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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_double.h"
  48 #include "kernelutil_x86_avx_256_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  82     real             scratch[4*DIM];
  83     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  84     real *           vdwioffsetptr0;
  85     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  86     real *           vdwioffsetptr1;
  87     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  88     real *           vdwioffsetptr2;
  89     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  90     real *           vdwioffsetptr3;
  91     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  99     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 100     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 104     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 107     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 108     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 109     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 110     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 111     real             *charge;
 112     int              nvdwtype;
 113     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 114     int              *vdwtype;
 115     real             *vdwparam;
 116     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 117     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 118     __m256d          dummy_mask,cutoff_mask;
 119     __m128           tmpmask0,tmpmask1;
 120     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 121     __m256d          one     = _mm256_set1_pd(1.0);
 122     __m256d          two     = _mm256_set1_pd(2.0);
 123     x                = xx[0];
 124     f                = ff[0];
 125
 126     nri              = nlist->nri;
 127     iinr             = nlist->iinr;
 128     jindex           = nlist->jindex;
 129     jjnr             = nlist->jjnr;
 130     shiftidx         = nlist->shift;
 131     gid              = nlist->gid;
 132     shiftvec         = fr->shift_vec[0];
 133     fshift           = fr->fshift[0];
 134     facel            = _mm256_set1_pd(fr->epsfac);
 135     charge           = mdatoms->chargeA;
 136     nvdwtype         = fr->ntype;
 137     vdwparam         = fr->nbfp;
 138     vdwtype          = mdatoms->typeA;
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 143     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 144     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 145     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 146
 147     jq1              = _mm256_set1_pd(charge[inr+1]);
 148     jq2              = _mm256_set1_pd(charge[inr+2]);
 149     jq3              = _mm256_set1_pd(charge[inr+3]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 152     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 153     qq11             = _mm256_mul_pd(iq1,jq1);
 154     qq12             = _mm256_mul_pd(iq1,jq2);
 155     qq13             = _mm256_mul_pd(iq1,jq3);
 156     qq21             = _mm256_mul_pd(iq2,jq1);
 157     qq22             = _mm256_mul_pd(iq2,jq2);
 158     qq23             = _mm256_mul_pd(iq2,jq3);
 159     qq31             = _mm256_mul_pd(iq3,jq1);
 160     qq32             = _mm256_mul_pd(iq3,jq2);
 161     qq33             = _mm256_mul_pd(iq3,jq3);
 162
 163     /* Avoid stupid compiler warnings */
 164     jnrA = jnrB = jnrC = jnrD = 0;
 165     j_coord_offsetA = 0;
 166     j_coord_offsetB = 0;
 167     j_coord_offsetC = 0;
 168     j_coord_offsetD = 0;
 169
 170     outeriter        = 0;
 171     inneriter        = 0;
 172
 173     for(iidx=0;iidx<4*DIM;iidx++)
 174     {
 175         scratch[iidx] = 0.0;
 176     }
 177
 178     /* Start outer loop over neighborlists */
 179     for(iidx=0; iidx<nri; iidx++)
 180     {
 181         /* Load shift vector for this list */
 182         i_shift_offset   = DIM*shiftidx[iidx];
 183
 184         /* Load limits for loop over neighbors */
 185         j_index_start    = jindex[iidx];
 186         j_index_end      = jindex[iidx+1];
 187
 188         /* Get outer coordinate index */
 189         inr              = iinr[iidx];
 190         i_coord_offset   = DIM*inr;
 191
 192         /* Load i particle coords and add shift vector */
 193         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 194                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 195
 196         fix0             = _mm256_setzero_pd();
 197         fiy0             = _mm256_setzero_pd();
 198         fiz0             = _mm256_setzero_pd();
 199         fix1             = _mm256_setzero_pd();
 200         fiy1             = _mm256_setzero_pd();
 201         fiz1             = _mm256_setzero_pd();
 202         fix2             = _mm256_setzero_pd();
 203         fiy2             = _mm256_setzero_pd();
 204         fiz2             = _mm256_setzero_pd();
 205         fix3             = _mm256_setzero_pd();
 206         fiy3             = _mm256_setzero_pd();
 207         fiz3             = _mm256_setzero_pd();
 208
 209         /* Reset potential sums */
 210         velecsum         = _mm256_setzero_pd();
 211         vvdwsum          = _mm256_setzero_pd();
 212
 213         /* Start inner kernel loop */
 214         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 215         {
 216
 217             /* Get j neighbor index, and coordinate index */
 218             jnrA             = jjnr[jidx];
 219             jnrB             = jjnr[jidx+1];
 220             jnrC             = jjnr[jidx+2];
 221             jnrD             = jjnr[jidx+3];
 222             j_coord_offsetA  = DIM*jnrA;
 223             j_coord_offsetB  = DIM*jnrB;
 224             j_coord_offsetC  = DIM*jnrC;
 225             j_coord_offsetD  = DIM*jnrD;
 226
 227             /* load j atom coordinates */
 228             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 229                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 230                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 231                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 232
 233             /* Calculate displacement vector */
 234             dx00             = _mm256_sub_pd(ix0,jx0);
 235             dy00             = _mm256_sub_pd(iy0,jy0);
 236             dz00             = _mm256_sub_pd(iz0,jz0);
 237             dx11             = _mm256_sub_pd(ix1,jx1);
 238             dy11             = _mm256_sub_pd(iy1,jy1);
 239             dz11             = _mm256_sub_pd(iz1,jz1);
 240             dx12             = _mm256_sub_pd(ix1,jx2);
 241             dy12             = _mm256_sub_pd(iy1,jy2);
 242             dz12             = _mm256_sub_pd(iz1,jz2);
 243             dx13             = _mm256_sub_pd(ix1,jx3);
 244             dy13             = _mm256_sub_pd(iy1,jy3);
 245             dz13             = _mm256_sub_pd(iz1,jz3);
 246             dx21             = _mm256_sub_pd(ix2,jx1);
 247             dy21             = _mm256_sub_pd(iy2,jy1);
 248             dz21             = _mm256_sub_pd(iz2,jz1);
 249             dx22             = _mm256_sub_pd(ix2,jx2);
 250             dy22             = _mm256_sub_pd(iy2,jy2);
 251             dz22             = _mm256_sub_pd(iz2,jz2);
 252             dx23             = _mm256_sub_pd(ix2,jx3);
 253             dy23             = _mm256_sub_pd(iy2,jy3);
 254             dz23             = _mm256_sub_pd(iz2,jz3);
 255             dx31             = _mm256_sub_pd(ix3,jx1);
 256             dy31             = _mm256_sub_pd(iy3,jy1);
 257             dz31             = _mm256_sub_pd(iz3,jz1);
 258             dx32             = _mm256_sub_pd(ix3,jx2);
 259             dy32             = _mm256_sub_pd(iy3,jy2);
 260             dz32             = _mm256_sub_pd(iz3,jz2);
 261             dx33             = _mm256_sub_pd(ix3,jx3);
 262             dy33             = _mm256_sub_pd(iy3,jy3);
