               = _mm256_mul_pd(fscal,dy32);
 973             tz               = _mm256_mul_pd(fscal,dz32);
 974
 975             /* Update vectorial force */
 976             fix3             = _mm256_add_pd(fix3,tx);
 977             fiy3             = _mm256_add_pd(fiy3,ty);
 978             fiz3             = _mm256_add_pd(fiz3,tz);
 979
 980             fjx2             = _mm256_add_pd(fjx2,tx);
 981             fjy2             = _mm256_add_pd(fjy2,ty);
 982             fjz2             = _mm256_add_pd(fjz2,tz);
 983
 984             /**************************
 985              * CALCULATE INTERACTIONS *
 986              **************************/
 987
 988             /* COULOMB ELECTROSTATICS */
 989             velec            = _mm256_mul_pd(qq33,rinv33);
 990             felec            = _mm256_mul_pd(velec,rinvsq33);
 991
 992             /* Update potential sum for this i atom from the interaction with this j atom. */
 993             velec            = _mm256_andnot_pd(dummy_mask,velec);
 994             velecsum         = _mm256_add_pd(velecsum,velec);
 995
 996             fscal            = felec;
 997
 998             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 999
1000             /* Calculate temporary vectorial force */
1001             tx               = _mm256_mul_pd(fscal,dx33);
1002             ty               = _mm256_mul_pd(fscal,dy33);
1003             tz               = _mm256_mul_pd(fscal,dz33);
1004
1005             /* Update vectorial force */
1006             fix3             = _mm256_add_pd(fix3,tx);
1007             fiy3             = _mm256_add_pd(fiy3,ty);
1008             fiz3             = _mm256_add_pd(fiz3,tz);
1009
1010             fjx3             = _mm256_add_pd(fjx3,tx);
1011             fjy3             = _mm256_add_pd(fjy3,ty);
1012             fjz3             = _mm256_add_pd(fjz3,tz);
1013
1014             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1015             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1016             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1017             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1018
1019             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1020                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1021                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1022
1023             /* Inner loop uses 278 flops */
1024         }
1025
1026         /* End of innermost loop */
1027
1028         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1029                                                  f+i_coord_offset,fshift+i_shift_offset);
1030
1031         ggid                        = gid[iidx];
1032         /* Update potential energies */
1033         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1034         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1035
1036         /* Increment number of inner iterations */
1037         inneriter                  += j_index_end - j_index_start;
1038
1039         /* Outer loop uses 26 flops */
1040     }
1041
1042     /* Increment number of outer iterations */
1043     outeriter        += nri;
1044
1045     /* Update outer/inner flops */
1046
1047     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1048 }
1049 /*
1050  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1051  * Electrostatics interaction: Coulomb
1052  * VdW interaction:            LennardJones
1053  * Geometry:                   Water4-Water4
1054  * Calculate force/pot:        Force
1055  */
1056 void
1057 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1058                     (t_nblist                    * gmx_restrict       nlist,
1059                      rvec                        * gmx_restrict          xx,
1060                      rvec                        * gmx_restrict          ff,
1061                      t_forcerec                  * gmx_restrict          fr,
1062                      t_mdatoms                   * gmx_restrict     mdatoms,
1063                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1064                      t_nrnb                      * gmx_restrict        nrnb)
1065 {
1066     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1067      * just 0 for non-waters.
1068      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1069      * jnr indices corresponding to data put in the four positions in the SIMD register.