 263             dz33             = _mm256_sub_pd(iz3,jz3);
 264
 265             /* Calculate squared distance and things based on it */
 266             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 267             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 268             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 269             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 270             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 271             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 272             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 273             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 274             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 275             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 276
 277             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 278             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 279             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 280             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 281             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 282             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 283             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 284             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 285             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 286
 287             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 288             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 289             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 290             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 291             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 292             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 293             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 294             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 295             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 296             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 297
 298             fjx0             = _mm256_setzero_pd();
 299             fjy0             = _mm256_setzero_pd();
 300             fjz0             = _mm256_setzero_pd();
 301             fjx1             = _mm256_setzero_pd();
 302             fjy1             = _mm256_setzero_pd();
 303             fjz1             = _mm256_setzero_pd();
 304             fjx2             = _mm256_setzero_pd();
 305             fjy2             = _mm256_setzero_pd();
 306             fjz2             = _mm256_setzero_pd();
 307             fjx3             = _mm256_setzero_pd();
 308             fjy3             = _mm256_setzero_pd();
 309             fjz3             = _mm256_setzero_pd();
 310
 311             /**************************
 312              * CALCULATE INTERACTIONS *
 313              **************************/
 314
 315             /* LENNARD-JONES DISPERSION/REPULSION */
 316
 317             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 318             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 319             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 320             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 321             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 322
 323             /* Update potential sum for this i atom from the interaction with this j atom. */
 324             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 325
 326             fscal            = fvdw;
 327
 328             /* Calculate temporary vectorial force */
 329             tx               = _mm256_mul_pd(fscal,dx00);
 330             ty               = _mm256_mul_pd(fscal,dy00);
 331             tz               = _mm256_mul_pd(fscal,dz00);
 332
 333             /* Update vectorial force */
 334             fix0             = _mm256_add_pd(fix0,tx);
 335             fiy0             = _mm256_add_pd(fiy0,ty);
 336             fiz0             = _mm256_add_pd(fiz0,tz);
 337
 338             fjx0             = _mm256_add_pd(fjx0,tx);
 339             fjy0             = _mm256_add_pd(fjy0,ty);
 340             fjz0             = _mm256_add_pd(fjz0,tz);
 341
 342             /**************************
 343              * CALCULATE INTERACTIONS *
 344              **************************/
 345
 346             /* COULOMB ELECTROSTATICS */
 347             velec            = _mm256_mul_pd(qq11,rinv11);
 348             felec            = _mm256_mul_pd(velec,rinvsq11);
 349
 350             /* Update potential sum for this i atom from the interaction with this j atom. */
 351             velecsum         = _mm256_add_pd(velecsum,velec);
 352
 353             fscal            = felec;
 354
 355             /* Calculate temporary vectorial force */
 356             tx               = _mm256_mul_pd(fscal,dx11);
 357             ty               = _mm256_mul_pd(fscal,dy11);
 358             tz               = _mm256_mul_pd(fscal,dz11);
 359
 360             /* Update vectorial force */
 361             fix1             = _mm256_add_pd(fix1,tx);
 362             fiy1             = _mm256_add_pd(fiy1,ty);
 363             fiz1             = _mm256_add_pd(fiz1,tz);
 364
 365             fjx1             = _mm256_add_pd(fjx1,tx);
 366             fjy1             = _mm256_add_pd(fjy1,ty);
 367             fjz1             = _mm256_add_pd(fjz1,tz);
 368
 369             /**************************
 370              * CALCULATE INTERACTIONS *
 371              **************************/
 372
 373             /* COULOMB ELECTROSTATICS */
 374             velec            = _mm256_mul_pd(qq12,rinv12);
 375             felec            = _mm256_mul_pd(velec,rinvsq12);
 376
 377             /* Update potential sum for this i atom from the interaction with this j atom. */
 378             velecsum         = _mm256_add_pd(velecsum,velec);
 379
 380             fscal            = felec;
 381
 382             /* Calculate temporary vectorial force */
 383             tx               = _mm256_mul_pd(fscal,dx12);
 384             ty               = _mm256_mul_pd(fscal,dy12);
 385             tz               = _mm256_mul_pd(fscal,dz12);
 386
 387             /* Update vectorial force */
 388             fix1             = _mm256_add_pd(fix1,tx);
 389             fiy1             = _mm256_add_pd(fiy1,ty);
 390             fiz1             = _mm256_add_pd(fiz1,tz);
 391
 392             fjx2             = _mm256_add_pd(fjx2,tx);
 393             fjy2             = _mm256_add_pd(fjy2,ty);
 394             fjz2             = _mm256_add_pd(fjz2,tz);
 395
 396             /**************************
 397              * CALCULATE INTERACTIONS *
 398              **************************/
 399
 400             /* COULOMB ELECTROSTATICS */
 401             velec            = _mm256_mul_pd(qq13,rinv13);
 402             felec            = _mm256_mul_pd(velec,rinvsq13);
 403
 404             /* Update potential sum for this i atom from the interaction with this j atom. */
 405             velecsum         = _mm256_add_pd(velecsum,velec);
 406
 407             fscal            = felec;
 408
 409             /* Calculate temporary vectorial force */
 410             tx               = _mm256_mul_pd(fscal,dx13);
 411             ty               = _mm256_mul_pd(fscal,dy13);
 412             tz               = _mm256_mul_pd(fscal,dz13);
 413
 414             /* Update vectorial force */
 415             fix1             = _mm256_add_pd(fix1,tx);
 416             fiy1             = _mm256_add_pd(fiy1,ty);
 417             fiz1             = _mm256_add_pd(fiz1,tz);
 418
 419             fjx3             = _mm256_add_pd(fjx3,tx);
 420             fjy3             = _mm256_add_pd(fjy3,ty);
 421             fjz3             = _mm256_add_pd(fjz3,tz);
 422
 423             /**************************
 424              * CALCULATE INTERACTIONS *
 425              **************************/
 426
 427             /* COULOMB ELECTROSTATICS */
 428             velec            = _mm256_mul_pd(qq21,rinv21);
 429             felec            = _mm256_mul_pd(velec,rinvsq21);
 430
 431             /* Update potential sum for this i atom from the interaction with this j atom. */
 432             velecsum         = _mm256_add_pd(velecsum,velec);
 433
 434             fscal            = felec;
 435
 436             /* Calculate temporary vectorial force */
 437             tx               = _mm256_mul_pd(fscal,dx21);
 438             ty               = _mm256_mul_pd(fscal,dy21);
 439             tz               = _mm256_mul_pd(fscal,dz21);
 440
 441             /* Update vectorial force */
 442             fix2             = _mm256_add_pd(fix2,tx);
 443             fiy2             = _mm256_add_pd(fiy2,ty);
 444             fiz2             = _mm256_add_pd(fiz2,tz);
 445
 446             fjx1             = _mm256_add_pd(fjx1,tx);
 447             fjy1             = _mm256_add_pd(fjy1,ty);
 448             fjz1             = _mm256_add_pd(fjz1,tz);
 449
 450             /**************************
 451              * CALCULATE INTERACTIONS *
 452              **************************/
 453
 454             /* COULOMB ELECTROSTATICS */
 455             velec            = _mm256_mul_pd(qq22,rinv22);
 456             felec            = _mm256_mul_pd(velec,rinvsq22);
 457
 458             /* Update potential sum for this i atom from the interaction with this j atom. */
 459             velecsum         = _mm256_add_pd(velecsum,velec);
 460
 461             fscal            = felec;
 462
 463             /* Calculate temporary vectorial force */
 464             tx               = _mm256_mul_pd(fscal,dx22);
 465             ty               = _mm256_mul_pd(fscal,dy22);
 466             tz               = _mm256_mul_pd(fscal,dz22);
 467
 468             /* Update vectorial force */
 469             fix2             = _mm256_add_pd(fix2,tx);
 470             fiy2             = _mm256_add_pd(fiy2,ty);
 471             fiz2             = _mm256_add_pd(fiz2,tz);
 472
 473             fjx2             = _mm256_add_pd(fjx2,tx);
 474             fjy2             = _mm256_add_pd(fjy2,ty);
 475             fjz2             = _mm256_add_pd(fjz2,tz);
 476
 477             /**************************
 478              * CALCULATE INTERACTIONS *
 479              **************************/
 480
 481             /* COULOMB ELECTROSTATICS */
 482             velec            = _mm256_mul_pd(qq23,rinv23);
 483             felec            = _mm256_mul_pd(velec,rinvsq23);
 484
 485             /* Update potential sum for this i atom from the interaction with this j atom. */
 486             velecsum         = _mm256_add_pd(velecsum,velec);
 487
 488             fscal            = felec;
 489
 490             /* Calculate temporary vectorial force */
 491             tx               = _mm256_mul_pd(fscal,dx23);
 492             ty               = _mm256_mul_pd(fscal,dy23);
 493             tz               = _mm256_mul_pd(fscal,dz23);
 494
 495             /* Update vectorial force */
 496             fix2             = _mm256_add_pd(fix2,tx);
 497             fiy2             = _mm256_add_pd(fiy2,ty);
 498             fiz2             = _mm256_add_pd(fiz2,tz);
 499
 500             fjx3             = _mm256_add_pd(fjx3,tx);
 501             fjy3             = _mm256_add_pd(fjy3,ty);
 502             fjz3             = _mm256_add_pd(fjz3,tz);
 503
 504             /**************************
 505              * CALCULATE INTERACTIONS *
 506              **************************/
 507
 508             /* COULOMB ELECTROSTATICS */
 509             velec            = _mm256_mul_pd(qq31,rinv31);
 510             felec            = _mm256_mul_pd(velec,rinvsq31);
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velecsum         = _mm256_add_pd(velecsum,velec);
 514
 515             fscal            = felec;
 516
 517             /* Calculate temporary vectorial force */
 518             tx               = _mm256_mul_pd(fscal,dx31);
 519             ty               = _mm256_mul_pd(fscal,dy31);
 520             tz               = _mm256_mul_pd(fscal,dz31);
 521
 522             /* Update vectorial force */
 523             fix3             = _mm256_add_pd(fix3,tx);
 524             fiy3             = _mm256_add_pd(fiy3,ty);
 525             fiz3             = _mm256_add_pd(fiz3,tz);
 526
 527             fjx1             = _mm256_add_pd(fjx1,tx);
 528             fjy1             = _mm256_add_pd(fjy1,ty);
 529             fjz1             = _mm256_add_pd(fjz1,tz);
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             /* COULOMB ELECTROSTATICS */
 536             velec            = _mm256_mul_pd(qq32,rinv32);
 537             felec            = _mm256_mul_pd(velec,rinvsq32);
 538
 539             /* Update potential sum for this i atom from the interaction with this j atom. */
 540             velecsum         = _mm256_add_pd(velecsum,velec);
 541
 542             fscal            = felec;
 543
 544             /* Calculate temporary vectorial force */
 545             tx               = _mm256_mul_pd(fscal,dx32);
 546             ty               = _mm256_mul_pd(fscal,dy32);
 547             tz               = _mm256_mul_pd(fscal,dz32);
 548
 549             /* Update vectorial force */
 550             fix3             = _mm256_add_pd(fix3,tx);
 551             fiy3             = _mm256_add_pd(fiy3,ty);
 552             fiz3             = _mm256_add_pd(fiz3,tz);
 553
 554             fjx2             = _mm256_add_pd(fjx2,tx);
 555             fjy2             = _mm256_add_pd(fjy2,ty);
 556             fjz2             = _mm256_add_pd(fjz2,tz);
 557
 558             /**************************
 559              * CALCULATE INTERACTIONS *
 560              **************************/
 561
 562             /* COULOMB ELECTROSTATICS */
 563             velec            = _mm256_mul_pd(qq33,rinv33);
 564             felec            = _mm256_mul_pd(velec,rinvsq33);
 565
 566             /* Update potential sum for this i atom from the interaction with this j atom. */
 567             velecsum         = _mm256_add_pd(velecsum,velec);
 568
 569             fscal            = felec;
 570
 571             /* Calculate temporary vectorial force */
 572             tx               = _mm256_mul_pd(fscal,dx33);
 573             ty               = _mm256_mul_pd(fscal,dy33);
 574             tz               = _mm256_mul_pd(fscal,dz33);
 575
 576             /* Update vectorial force */
 577             fix3             = _mm256_add_pd(fix3,tx);
 578             fiy3             = _mm256_add_pd(fiy3,ty);
 579             fiz3             = _mm256_add_pd(fiz3,tz);
 580
 581             fjx3             = _mm256_add_pd(fjx3,tx);
 582             fjy3             = _mm256_add_pd(fjy3,ty);
 583             fjz3             = _mm256_add_pd(fjz3,tz);
 584
 585             fjptrA             = f+j_coord_offsetA;
 586             fjptrB             = f+j_coord_offsetB;
 587             fjptrC             = f+j_coord_offsetC;
 588             fjptrD             = f+j_coord_offsetD;
 589
 590             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 591                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 592                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 593
 594             /* Inner loop uses 278 flops */
 595         }
 596
 597         if(jidx<j_index_end)
 598         {
 599
 600             /* Get j neighbor index, and coordinate index */
 601             jnrlistA         = jjnr[jidx];
 602             jnrlistB         = jjnr[jidx+1];
 603             jnrlistC         = jjnr[jidx+2];
 604             jnrlistD         = jjnr[jidx+3];
 605             /* Sign of each element will be negative for non-real atoms.