1070      */
1071     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1072     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1073     int              jnrA,jnrB,jnrC,jnrD;
1074     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1075     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1076     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1077     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1078     real             rcutoff_scalar;
1079     real             *shiftvec,*fshift,*x,*f;
1080     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1081     real             scratch[4*DIM];
1082     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1083     real *           vdwioffsetptr0;
1084     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1085     real *           vdwioffsetptr1;
1086     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1087     real *           vdwioffsetptr2;
1088     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1089     real *           vdwioffsetptr3;
1090     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1091     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1092     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1093     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1094     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1095     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1096     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1097     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1098     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1099     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1100     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1101     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1102     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1103     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1104     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1105     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1106     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1107     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1108     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1109     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1110     real             *charge;
1111     int              nvdwtype;
1112     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1113     int              *vdwtype;
1114     real             *vdwparam;
1115     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1116     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1117     __m256d          dummy_mask,cutoff_mask;
1118     __m128           tmpmask0,tmpmask1;
1119     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1120     __m256d          one     = _mm256_set1_pd(1.0);
1121     __m256d          two     = _mm256_set1_pd(2.0);
1122     x                = xx[0];
1123     f                = ff[0];
1124
1125     nri              = nlist->nri;
1126     iinr             = nlist->iinr;
1127     jindex           = nlist->jindex;
1128     jjnr             = nlist->jjnr;
1129     shiftidx         = nlist->shift;
1130     gid              = nlist->gid;
1131     shiftvec         = fr->shift_vec[0];
1132     fshift           = fr->fshift[0];
1133     facel            = _mm256_set1_pd(fr->epsfac);
1134     charge           = mdatoms->chargeA;
1135     nvdwtype         = fr->ntype;
1136     vdwparam         = fr->nbfp;
1137     vdwtype          = mdatoms->typeA;
1138
1139     /* Setup water-specific parameters */
1140     inr              = nlist->iinr[0];
1141     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1142     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1143     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1144     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1145
1146     jq1              = _mm256_set1_pd(charge[inr+1]);
1147     jq2              = _mm256_set1_pd(charge[inr+2]);
1148     jq3              = _mm256_set1_pd(charge[inr+3]);
1149     vdwjidx0A        = 2*vdwtype[inr+0];
1150     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1151     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1152     qq11             = _mm256_mul_pd(iq1,jq1);
1153     qq12             = _mm256_mul_pd(iq1,jq2);
1154     qq13             = _mm256_mul_pd(iq1,jq3);
1155     qq21             = _mm256_mul_pd(iq2,jq1);
1156     qq22             = _mm256_mul_pd(iq2,jq2);
1157     qq23             = _mm256_mul_pd(iq2,jq3);
1158     qq31             = _mm256_mul_pd(iq3,jq1);
1159     qq32             = _mm256_mul_pd(iq3,jq2);
1160     qq33             = _mm256_mul_pd(iq3,jq3);
1161
1162     /* Avoid stupid compiler warnings */
1163     jnrA = jnrB = jnrC = jnrD = 0;
1164     j_coord_offsetA = 0;
1165     j_coord_offsetB = 0;
1166     j_coord_offsetC = 0;
1167     j_coord_offsetD = 0;
1168
1169     outeriter        = 0;
1170     inneriter        = 0;
1171
1172     for(iidx=0;iidx<4*DIM;iidx++)
1173     {
1174         scratch[iidx] = 0.0;
1175     }
1176
1177     /* Start outer loop over neighborlists */
1178     for(iidx=0; iidx<nri; iidx++)
1179     {
1180         /* Load shift vector for this list */
1181         i_shift_offset   = DIM*shiftidx[iidx];
1182
1183         /* Load limits for loop over neighbors */
1184         j_index_start    = jindex[iidx];
1185         j_index_end      = jindex[iidx+1];
1186
1187         /* Get outer coordinate index */
1188         inr              = iinr[iidx];
1189         i_coord_offset   = DIM*inr;
1190
1191         /* Load i particle coords and add shift vector */
1192         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1193                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1194
1195         fix0             = _mm256_setzero_pd();