 606              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 607              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 608              */
 609             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 610
 611             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 612             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 613             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 614
 615             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 616             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 617             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 618             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 619             j_coord_offsetA  = DIM*jnrA;
 620             j_coord_offsetB  = DIM*jnrB;
 621             j_coord_offsetC  = DIM*jnrC;
 622             j_coord_offsetD  = DIM*jnrD;
 623
 624             /* load j atom coordinates */
 625             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 626                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 627                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 628                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 629
 630             /* Calculate displacement vector */
 631             dx00             = _mm256_sub_pd(ix0,jx0);
 632             dy00             = _mm256_sub_pd(iy0,jy0);
 633             dz00             = _mm256_sub_pd(iz0,jz0);
 634             dx11             = _mm256_sub_pd(ix1,jx1);
 635             dy11             = _mm256_sub_pd(iy1,jy1);
 636             dz11             = _mm256_sub_pd(iz1,jz1);
 637             dx12             = _mm256_sub_pd(ix1,jx2);
 638             dy12             = _mm256_sub_pd(iy1,jy2);
 639             dz12             = _mm256_sub_pd(iz1,jz2);
 640             dx13             = _mm256_sub_pd(ix1,jx3);
 641             dy13             = _mm256_sub_pd(iy1,jy3);
 642             dz13             = _mm256_sub_pd(iz1,jz3);
 643             dx21             = _mm256_sub_pd(ix2,jx1);
 644             dy21             = _mm256_sub_pd(iy2,jy1);
 645             dz21             = _mm256_sub_pd(iz2,jz1);
 646             dx22             = _mm256_sub_pd(ix2,jx2);
 647             dy22             = _mm256_sub_pd(iy2,jy2);
 648             dz22             = _mm256_sub_pd(iz2,jz2);
 649             dx23             = _mm256_sub_pd(ix2,jx3);
 650             dy23             = _mm256_sub_pd(iy2,jy3);
 651             dz23             = _mm256_sub_pd(iz2,jz3);
 652             dx31             = _mm256_sub_pd(ix3,jx1);
 653             dy31             = _mm256_sub_pd(iy3,jy1);
 654             dz31             = _mm256_sub_pd(iz3,jz1);
 655             dx32             = _mm256_sub_pd(ix3,jx2);
 656             dy32             = _mm256_sub_pd(iy3,jy2);
 657             dz32             = _mm256_sub_pd(iz3,jz2);
 658             dx33             = _mm256_sub_pd(ix3,jx3);
 659             dy33             = _mm256_sub_pd(iy3,jy3);
 660             dz33             = _mm256_sub_pd(iz3,jz3);
 661
 662             /* Calculate squared distance and things based on it */
 663             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 664             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 665             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 666             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 667             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 668             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 669             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 670             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 671             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 672             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 673
 674             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 675             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 676             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 677             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 678             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 679             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 680             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 681             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 682             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 683
 684             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 685             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 686             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 687             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 688             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 689             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 690             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 691             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 692             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 693             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 694
 695             fjx0             = _mm256_setzero_pd();
 696             fjy0             = _mm256_setzero_pd();
 697             fjz0             = _mm256_setzero_pd();
 698             fjx1             = _mm256_setzero_pd();
 699             fjy1             = _mm256_setzero_pd();
 700             fjz1             = _mm256_setzero_pd();
 701             fjx2             = _mm256_setzero_pd();
 702             fjy2             = _mm256_setzero_pd();
 703             fjz2             = _mm256_setzero_pd();
 704             fjx3             = _mm256_setzero_pd();
 705             fjy3             = _mm256_setzero_pd();
 706             fjz3             = _mm256_setzero_pd();
 707
 708             /**************************
 709              * CALCULATE INTERACTIONS *
 710              **************************/
 711
 712             /* LENNARD-JONES DISPERSION/REPULSION */
 713
 714             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 715             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 716             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 717             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 718             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 719
 720             /* Update potential sum for this i atom from the interaction with this j atom. */
 721             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 722             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 723
 724             fscal            = fvdw;
 725
 726             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 727
 728             /* Calculate temporary vectorial force */
 729             tx               = _mm256_mul_pd(fscal,dx00);
 730             ty               = _mm256_mul_pd(fscal,dy00);
 731             tz               = _mm256_mul_pd(fscal,dz00);
 732
 733             /* Update vectorial force */
 734             fix0             = _mm256_add_pd(fix0,tx);
 735             fiy0             = _mm256_add_pd(fiy0,ty);
 736             fiz0             = _mm256_add_pd(fiz0,tz);
 737
 738             fjx0             = _mm256_add_pd(fjx0,tx);
 739             fjy0             = _mm256_add_pd(fjy0,ty);
 740             fjz0             = _mm256_add_pd(fjz0,tz);
 741
 742             /**************************
 743              * CALCULATE INTERACTIONS *
 744              **************************/
 745
 746             /* COULOMB ELECTROSTATICS */
 747             velec            = _mm256_mul_pd(qq11,rinv11);
 748             felec            = _mm256_mul_pd(velec,rinvsq11);
 749
 750             /* Update potential sum for this i atom from the interaction with this j atom. */
 751             velec            = _mm256_andnot_pd(dummy_mask,velec);
 752             velecsum         = _mm256_add_pd(velecsum,velec);
 753
 754             fscal            = felec;
 755
 756             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 757
 758             /* Calculate temporary vectorial force */
 759             tx               = _mm256_mul_pd(fscal,dx11);
 760             ty               = _mm256_mul_pd(fscal,dy11);
 761             tz               = _mm256_mul_pd(fscal,dz11);
 762
 763             /* Update vectorial force */
 764             fix1             = _mm256_add_pd(fix1,tx);
 765             fiy1             = _mm256_add_pd(fiy1,ty);
 766             fiz1             = _mm256_add_pd(fiz1,tz);
 767
 768             fjx1             = _mm256_add_pd(fjx1,tx);
 769             fjy1             = _mm256_add_pd(fjy1,ty);
 770             fjz1             = _mm256_add_pd(fjz1,tz);
 771
 772             /**************************
 773              * CALCULATE INTERACTIONS *
 774              **************************/
 775
 776             /* COULOMB ELECTROSTATICS */
 777             velec            = _mm256_mul_pd(qq12,rinv12);
 778             felec            = _mm256_mul_pd(velec,rinvsq12);
 779
 780             /* Update potential sum for this i atom from the interaction with this j atom. */
 781             velec            = _mm256_andnot_pd(dummy_mask,velec);
 782             velecsum         = _mm256_add_pd(velecsum,velec);
 783
 784             fscal            = felec;
 785
 786             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 787
 788             /* Calculate temporary vectorial force */
 789             tx               = _mm256_mul_pd(fscal,dx12);
 790             ty               = _mm256_mul_pd(fscal,dy12);
 791             tz               = _mm256_mul_pd(fscal,dz12);
 792
 793             /* Update vectorial force */
 794             fix1             = _mm256_add_pd(fix1,tx);
 795             fiy1             = _mm256_add_pd(fiy1,ty);
 796             fiz1             = _mm256_add_pd(fiz1,tz);
 797
 798             fjx2             = _mm256_add_pd(fjx2,tx);
 799             fjy2             = _mm256_add_pd(fjy2,ty);
 800             fjz2             = _mm256_add_pd(fjz2,tz);
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             /* COULOMB ELECTROSTATICS */
 807             velec            = _mm256_mul_pd(qq13,rinv13);
 808             felec            = _mm256_mul_pd(velec,rinvsq13);
 809
 810             /* Update potential sum for this i atom from the interaction with this j atom. */
 811             velec            = _mm256_andnot_pd(dummy_mask,velec);
 812             velecsum         = _mm256_add_pd(velecsum,velec);
 813
 814             fscal            = felec;
 815
 816             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 817
 818             /* Calculate temporary vectorial force */
 819             tx               = _mm256_mul_pd(fscal,dx13);
 820             ty               = _mm256_mul_pd(fscal,dy13);
 821             tz               = _mm256_mul_pd(fscal,dz13);
 822
 823             /* Update vectorial force */
 824             fix1             = _mm256_add_pd(fix1,tx);
 825             fiy1             = _mm256_add_pd(fiy1,ty);
 826             fiz1             = _mm256_add_pd(fiz1,tz);
 827
 828             fjx3             = _mm256_add_pd(fjx3,tx);
 829             fjy3             = _mm256_add_pd(fjy3,ty);
 830             fjz3             = _mm256_add_pd(fjz3,tz);
 831
 832             /**************************
 833              * CALCULATE INTERACTIONS *
 834              **************************/
 835
 836             /* COULOMB ELECTROSTATICS */
 837             velec            = _mm256_mul_pd(qq21,rinv21);
 838             felec            = _mm256_mul_pd(velec,rinvsq21);
 839
 840             /* Update potential sum for this i atom from the interaction with this j atom. */
 841             velec            = _mm256_andnot_pd(dummy_mask,velec);
 842             velecsum         = _mm256_add_pd(velecsum,velec);
 843
 844             fscal            = felec;
 845
 846             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 847
 848             /* Calculate temporary vectorial force */
 849             tx               = _mm256_mul_pd(fscal,dx21);
 850             ty               = _mm256_mul_pd(fscal,dy21);
 851             tz               = _mm256_mul_pd(fscal,dz21);
 852
 853             /* Update vectorial force */
 854             fix2             = _mm256_add_pd(fix2,tx);
 855             fiy2             = _mm256_add_pd(fiy2,ty);
 856             fiz2             = _mm256_add_pd(fiz2,tz);
 857
 858             fjx1             = _mm256_add_pd(fjx1,tx);
 859             fjy1             = _mm256_add_pd(fjy1,ty);
 860             fjz1             = _mm256_add_pd(fjz1,tz);
 861
 862             /**************************
 863              * CALCULATE INTERACTIONS *
 864              **************************/
 865
 866             /* COULOMB ELECTROSTATICS */
 867             velec            = _mm256_mul_pd(qq22,rinv22);
 868             felec            = _mm256_mul_pd(velec,rinvsq22);
 869
 870             /* Update potential sum for this i atom from the interaction with this j atom. */
 871             velec            = _mm256_andnot_pd(dummy_mask,velec);
 872             velecsum         = _mm256_add_pd(velecsum,velec);
 873
 874             fscal            = felec;
 875
 876             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 877
 878             /* Calculate temporary vectorial force */
 879             tx               = _mm256_mul_pd(fscal,dx22);
 880             ty               = _mm256_mul_pd(fscal,dy22);
 881             tz               = _mm256_mul_pd(fscal,dz22);
 882
 883             /* Update vectorial force */
 884             fix2             = _mm256_add_pd(fix2,tx);
 885             fiy2             = _mm256_add_pd(fiy2,ty);
 886             fiz2             = _mm256_add_pd(fiz2,tz);
 887
 888             fjx2             = _mm256_add_pd(fjx2,tx);
 889             fjy2             = _mm256_add_pd(fjy2,ty);
 890             fjz2             = _mm256_add_pd(fjz2,tz);
 891
 892             /**************************
 893              * CALCULATE INTERACTIONS *
 894              **************************/
 895
 896             /* COULOMB ELECTROSTATICS */
 897             velec            = _mm256_mul_pd(qq23,rinv23);
 898             felec            = _mm256_mul_pd(velec,rinvsq23);
 899
 900             /* Update potential sum for this i atom from the interaction with this j atom. */
 901             velec            = _mm256_andnot_pd(dummy_mask,velec);
 902             velecsum         = _mm256_add_pd(velecsum,velec);
 903
 904             fscal            = felec;
 905
 906             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 907
 908             /* Calculate temporary vectorial force */
 909             tx               = _mm256_mul_pd(fscal,dx23);
 910             ty               = _mm256_mul_pd(fscal,dy23);
 911             tz               = _mm256_mul_pd(fscal,dz23);
 912
 913             /* Update vectorial force */
 914             fix2             = _mm256_add_pd(fix2,tx);
 915             fiy2             = _mm256_add_pd(fiy2,ty);
 916             fiz2             = _mm256_add_pd(fiz2,tz);
 917
 918             fjx3             = _mm256_add_pd(fjx3,tx);
 919             fjy3             = _mm256_add_pd(fjy3,ty);
 920             fjz3             = _mm256_add_pd(fjz3,tz);
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             /* COULOMB ELECTROSTATICS */
 927             velec            = _mm256_mul_pd(qq31,rinv31);
 928             felec            = _mm256_mul_pd(velec,rinvsq31);
 929
 930             /* Update potential sum for this i atom from the interaction with this j atom. */
 931             velec            = _mm256_andnot_pd(dummy_mask,velec);
 932             velecsum         = _mm256_add_pd(velecsum,velec);
 933
 934             fscal            = felec;
 935
 936             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 937
 938             /* Calculate temporary vectorial force */
 939             tx               = _mm256_mul_pd(fscal,dx31);
 940             ty               = _mm256_mul_pd(fscal,dy31);
 941             tz               = _mm256_mul_pd(fscal,dz31);
 942
 943             /* Update vectorial force */
 944             fix3             = _mm256_add_pd(fix3,tx);
 945             fiy3             = _mm256_add_pd(fiy3,ty);
 946             fiz3             = _mm256_add_pd(fiz3,tz);
 947
 948             fjx1             = _mm256_add_pd(fjx1,tx);
 949             fjy1             = _mm256_add_pd(fjy1,ty);
 950             fjz1             = _mm256_add_pd(fjz1,tz);
 951
 952             /**************************
 953              * CALCULATE INTERACTIONS *
 954              **************************/
 955
 956             /* COULOMB ELECTROSTATICS */
 957             velec            = _mm256_mul_pd(qq32,rinv32);
 958             felec            = _mm256_mul_pd(velec,rinvsq32);
 959
 960             /* Update potential sum for this i atom from the interaction with this j atom. */
 961             velec            = _mm256_andnot_pd(dummy_mask,velec);
 962             velecsum         = _mm256_add_pd(velecsum,velec);
 963
 964             fscal            = felec;
 965
 966             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 967
 968             /* Calculate temporary vectorial force */
 969             tx               = _mm256_mul_pd(fscal,dx32);
 970             ty               = _mm256_mul_pd(fscal,dy32);
 971             tz               = _mm256_mul_pd(fscal,dz32);
 972
 973             /* Update vectorial force */
 974             fix3             = _mm256_add_pd(fix3,tx);
 975             fiy3             = _mm256_add_pd(fiy3,ty);
 976             fiz3             = _mm256_add_pd(fiz3,tz);
 977
 978             fjx2             = _mm256_add_pd(fjx2,tx);
 979             fjy2             = _mm256_add_pd(fjy2,ty);
 980             fjz2             = _mm256_add_pd(fjz2,tz);
 981
 982             /**************************
 983              * CALCULATE INTERACTIONS *
 984              **************************/
 985
 986             /* COULOMB ELECTROSTATICS */
 987             velec            = _mm256_mul_pd(qq33,rinv33);
 988             felec            = _mm256_mul_pd(velec,rinvsq33);
 989
 990             /* Update potential sum for this i atom from the interaction with this j atom. */
 991             velec            = _mm256_andnot_pd(dummy_mask,velec);
 992             velecsum         = _mm256_add_pd(velecsum,velec);
 993
 994             fscal            = felec;
 995
 996             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 997
 998             /* Calculate temporary vectorial force */
 999             tx               = _mm256_mul_pd(fscal,dx33);
1000             ty               = _mm256_mul_pd(fscal,dy33);
1001             tz               = _mm256_mul_pd(fscal,dz33);
1002
1003             /* Update vectorial force */
1004             fix3             = _mm256_add_pd(fix3,tx);
1005             fiy3             = _mm256_add_pd(fiy3,ty);
1006             fiz3             = _mm256_add_pd(fiz3,tz);
1007
1008             fjx3             = _mm256_add_pd(fjx3,tx);
1009             fjy3             = _mm256_add_pd(fjy3,ty);
1010             fjz3             = _mm256_add_pd(fjz3,tz);
1011
1012             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1013             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1014             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1015             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1016
1017             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1018                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1019                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1020
1021             /* Inner loop uses 278 flops */
1022         }
1023
1024         /* End of innermost loop */
1025
1026         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1027                                                  f+i_coord_offset,fshift+i_shift_offset);
1028
1029         ggid                        = gid[iidx];
1030         /* Update potential energies */
1031         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1032         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1033
1034         /* Increment number of inner iterations */
1035         inneriter                  += j_index_end - j_index_start;
1036
1037         /* Outer loop uses 26 flops */
1038     }
1039
1040     /* Increment number of outer iterations */
1041     outeriter        += nri;
1042
1043     /* Update outer/inner flops */
1044
1045     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1046 }
1047 /*
1048  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1049  * Electrostatics interaction: Coulomb
1050  * VdW interaction:            LennardJones
1051  * Geometry:                   Water4-Water4
1052  * Calculate force/pot:        Force
1053  */
1054 void
1055 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_double
1056                     (t_nblist                    * gmx_restrict       nlist,
1057                      rvec                        * gmx_restrict          xx,
1058                      rvec                        * gmx_restrict          ff,
1059                      t_forcerec                  * gmx_restrict          fr,
1060                      t_mdatoms                   * gmx_restrict     mdatoms,
1061                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1062                      t_nrnb                      * gmx_restrict        nrnb)
1063 {
1064     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1065      * just 0 for non-waters.