1196         fiy0             = _mm256_setzero_pd();
1197         fiz0             = _mm256_setzero_pd();
1198         fix1             = _mm256_setzero_pd();
1199         fiy1             = _mm256_setzero_pd();
1200         fiz1             = _mm256_setzero_pd();
1201         fix2             = _mm256_setzero_pd();
1202         fiy2             = _mm256_setzero_pd();
1203         fiz2             = _mm256_setzero_pd();
1204         fix3             = _mm256_setzero_pd();
1205         fiy3             = _mm256_setzero_pd();
1206         fiz3             = _mm256_setzero_pd();
1207
1208         /* Start inner kernel loop */
1209         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1210         {
1211
1212             /* Get j neighbor index, and coordinate index */
1213             jnrA             = jjnr[jidx];
1214             jnrB             = jjnr[jidx+1];
1215             jnrC             = jjnr[jidx+2];
1216             jnrD             = jjnr[jidx+3];
1217             j_coord_offsetA  = DIM*jnrA;
1218             j_coord_offsetB  = DIM*jnrB;
1219             j_coord_offsetC  = DIM*jnrC;
1220             j_coord_offsetD  = DIM*jnrD;
1221
1222             /* load j atom coordinates */
1223             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1224                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1225                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1226                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1227
1228             /* Calculate displacement vector */
1229             dx00             = _mm256_sub_pd(ix0,jx0);
1230             dy00             = _mm256_sub_pd(iy0,jy0);
1231             dz00             = _mm256_sub_pd(iz0,jz0);
1232             dx11             = _mm256_sub_pd(ix1,jx1);
1233             dy11             = _mm256_sub_pd(iy1,jy1);
1234             dz11             = _mm256_sub_pd(iz1,jz1);
1235             dx12             = _mm256_sub_pd(ix1,jx2);
1236             dy12             = _mm256_sub_pd(iy1,jy2);
1237             dz12             = _mm256_sub_pd(iz1,jz2);
1238             dx13             = _mm256_sub_pd(ix1,jx3);
1239             dy13             = _mm256_sub_pd(iy1,jy3);
1240             dz13             = _mm256_sub_pd(iz1,jz3);
1241             dx21             = _mm256_sub_pd(ix2,jx1);
1242             dy21             = _mm256_sub_pd(iy2,jy1);
1243             dz21             = _mm256_sub_pd(iz2,jz1);
1244             dx22             = _mm256_sub_pd(ix2,jx2);
1245             dy22             = _mm256_sub_pd(iy2,jy2);
1246             dz22             = _mm256_sub_pd(iz2,jz2);
1247             dx23             = _mm256_sub_pd(ix2,jx3);
1248             dy23             = _mm256_sub_pd(iy2,jy3);
1249             dz23             = _mm256_sub_pd(iz2,jz3);
1250             dx31             = _mm256_sub_pd(ix3,jx1);
1251             dy31             = _mm256_sub_pd(iy3,jy1);
1252             dz31             = _mm256_sub_pd(iz3,jz1);
1253             dx32             = _mm256_sub_pd(ix3,jx2);
1254             dy32             = _mm256_sub_pd(iy3,jy2);
1255             dz32             = _mm256_sub_pd(iz3,jz2);
1256             dx33             = _mm256_sub_pd(ix3,jx3);
1257             dy33             = _mm256_sub_pd(iy3,jy3);
1258             dz33             = _mm256_sub_pd(iz3,jz3);
1259
1260             /* Calculate squared distance and things based on it */
1261             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1262             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1263             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1264             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1265             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1266             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1267             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1268             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1269             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1270             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1271
1272             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1273             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1274             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1275             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1276             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1277             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1278             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1279             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1280             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1281
1282             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1283             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1284             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1285             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1286             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1287             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1288             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1289             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1290             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1291             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1292
1293             fjx0             = _mm256_setzero_pd();
1294             fjy0             = _mm256_setzero_pd();
1295             fjz0             = _mm256_setzero_pd();
1296             fjx1             = _mm256_setzero_pd();
1297             fjy1             = _mm256_setzero_pd();
1298             fjz1             = _mm256_setzero_pd();