1066      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1067      * jnr indices corresponding to data put in the four positions in the SIMD register.
1068      */
1069     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1070     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1071     int              jnrA,jnrB,jnrC,jnrD;
1072     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1073     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1074     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1075     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1076     real             rcutoff_scalar;
1077     real             *shiftvec,*fshift,*x,*f;
1078     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1079     real             scratch[4*DIM];
1080     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1081     real *           vdwioffsetptr0;
1082     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1083     real *           vdwioffsetptr1;
1084     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1085     real *           vdwioffsetptr2;
1086     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1087     real *           vdwioffsetptr3;
1088     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1089     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1090     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1091     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1092     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1093     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1094     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1095     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1096     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1097     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1098     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1099     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1100     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1101     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1102     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1103     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1104     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1105     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1106     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1107     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1108     real             *charge;
1109     int              nvdwtype;
1110     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1111     int              *vdwtype;
1112     real             *vdwparam;
1113     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1114     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1115     __m256d          dummy_mask,cutoff_mask;
1116     __m128           tmpmask0,tmpmask1;
1117     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1118     __m256d          one     = _mm256_set1_pd(1.0);
1119     __m256d          two     = _mm256_set1_pd(2.0);
1120     x                = xx[0];
1121     f                = ff[0];
1122
1123     nri              = nlist->nri;
1124     iinr             = nlist->iinr;
1125     jindex           = nlist->jindex;
1126     jjnr             = nlist->jjnr;
1127     shiftidx         = nlist->shift;
1128     gid              = nlist->gid;
1129     shiftvec         = fr->shift_vec[0];
1130     fshift           = fr->fshift[0];
1131     facel            = _mm256_set1_pd(fr->epsfac);
1132     charge           = mdatoms->chargeA;
1133     nvdwtype         = fr->ntype;
1134     vdwparam         = fr->nbfp;
1135     vdwtype          = mdatoms->typeA;
1136
1137     /* Setup water-specific parameters */
1138     inr              = nlist->iinr[0];
1139     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1140     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1141     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1142     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1143
1144     jq1              = _mm256_set1_pd(charge[inr+1]);
1145     jq2              = _mm256_set1_pd(charge[inr+2]);
1146     jq3              = _mm256_set1_pd(charge[inr+3]);
1147     vdwjidx0A        = 2*vdwtype[inr+0];
1148     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1149     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1150     qq11             = _mm256_mul_pd(iq1,jq1);
1151     qq12             = _mm256_mul_pd(iq1,jq2);
1152     qq13             = _mm256_mul_pd(iq1,jq3);
1153     qq21             = _mm256_mul_pd(iq2,jq1);
1154     qq22             = _mm256_mul_pd(iq2,jq2);
1155     qq23             = _mm256_mul_pd(iq2,jq3);
1156     qq31             = _mm256_mul_pd(iq3,jq1);
1157     qq32             = _mm256_mul_pd(iq3,jq2);
1158     qq33             = _mm256_mul_pd(iq3,jq3);
1159
1160     /* Avoid stupid compiler warnings */
1161     jnrA = jnrB = jnrC = jnrD = 0;
1162     j_coord_offsetA = 0;
1163     j_coord_offsetB = 0;
1164     j_coord_offsetC = 0;
1165     j_coord_offsetD = 0;
1166
1167     outeriter        = 0;
1168     inneriter        = 0;
1169
1170     for(iidx=0;iidx<4*DIM;iidx++)
1171     {
1172         scratch[iidx] = 0.0;
1173     }
1174
1175     /* Start outer loop over neighborlists */
1176     for(iidx=0; iidx<nri; iidx++)
1177     {
1178         /* Load shift vector for this list */
1179         i_shift_offset   = DIM*shiftidx[iidx];
1180
1181         /* Load limits for loop over neighbors */
1182         j_index_start    = jindex[iidx];
1183         j_index_end      = jindex[iidx+1];
1184
1185         /* Get outer coordinate index */
1186         inr              = iinr[iidx];
1187         i_coord_offset   = DIM*inr;
1188
1189         /* Load i particle coords and add shift vector */
1190         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1191                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1192
1193         fix0             = _mm256_setzero_pd();
1194         fiy0             = _mm256_setzero_pd();
1195         fiz0             = _mm256_setzero_pd();
1196         fix1             = _mm256_setzero_pd();
1197         fiy1             = _mm256_setzero_pd();
1198         fiz1             = _mm256_setzero_pd();
1199         fix2             = _mm256_setzero_pd();
1200         fiy2             = _mm256_setzero_pd();
1201         fiz2             = _mm256_setzero_pd();
1202         fix3             = _mm256_setzero_pd();
1203         fiy3             = _mm256_setzero_pd();
1204         fiz3             = _mm256_setzero_pd();
1205
1206         /* Start inner kernel loop */
1207         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1208         {
1209
1210             /* Get j neighbor index, and coordinate index */
1211             jnrA             = jjnr[jidx];
1212             jnrB             = jjnr[jidx+1];
1213             jnrC             = jjnr[jidx+2];
1214             jnrD             = jjnr[jidx+3];
1215             j_coord_offsetA  = DIM*jnrA;
1216             j_coord_offsetB  = DIM*jnrB;
1217             j_coord_offsetC  = DIM*jnrC;
1218             j_coord_offsetD  = DIM*jnrD;
1219
1220             /* load j atom coordinates */
1221             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1222                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1223                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1224                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1225
1226             /* Calculate displacement vector */
1227             dx00             = _mm256_sub_pd(ix0,jx0);
1228             dy00             = _mm256_sub_pd(iy0,jy0);
1229             dz00             = _mm256_sub_pd(iz0,jz0);
1230             dx11             = _mm256_sub_pd(ix1,jx1);
1231             dy11             = _mm256_sub_pd(iy1,jy1);
1232             dz11             = _mm256_sub_pd(iz1,jz1);
1233             dx12             = _mm256_sub_pd(ix1,jx2);
1234             dy12             = _mm256_sub_pd(iy1,jy2);
1235             dz12             = _mm256_sub_pd(iz1,jz2);
1236             dx13             = _mm256_sub_pd(ix1,jx3);
1237             dy13             = _mm256_sub_pd(iy1,jy3);
1238             dz13             = _mm256_sub_pd(iz1,jz3);
1239             dx21             = _mm256_sub_pd(ix2,jx1);
1240             dy21             = _mm256_sub_pd(iy2,jy1);
1241             dz21             = _mm256_sub_pd(iz2,jz1);
1242             dx22             = _mm256_sub_pd(ix2,jx2);
1243             dy22             = _mm256_sub_pd(iy2,jy2);
1244             dz22             = _mm256_sub_pd(iz2,jz2);
1245             dx23             = _mm256_sub_pd(ix2,jx3);
1246             dy23             = _mm256_sub_pd(iy2,jy3);
1247             dz23             = _mm256_sub_pd(iz2,jz3);
1248             dx31             = _mm256_sub_pd(ix3,jx1);
1249             dy31             = _mm256_sub_pd(iy3,jy1);
1250             dz31             = _mm256_sub_pd(iz3,jz1);
1251             dx32             = _mm256_sub_pd(ix3,jx2);
1252             dy32             = _mm256_sub_pd(iy3,jy2);
1253             dz32             = _mm256_sub_pd(iz3,jz2);
1254             dx33             = _mm256_sub_pd(ix3,jx3);
1255             dy33             = _mm256_sub_pd(iy3,jy3);
1256             dz33             = _mm256_sub_pd(iz3,jz3);
1257
1258             /* Calculate squared distance and things based on it */
1259             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1260             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1261             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1262             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1263             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1264             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1265             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1266             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1267             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1268             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1269
1270             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1271             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1272             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1273             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1274             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1275             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1276             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1277             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1278             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1279
1280             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1281             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1282             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1283             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1284             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1285             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1286             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1287             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1288             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1289             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1290
1291             fjx0             = _mm256_setzero_pd();
1292             fjy0             = _mm256_setzero_pd();
1293             fjz0             = _mm256_setzero_pd();
1294             fjx1             = _mm256_setzero_pd();
1295             fjy1             = _mm256_setzero_pd();
1296             fjz1             = _mm256_setzero_pd();
1297             fjx2             = _mm256_setzero_pd();
1298             fjy2             = _mm256_setzero_pd();
1299             fjz2             = _mm256_setzero_pd();
1300             fjx3             = _mm256_setzero_pd();
1301             fjy3             = _mm256_setzero_pd();
1302             fjz3             = _mm256_setzero_pd();
1303
1304             /**************************
1305              * CALCULATE INTERACTIONS *
1306              **************************/
1307
1308             /* LENNARD-JONES DISPERSION/REPULSION */
1309
1310             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1311             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1312
1313             fscal            = fvdw;
1314
1315             /* Calculate temporary vectorial force */
1316             tx               = _mm256_mul_pd(fscal,dx00);
1317             ty               = _mm256_mul_pd(fscal,dy00);
1318             tz               = _mm256_mul_pd(fscal,dz00);
1319
1320             /* Update vectorial force */
1321             fix0             = _mm256_add_pd(fix0,tx);
1322             fiy0             = _mm256_add_pd(fiy0,ty);
1323             fiz0             = _mm256_add_pd(fiz0,tz);
1324
1325             fjx0             = _mm256_add_pd(fjx0,tx);
1326             fjy0             = _mm256_add_pd(fjy0,ty);
1327             fjz0             = _mm256_add_pd(fjz0,tz);
1328
1329             /**************************
1330              * CALCULATE INTERACTIONS *
1331              **************************/
1332
1333             /* COULOMB ELECTROSTATICS */
1334             velec            = _mm256_mul_pd(qq11,rinv11);
1335             felec            = _mm256_mul_pd(velec,rinvsq11);
1336
1337             fscal            = felec;
1338
1339             /* Calculate temporary vectorial force */
1340             tx               = _mm256_mul_pd(fscal,dx11);
1341             ty               = _mm256_mul_pd(fscal,dy11);
1342             tz               = _mm256_mul_pd(fscal,dz11);
1343
1344             /* Update vectorial force */
1345             fix1             = _mm256_add_pd(fix1,tx);
1346             fiy1             = _mm256_add_pd(fiy1,ty);
1347             fiz1             = _mm256_add_pd(fiz1,tz);
1348
1349             fjx1             = _mm256_add_pd(fjx1,tx);
1350             fjy1             = _mm256_add_pd(fjy1,ty);
1351             fjz1             = _mm256_add_pd(fjz1,tz);
1352
1353             /**************************
1354              * CALCULATE INTERACTIONS *
1355              **************************/
1356
1357             /* COULOMB ELECTROSTATICS */
1358             velec            = _mm256_mul_pd(qq12,rinv12);
1359             felec            = _mm256_mul_pd(velec,rinvsq12);
1360
1361             fscal            = felec;
1362
1363             /* Calculate temporary vectorial force */
1364             tx               = _mm256_mul_pd(fscal,dx12);
1365             ty               = _mm256_mul_pd(fscal,dy12);
1366             tz               = _mm256_mul_pd(fscal,dz12);
1367
1368             /* Update vectorial force */
1369             fix1             = _mm256_add_pd(fix1,tx);
1370             fiy1             = _mm256_add_pd(fiy1,ty);
1371             fiz1             = _mm256_add_pd(fiz1,tz);
1372
1373             fjx2             = _mm256_add_pd(fjx2,tx);
1374             fjy2             = _mm256_add_pd(fjy2,ty);
1375             fjz2             = _mm256_add_pd(fjz2,tz);
1376
1377             /**************************
1378              * CALCULATE INTERACTIONS *
1379              **************************/
1380
1381             /* COULOMB ELECTROSTATICS */
1382             velec            = _mm256_mul_pd(qq13,rinv13);
1383             felec            = _mm256_mul_pd(velec,rinvsq13);
1384
1385             fscal            = felec;
1386
1387             /* Calculate temporary vectorial force */
1388             tx               = _mm256_mul_pd(fscal,dx13);
1389             ty               = _mm256_mul_pd(fscal,dy13);
1390             tz               = _mm256_mul_pd(fscal,dz13);
1391
1392             /* Update vectorial force */
1393             fix1             = _mm256_add_pd(fix1,tx);
1394             fiy1             = _mm256_add_pd(fiy1,ty);
1395             fiz1             = _mm256_add_pd(fiz1,tz);
1396
1397             fjx3             = _mm256_add_pd(fjx3,tx);
1398             fjy3             = _mm256_add_pd(fjy3,ty);
1399             fjz3             = _mm256_add_pd(fjz3,tz);
1400
1401             /**************************
1402              * CALCULATE INTERACTIONS *
1403              **************************/
1404
1405             /* COULOMB ELECTROSTATICS */
1406             velec            = _mm256_mul_pd(qq21,rinv21);
1407             felec            = _mm256_mul_pd(velec,rinvsq21);
1408
1409             fscal            = felec;
1410
1411             /* Calculate temporary vectorial force */
1412             tx               = _mm256_mul_pd(fscal,dx21);
1413             ty               = _mm256_mul_pd(fscal,dy21);
1414             tz               = _mm256_mul_pd(fscal,dz21);
1415
1416             /* Update vectorial force */
1417             fix2             = _mm256_add_pd(fix2,tx);
1418             fiy2             = _mm256_add_pd(fiy2,ty);
1419             fiz2             = _mm256_add_pd(fiz2,tz);
1420
1421             fjx1             = _mm256_add_pd(fjx1,tx);
1422             fjy1             = _mm256_add_pd(fjy1,ty);
1423             fjz1             = _mm256_add_pd(fjz1,tz);
1424
1425             /**************************
1426              * CALCULATE INTERACTIONS *
1427              **************************/
1428
1429             /* COULOMB ELECTROSTATICS */
1430             velec            = _mm256_mul_pd(qq22,rinv22);
1431             felec            = _mm256_mul_pd(velec,rinvsq22);
1432
1433             fscal            = felec;
1434
1435             /* Calculate temporary vectorial force */
1436             tx               = _mm256_mul_pd(fscal,dx22);
1437             ty               = _mm256_mul_pd(fscal,dy22);
1438             tz               = _mm256_mul_pd(fscal,dz22);
1439
1440             /* Update vectorial force */