1299             fjx2             = _mm256_setzero_pd();
1300             fjy2             = _mm256_setzero_pd();
1301             fjz2             = _mm256_setzero_pd();
1302             fjx3             = _mm256_setzero_pd();
1303             fjy3             = _mm256_setzero_pd();
1304             fjz3             = _mm256_setzero_pd();
1305
1306             /**************************
1307              * CALCULATE INTERACTIONS *
1308              **************************/
1309
1310             /* LENNARD-JONES DISPERSION/REPULSION */
1311
1312             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1313             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1314
1315             fscal            = fvdw;
1316
1317             /* Calculate temporary vectorial force */
1318             tx               = _mm256_mul_pd(fscal,dx00);
1319             ty               = _mm256_mul_pd(fscal,dy00);
1320             tz               = _mm256_mul_pd(fscal,dz00);
1321
1322             /* Update vectorial force */
1323             fix0             = _mm256_add_pd(fix0,tx);
1324             fiy0             = _mm256_add_pd(fiy0,ty);
1325             fiz0             = _mm256_add_pd(fiz0,tz);
1326
1327             fjx0             = _mm256_add_pd(fjx0,tx);
1328             fjy0             = _mm256_add_pd(fjy0,ty);
1329             fjz0             = _mm256_add_pd(fjz0,tz);
1330
1331             /**************************
1332              * CALCULATE INTERACTIONS *
1333              **************************/
1334
1335             /* COULOMB ELECTROSTATICS */
1336             velec            = _mm256_mul_pd(qq11,rinv11);
1337             felec            = _mm256_mul_pd(velec,rinvsq11);
1338
1339             fscal            = felec;
1340
1341             /* Calculate temporary vectorial force */
1342             tx               = _mm256_mul_pd(fscal,dx11);
1343             ty               = _mm256_mul_pd(fscal,dy11);
1344             tz               = _mm256_mul_pd(fscal,dz11);
1345
1346             /* Update vectorial force */
1347             fix1             = _mm256_add_pd(fix1,tx);
1348             fiy1             = _mm256_add_pd(fiy1,ty);
1349             fiz1             = _mm256_add_pd(fiz1,tz);
1350
1351             fjx1             = _mm256_add_pd(fjx1,tx);
1352             fjy1             = _mm256_add_pd(fjy1,ty);
1353             fjz1             = _mm256_add_pd(fjz1,tz);
1354
1355             /**************************
1356              * CALCULATE INTERACTIONS *
1357              **************************/
1358
1359             /* COULOMB ELECTROSTATICS */
1360             velec            = _mm256_mul_pd(qq12,rinv12);
1361             felec            = _mm256_mul_pd(velec,rinvsq12);
1362
1363             fscal            = felec;
1364
1365             /* Calculate temporary vectorial force */
1366             tx               = _mm256_mul_pd(fscal,dx12);
1367             ty               = _mm256_mul_pd(fscal,dy12);
1368             tz               = _mm256_mul_pd(fscal,dz12);
1369
1370             /* Update vectorial force */
1371             fix1             = _mm256_add_pd(fix1,tx);
1372             fiy1             = _mm256_add_pd(fiy1,ty);
1373             fiz1             = _mm256_add_pd(fiz1,tz);
1374
1375             fjx2             = _mm256_add_pd(fjx2,tx);
1376             fjy2             = _mm256_add_pd(fjy2,ty);
1377             fjz2             = _mm256_add_pd(fjz2,tz);
1378
1379             /**************************
1380              * CALCULATE INTERACTIONS *
1381              **************************/
1382
1383             /* COULOMB ELECTROSTATICS */
1384             velec            = _mm256_mul_pd(qq13,rinv13);
1385             felec            = _mm256_mul_pd(velec,rinvsq13);
1386
1387             fscal            = felec;
1388
1389             /* Calculate temporary vectorial force */
1390             tx               = _mm256_mul_pd(fscal,dx13);
1391             ty               = _mm256_mul_pd(fscal,dy13);
1392             tz               = _mm256_mul_pd(fscal,dz13);
1393
1394             /* Update vectorial force */
1395             fix1             = _mm256_add_pd(fix1,tx);
1396             fiy1             = _mm256_add_pd(fiy1,ty);
1397             fiz1             = _mm256_add_pd(fiz1,tz);
1398
1399             fjx3             = _mm256_add_pd(fjx3,tx);
1400             fjy3             = _mm256_add_pd(fjy3,ty);
1401             fjz3             = _mm256_add_pd(fjz3,tz);
1402
1403             /**************************
1404              * CALCULATE INTERACTIONS *
1405              **************************/
1406
1407             /* COULOMB ELECTROSTATICS */
1408             velec            = _mm256_mul_pd(qq21,rinv21);
1409             felec            = _mm256_mul_pd(velec,rinvsq21);
1410
1411             fscal            = felec;
1412
1413             /* Calculate temporary vectorial force */
1414             tx               = _mm256_mul_pd(fscal,dx21);
1415             ty               = _mm256_mul_pd(fscal,dy21);
1416             tz               = _mm256_mul_pd(fscal,dz21);
1417
1418             /* Update vectorial force */
1419             fix2             = _mm256_add_pd(fix2,tx);
1420             fiy2             = _mm256_add_pd(fiy2,ty);
1421             fiz2             = _mm256_add_pd(fiz2,tz);
1422
1423             fjx1             = _mm256_add_pd(fjx1,tx);
1424             fjy1             = _mm256_add_pd(fjy1,ty);
1425             fjz1             = _mm256_add_pd(fjz1,tz);
1426
1427             /**************************
1428              * CALCULATE INTERACTIONS *
1429              **************************/
1430
1431             /* COULOMB ELECTROSTATICS */
1432             velec            = _mm256_mul_pd(qq22,rinv22);
1433             felec            = _mm256_mul_pd(velec,rinvsq22);
1434
1435             fscal            = felec;
1436
1437             /* Calculate temporary vectorial force */
1438             tx               = _mm256_mul_pd(fscal,dx22);
1439             ty               = _mm256_mul_pd(fscal,dy22);
1440             tz               = _mm256_mul_pd(fscal,dz22);
1441
1442             /* Update vectorial force */
1443             fix2             = _mm256_add_pd(fix2,tx);
1444             fiy2             = _mm256_add_pd(fiy2,ty);
1445             fiz2             = _mm256_add_pd(fiz2,tz);
1446
1447             fjx2             = _mm256_add_pd(fjx2,tx);
1448             fjy2             = _mm256_add_pd(fjy2,ty);
1449             fjz2             = _mm256_add_pd(fjz2,tz);
1450
1451             /**************************
1452              * CALCULATE INTERACTIONS *
1453              **************************/
1454
1455             /* COULOMB ELECTROSTATICS */
1456             velec            = _mm256_mul_pd(qq23,rinv23);
1457             felec            = _mm256_mul_pd(velec,rinvsq23);
1458
1459             fscal            = felec;
1460
1461             /* Calculate temporary vectorial force */
1462             tx               = _mm256_mul_pd(fscal,dx23);
1463             ty               = _mm256_mul_pd(fscal,dy23);
1464             tz               = _mm256_mul_pd(fscal,dz23);
1465
1466             /* Update vectorial force */
1467             fix2             = _mm256_add_pd(fix2,tx);
1468             fiy2             = _mm256_add_pd(fiy2,ty);
1469             fiz2             = _mm256_add_pd(fiz2,tz);
1470
1471             fjx3             = _mm256_add_pd(fjx3,tx);
1472             fjy3             = _mm256_add_pd(fjy3,ty);
1473             fjz3             = _mm256_add_pd(fjz3,tz);
1474
1475             /**************************
1476              * CALCULATE INTERACTIONS *
1477              **************************/
1478
1479             /* COULOMB ELECTROSTATICS */
1480             velec            = _mm256_mul_pd(qq31,rinv31);
1481             felec            = _mm256_mul_pd(velec,rinvsq31);
1482
1483             fscal            = felec;
1484
1485             /* Calculate temporary vectorial force */
1486             tx               = _mm256_mul_pd(fscal,dx31);
1487             ty               = _mm256_mul_pd(fscal,dy31);
1488             tz               = _mm256_mul_pd(fscal,dz31);
1489
1490             /* Update vectorial force */
1491             fix3             = _mm256_add_pd(fix3,tx);
1492             fiy3             = _mm256_add_pd(fiy3,ty);
1493             fiz3             = _mm256_add_pd(fiz3,tz);
1494
1495             fjx1             = _mm256_add_pd(fjx1,tx);
1496             fjy1             = _mm256_add_pd(fjy1,ty);
1497             fjz1             = _mm256_add_pd(fjz1,tz);
1498
1499             /**************************
1500              * CALCULATE INTERACTIONS *
1501              **************************/
1502
1503             /* COULOMB ELECTROSTATICS */
1504             velec            = _mm256_mul_pd(qq32,rinv32);
1505             felec            = _mm256_mul_pd(velec,rinvsq32);
1506
1507             fscal            = felec;
1508
1509             /* Calculate temporary vectorial force */
1510             tx               = _mm256_mul_pd(fscal,dx32);
1511             ty               = _mm256_mul_pd(fscal,dy32);
1512             tz               = _mm256_mul_pd(fscal,dz32);
1513
1514             /* Update vectorial force */
1515             fix3             = _mm256_add_pd(fix3,tx);
1516             fiy3             = _mm256_add_pd(fiy3,ty);
1517             fiz3             = _mm256_add_pd(fiz3,tz);
1518
1519             fjx2             = _mm256_add_pd(fjx2,tx);
1520             fjy2             = _mm256_add_pd(fjy2,ty);
1521             fjz2             = _mm256_add_pd(fjz2,tz);
1522
1523             /**************************
1524              * CALCULATE INTERACTIONS *
1525              **************************/
1526
1527             /* COULOMB ELECTROSTATICS */
1528             velec            = _mm256_mul_pd(qq33,rinv33);
1529             felec            = _mm256_mul_pd(velec,rinvsq33);
1530
1531             fscal            = felec;
1532
1533             /* Calculate temporary vectorial force */
1534             tx               = _mm256_mul_pd(fscal,dx33);
1535             ty               = _mm256_mul_pd(fscal,dy33);
1536             tz               = _mm256_mul_pd(fscal,dz33);
1537
1538             /* Update vectorial force */
1539             fix3             = _mm256_add_pd(fix3,tx);
1540             fiy3             = _mm256_add_pd(fiy3,ty);
1541             fiz3             = _mm256_add_pd(fiz3,tz);
1542
1543             fjx3             = _mm256_add_pd(fjx3,tx);
1544             fjy3             = _mm256_add_pd(fjy3,ty);
1545             fjz3             = _mm256_add_pd(fjz3,tz);
1546
1547             fjptrA             = f+j_coord_offsetA;
1548             fjptrB             = f+j_coord_offsetB;
1549             fjptrC             = f+j_coord_offsetC;
1550             fjptrD             = f+j_coord_offsetD;
1551
1552             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1553                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1554                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1555
1556             /* Inner loop uses 264 flops */
1557         }
1558
1559         if(jidx<j_index_end)
1560         {
1561
1562             /* Get j neighbor index, and coordinate index */
1563             jnrlistA         = jjnr[jidx];
1564             jnrlistB         = jjnr[jidx+1];
1565             jnrlistC         = jjnr[jidx+2];
1566             jnrlistD         = jjnr[jidx+3];
1567             /* Sign of each element will be negative for non-real atoms.
1568              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1569              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1570              */
1571             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1572
1573             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1574             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1575             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1576
1577             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1578             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1579             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1580             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1581             j_coord_offsetA  = DIM*jnrA;
1582             j_coord_offsetB  = DIM*jnrB;
1583             j_coord_offsetC  = DIM*jnrC;
1584             j_coord_offsetD  = DIM*jnrD;
1585
1586             /* load j atom coordinates */
1587             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1588                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1589                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1590                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1591
1592             /* Calculate displacement vector */
1593             dx00             = _mm256_sub_pd(ix0,jx0);
1594             dy00             = _mm256_sub_pd(iy0,jy0);
1595             dz00             = _mm256_sub_pd(iz0,jz0);
1596             dx11             = _mm256_sub_pd(ix1,jx1);
1597             dy11             = _mm256_sub_pd(iy1,jy1);
1598             dz11             = _mm256_sub_pd(iz1,jz1);
1599             dx12             = _mm256_sub_pd(ix1,jx2);
1600             dy12             = _mm256_sub_pd(iy1,jy2);
1601             dz12             = _mm256_sub_pd(iz1,jz2);
1602             dx13             = _mm256_sub_pd(ix1,jx3);
1603             dy13             = _mm256_sub_pd(iy1,jy3);
1604             dz13             = _mm256_sub_pd(iz1,jz3);
1605             dx21             = _mm256_sub_pd(ix2,jx1);
1606             dy21             = _mm256_sub_pd(iy2,jy1);
1607             dz21             = _mm256_sub_pd(iz2,jz1);
1608             dx22             = _mm256_sub_pd(ix2,jx2);
1609             dy22             = _mm256_sub_pd(iy2,jy2);
1610             dz22             = _mm256_sub_pd(iz2,jz2);
1611             dx23             = _mm256_sub_pd(ix2,jx3);
1612             dy23             = _mm256_sub_pd(iy2,jy3);
1613             dz23             = _mm256_sub_pd(iz2,jz3);
1614             dx31             = _mm256_sub_pd(ix3,jx1);
1615             dy31             = _mm256_sub_pd(iy3,jy1);
1616             dz31             = _mm256_sub_pd(iz3,jz1);
1617             dx32             = _mm256_sub_pd(ix3,jx2);
1618             dy32             = _mm256_sub_pd(iy3,jy2);
1619             dz32             = _mm256_sub_pd(iz3,jz2);
1620             dx33             = _mm256_sub_pd(ix3,jx3);
1621             dy33             = _mm256_sub_pd(iy3,jy3);
1622             dz33             = _mm256_sub_pd(iz3,jz3);
1623
1624             /* Calculate squared distance and things based on it */
1625             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1626             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1627             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1628             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1629             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1630             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1631             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1632             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1633             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1634             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1635
1636             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1637             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1638             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1639             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1640             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1641             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1642             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1643             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1644             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1645
1646             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1647             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1648             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1649             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1650             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1651             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1652             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1653             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1654             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1655             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1656
1657             fjx0             = _mm256_setzero_pd();
1658             fjy0             = _mm256_setzero_pd();
1659             fjz0             = _mm256_setzero_pd();
1660             fjx1             = _mm256_setzero_pd();
1661             fjy1             = _mm256_setzero_pd();
1662             fjz1             = _mm256_setzero_pd();
1663             fjx2             = _mm256_setzero_pd();
1664             fjy2             = _mm256_setzero_pd();
1665             fjz2             = _mm256_setzero_pd();
1666             fjx3             = _mm256_setzero_pd();
1667             fjy3             = _mm256_setzero_pd();
1668             fjz3             = _mm256_setzero_pd();
1669
1670             /**************************
1671              * CALCULATE INTERACTIONS *
1672              **************************/
1673
1674             /* LENNARD-JONES DISPERSION/REPULSION */
1675
1676             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1677             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1678
1679             fscal            = fvdw;
1680
1681             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1682
1683             /* Calculate temporary vectorial force */
1684             tx               = _mm256_mul_pd(fscal,dx00);
1685             ty               = _mm256_mul_pd(fscal,dy00);
1686             tz               = _mm256_mul_pd(fscal,dz00);
1687
1688             /* Update vectorial force */
1689             fix0             = _mm256_add_pd(fix0,tx);
1690             fiy0             = _mm256_add_pd(fiy0,ty);
1691             fiz0             = _mm256_add_pd(fiz0,tz);
1692
1693             fjx0             = _mm256_add_pd(fjx0,tx);
1694             fjy0             = _mm256_add_pd(fjy0,ty);
1695             fjz0             = _mm256_add_pd(fjz0,tz);
1696
1697             /**************************
1698              * CALCULATE INTERACTIONS *
1699              **************************/
1700
1701             /* COULOMB ELECTROSTATICS */
1702             velec            = _mm256_mul_pd(qq11,rinv11);
1703             felec            = _mm256_mul_pd(velec,rinvsq11);
1704
1705             fscal            = felec;
1706
1707             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1708
1709             /* Calculate temporary vectorial force */
1710             tx               = _mm256_mul_pd(fscal,dx11);
1711             ty               = _mm256_mul_pd(fscal,dy11);
1712             tz               = _mm256_mul_pd(fscal,dz11);
1713
1714             /* Update vectorial force */
1715             fix1             = _mm256_add_pd(fix1,tx);
1716             fiy1             = _mm256_add_pd(fiy1,ty);
1717             fiz1             = _mm256_add_pd(fiz1,tz);
1718
1719             fjx1             = _mm256_add_pd(fjx1,tx);
1720             fjy1             = _mm256_add_pd(fjy1,ty);
1721             fjz1             = _mm256_add_pd(fjz1,tz);
1722
1723             /**************************
1724              * CALCULATE INTERACTIONS *
1725              **************************/
1726
1727             /* COULOMB ELECTROSTATICS */
1728             velec            = _mm256_mul_pd(qq12,rinv12);
1729             felec            = _mm256_mul_pd(velec,rinvsq12);
1730
1731             fscal            = felec;
1732
1733             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1734
1735             /* Calculate temporary vectorial force */
1736             tx               = _mm256_mul_pd(fscal,dx12);
1737             ty               = _mm256_mul_pd(fscal,dy12);
1738             tz               = _mm256_mul_pd(fscal,dz12);
1739
1740             /* Update vectorial force */
1741             fix1             = _mm256_add_pd(fix1,tx);
1742             fiy1             = _mm256_add_pd(fiy1,ty);
1743             fiz1             = _mm256_add_pd(fiz1,tz);
1744
1745             fjx2             = _mm256_add_pd(fjx2,tx);
1746             fjy2             = _mm256_add_pd(fjy2,ty);
1747             fjz2             = _mm256_add_pd(fjz2,tz);
1748
1749             /**************************
1750              * CALCULATE INTERACTIONS *
1751              **************************/
1752
1753             /* COULOMB ELECTROSTATICS */
1754             velec            = _mm256_mul_pd(qq13,rinv13);
1755             felec            = _mm256_mul_pd(velec,rinvsq13);
1756
1757             fscal            = felec;
1758
1759             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1760
1761             /* Calculate temporary vectorial force */
1762             tx               = _mm256_mul_pd(fscal,dx13);
1763             ty               = _mm256_mul_pd(fscal,dy13);
1764             tz               = _mm256_mul_pd(fscal,dz13);
1765
1766             /* Update vectorial force */
1767             fix1             = _mm256_add_pd(fix1,tx);
1768             fiy1             = _mm256_add_pd(fiy1,ty);
1769             fiz1             = _mm256_add_pd(fiz1,tz);
1770
1771             fjx3             = _mm256_add_pd(fjx3,tx);
1772             fjy3             = _mm256_add_pd(fjy3,ty);
1773             fjz3             = _mm256_add_pd(fjz3,tz);
1774
1775             /**************************
1776              * CALCULATE INTERACTIONS *
1777              **************************/
1778
1779             /* COULOMB ELECTROSTATICS */
1780             velec            = _mm256_mul_pd(qq21,rinv21);
1781             felec            = _mm256_mul_pd(velec,rinvsq21);
1782
1783             fscal            = felec;
1784
1785             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1786
1787             /* Calculate temporary vectorial force */
1788             tx               = _mm256_mul_pd(fscal,dx21);
1789             ty               = _mm256_mul_pd(fscal,dy21);
1790             tz               = _mm256_mul_pd(fscal,dz21);
1791
1792             /* Update vectorial force */
1793             fix2             = _mm256_add_pd(fix2,tx);
1794             fiy2             = _mm256_add_pd(fiy2,ty);
1795             fiz2             = _mm256_add_pd(fiz2,tz);
1796
1797             fjx1             = _mm256_add_pd(fjx1,tx);
1798             fjy1             = _mm256_add_pd(fjy1,ty);
1799             fjz1             = _mm256_add_pd(fjz1,tz);
1800
1801             /**************************
1802              * CALCULATE INTERACTIONS *
1803              **************************/
1804
1805             /* COULOMB ELECTROSTATICS */
1806             velec            = _mm256_mul_pd(qq22,rinv22);
1807             felec            = _mm256_mul_pd(velec,rinvsq22);
1808
1809             fscal            = felec;
1810
1811             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1812
1813             /* Calculate temporary vectorial force */
1814             tx               = _mm256_mul_pd(fscal,dx22);
1815             ty               = _mm256_mul_pd(fscal,dy22);
1816             tz               = _mm256_mul_pd(fscal,dz22);
1817
1818             /* Update vectorial force */
1819             fix2             = _mm256_add_pd(fix2,tx);
1820             fiy2             = _mm256_add_pd(fiy2,ty);
1821             fiz2             = _mm256_add_pd(fiz2,tz);
1822
1823             fjx2             = _mm256_add_pd(fjx2,tx);
1824             fjy2             = _mm256_add_pd(fjy2,ty);
1825             fjz2             = _mm256_add_pd(fjz2,tz);
1826
1827             /**************************
1828              * CALCULATE INTERACTIONS *
1829              **************************/
1830
1831             /* COULOMB ELECTROSTATICS */
1832             velec            = _mm256_mul_pd(qq23,rinv23);
1833             felec            = _mm256_mul_pd(velec,rinvsq23);
1834
1835             fscal            = felec;
1836
1837             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1838
1839             /* Calculate temporary vectorial force */
1840             tx               = _mm256_mul_pd(fscal,dx23);
1841             ty               = _mm256_mul_pd(fscal,dy23);
1842             tz               = _mm256_mul_pd(fscal,dz23);
1843
1844             /* Update vectorial force */
1845             fix2             = _mm256_add_pd(fix2,tx);
1846             fiy2             = _mm256_add_pd(fiy2,ty);
1847             fiz2             = _mm256_add_pd(fiz2,tz);
1848
1849             fjx3             = _mm256_add_pd(fjx3,tx);
1850             fjy3             = _mm256_add_pd(fjy3,ty);
1851             fjz3             = _mm256_add_pd(fjz3,tz);
1852
1853             /**************************
1854              * CALCULATE INTERACTIONS *
1855              **************************/
1856
1857             /* COULOMB ELECTROSTATICS */
1858             velec            = _mm256_mul_pd(qq31,rinv31);
1859             felec            = _mm256_mul_pd(velec,rinvsq31);
1860
1861             fscal            = felec;
1862
1863             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1864
1865             /* Calculate temporary vectorial force */
1866             tx               = _mm256_mul_pd(fscal,dx31);
1867             ty               = _mm256_mul_pd(fscal,dy31);
1868             tz               = _mm256_mul_pd(fscal,dz31);
1869
1870             /* Update vectorial force */
1871             fix3             = _mm256_add_pd(fix3,tx);
1872             fiy3             = _mm256_add_pd(fiy3,ty);
1873             fiz3             = _mm256_add_pd(fiz3,tz);
1874
1875             fjx1             = _mm256_add_pd(fjx1,tx);
1876             fjy1             = _mm256_add_pd(fjy1,ty);
1877             fjz1             = _mm256_add_pd(fjz1,tz);
1878
1879             /**************************
1880              * CALCULATE INTERACTIONS *
1881              **************************/
1882
1883             /* COULOMB ELECTROSTATICS */
1884             velec            = _mm256_mul_pd(qq32,rinv32);
1885             felec            = _mm256_mul_pd(velec,rinvsq32);
1886
1887             fscal            = felec;
1888
1889             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1890
1891             /* Calculate temporary vectorial force */
1892             tx               = _mm256_mul_pd(fscal,dx32);
1893             ty               = _mm256_mul_pd(fscal,dy32);
1894             tz               = _mm256_mul_pd(fscal,dz32);
1895
1896             /* Update vectorial force */
1897             fix3             = _mm256_add_pd(fix3,tx);
1898             fiy3             = _mm256_add_pd(fiy3,ty);
1899             fiz3             = _mm256_add_pd(fiz3,tz);
1900
1901             fjx2             = _mm256_add_pd(fjx2,tx);
1902             fjy2             = _mm256_add_pd(fjy2,ty);
1903             fjz2             = _mm256_add_pd(fjz2,tz);
1904
1905             /**************************
1906              * CALCULATE INTERACTIONS *
1907              **************************/
1908
1909             /* COULOMB ELECTROSTATICS */
1910             velec            = _mm256_mul_pd(qq33,rinv33);
1911             felec            = _mm256_mul_pd(velec,rinvsq33);
1912
1913             fscal            = felec;
1914
1915             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1916
1917             /* Calculate temporary vectorial force */
1918             tx               = _mm256_mul_pd(fscal,dx33);
1919             ty               = _mm256_mul_pd(fscal,dy33);
1920             tz               = _mm256_mul_pd(fscal,dz33);
1921
1922             /* Update vectorial force */
1923             fix3             = _mm256_add_pd(fix3,tx);
1924             fiy3             = _mm256_add_pd(fiy3,ty);
1925             fiz3             = _mm256_add_pd(fiz3,tz);
1926
1927             fjx3             = _mm256_add_pd(fjx3,tx);
1928             fjy3             = _mm256_add_pd(fjy3,ty);
1929             fjz3             = _mm256_add_pd(fjz3,tz);
1930
1931             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1932             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1933             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1934             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1935
1936             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1937                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1938                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1939
1940             /* Inner loop uses 264 flops */
1941         }
1942
1943         /* End of innermost loop */
1944
1945         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1946                                                  f+i_coord_offset,fshift+i_shift_offset);
1947
1948         /* Increment number of inner iterations */
1949         inneriter                  += j_index_end - j_index_start;
1950
1951         /* Outer loop uses 24 flops */
1952     }
1953
1954     /* Increment number of outer iterations */
1955     outeriter        += nri;
1956
1957     /* Update outer/inner flops */
1958
1959     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
1960 }