1441             fix2             = _mm256_add_pd(fix2,tx);
1442             fiy2             = _mm256_add_pd(fiy2,ty);
1443             fiz2             = _mm256_add_pd(fiz2,tz);
1444
1445             fjx2             = _mm256_add_pd(fjx2,tx);
1446             fjy2             = _mm256_add_pd(fjy2,ty);
1447             fjz2             = _mm256_add_pd(fjz2,tz);
1448
1449             /**************************
1450              * CALCULATE INTERACTIONS *
1451              **************************/
1452
1453             /* COULOMB ELECTROSTATICS */
1454             velec            = _mm256_mul_pd(qq23,rinv23);
1455             felec            = _mm256_mul_pd(velec,rinvsq23);
1456
1457             fscal            = felec;
1458
1459             /* Calculate temporary vectorial force */
1460             tx               = _mm256_mul_pd(fscal,dx23);
1461             ty               = _mm256_mul_pd(fscal,dy23);
1462             tz               = _mm256_mul_pd(fscal,dz23);
1463
1464             /* Update vectorial force */
1465             fix2             = _mm256_add_pd(fix2,tx);
1466             fiy2             = _mm256_add_pd(fiy2,ty);
1467             fiz2             = _mm256_add_pd(fiz2,tz);
1468
1469             fjx3             = _mm256_add_pd(fjx3,tx);
1470             fjy3             = _mm256_add_pd(fjy3,ty);
1471             fjz3             = _mm256_add_pd(fjz3,tz);
1472
1473             /**************************
1474              * CALCULATE INTERACTIONS *
1475              **************************/
1476
1477             /* COULOMB ELECTROSTATICS */
1478             velec            = _mm256_mul_pd(qq31,rinv31);
1479             felec            = _mm256_mul_pd(velec,rinvsq31);
1480
1481             fscal            = felec;
1482
1483             /* Calculate temporary vectorial force */
1484             tx               = _mm256_mul_pd(fscal,dx31);
1485             ty               = _mm256_mul_pd(fscal,dy31);
1486             tz               = _mm256_mul_pd(fscal,dz31);
1487
1488             /* Update vectorial force */
1489             fix3             = _mm256_add_pd(fix3,tx);
1490             fiy3             = _mm256_add_pd(fiy3,ty);
1491             fiz3             = _mm256_add_pd(fiz3,tz);
1492
1493             fjx1             = _mm256_add_pd(fjx1,tx);
1494             fjy1             = _mm256_add_pd(fjy1,ty);
1495             fjz1             = _mm256_add_pd(fjz1,tz);
1496
1497             /**************************
1498              * CALCULATE INTERACTIONS *
1499              **************************/
1500
1501             /* COULOMB ELECTROSTATICS */
1502             velec            = _mm256_mul_pd(qq32,rinv32);
1503             felec            = _mm256_mul_pd(velec,rinvsq32);
1504
1505             fscal            = felec;
1506
1507             /* Calculate temporary vectorial force */
1508             tx               = _mm256_mul_pd(fscal,dx32);
1509             ty               = _mm256_mul_pd(fscal,dy32);
1510             tz               = _mm256_mul_pd(fscal,dz32);
1511
1512             /* Update vectorial force */
1513             fix3             = _mm256_add_pd(fix3,tx);
1514             fiy3             = _mm256_add_pd(fiy3,ty);
1515             fiz3             = _mm256_add_pd(fiz3,tz);
1516
1517             fjx2             = _mm256_add_pd(fjx2,tx);
1518             fjy2             = _mm256_add_pd(fjy2,ty);
1519             fjz2             = _mm256_add_pd(fjz2,tz);
1520
1521             /**************************
1522              * CALCULATE INTERACTIONS *
1523              **************************/
1524
1525             /* COULOMB ELECTROSTATICS */
1526             velec            = _mm256_mul_pd(qq33,rinv33);
1527             felec            = _mm256_mul_pd(velec,rinvsq33);
1528
1529             fscal            = felec;
1530
1531             /* Calculate temporary vectorial force */
1532             tx               = _mm256_mul_pd(fscal,dx33);
1533             ty               = _mm256_mul_pd(fscal,dy33);
1534             tz               = _mm256_mul_pd(fscal,dz33);
1535
1536             /* Update vectorial force */
1537             fix3             = _mm256_add_pd(fix3,tx);
1538             fiy3             = _mm256_add_pd(fiy3,ty);
1539             fiz3             = _mm256_add_pd(fiz3,tz);
1540
1541             fjx3             = _mm256_add_pd(fjx3,tx);
1542             fjy3             = _mm256_add_pd(fjy3,ty);
1543             fjz3             = _mm256_add_pd(fjz3,tz);
1544
1545             fjptrA             = f+j_coord_offsetA;
1546             fjptrB             = f+j_coord_offsetB;
1547             fjptrC             = f+j_coord_offsetC;
1548             fjptrD             = f+j_coord_offsetD;
1549
1550             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1551                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1552                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1553
1554             /* Inner loop uses 264 flops */
1555         }
1556
1557         if(jidx<j_index_end)
1558         {
1559
1560             /* Get j neighbor index, and coordinate index */
1561             jnrlistA         = jjnr[jidx];
1562             jnrlistB         = jjnr[jidx+1];
1563             jnrlistC         = jjnr[jidx+2];
1564             jnrlistD         = jjnr[jidx+3];
1565             /* Sign of each element will be negative for non-real atoms.
1566              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1567              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1568              */
1569             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1570
1571             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1572             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1573             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1574
1575             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1576             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1577             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1578             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1579             j_coord_offsetA  = DIM*jnrA;
1580             j_coord_offsetB  = DIM*jnrB;
1581             j_coord_offsetC  = DIM*jnrC;
1582             j_coord_offsetD  = DIM*jnrD;
1583
1584             /* load j atom coordinates */
1585             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1586                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1587                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1588                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1589
1590             /* Calculate displacement vector */
1591             dx00             = _mm256_sub_pd(ix0,jx0);
1592             dy00             = _mm256_sub_pd(iy0,jy0);
1593             dz00             = _mm256_sub_pd(iz0,jz0);
1594             dx11             = _mm256_sub_pd(ix1,jx1);
1595             dy11             = _mm256_sub_pd(iy1,jy1);
1596             dz11             = _mm256_sub_pd(iz1,jz1);
1597             dx12             = _mm256_sub_pd(ix1,jx2);
1598             dy12             = _mm256_sub_pd(iy1,jy2);
1599             dz12             = _mm256_sub_pd(iz1,jz2);
1600             dx13             = _mm256_sub_pd(ix1,jx3);
1601             dy13             = _mm256_sub_pd(iy1,jy3);
1602             dz13             = _mm256_sub_pd(iz1,jz3);
1603             dx21             = _mm256_sub_pd(ix2,jx1);
1604             dy21             = _mm256_sub_pd(iy2,jy1);
1605             dz21             = _mm256_sub_pd(iz2,jz1);
1606             dx22             = _mm256_sub_pd(ix2,jx2);
1607             dy22             = _mm256_sub_pd(iy2,jy2);
1608             dz22             = _mm256_sub_pd(iz2,jz2);
1609             dx23             = _mm256_sub_pd(ix2,jx3);
1610             dy23             = _mm256_sub_pd(iy2,jy3);
1611             dz23             = _mm256_sub_pd(iz2,jz3);
1612             dx31             = _mm256_sub_pd(ix3,jx1);
1613             dy31             = _mm256_sub_pd(iy3,jy1);
1614             dz31             = _mm256_sub_pd(iz3,jz1);
1615             dx32             = _mm256_sub_pd(ix3,jx2);
1616             dy32             = _mm256_sub_pd(iy3,jy2);
1617             dz32             = _mm256_sub_pd(iz3,jz2);
1618             dx33             = _mm256_sub_pd(ix3,jx3);
1619             dy33             = _mm256_sub_pd(iy3,jy3);
1620             dz33             = _mm256_sub_pd(iz3,jz3);
1621
1622             /* Calculate squared distance and things based on it */
1623             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1624             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1625             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1626             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1627             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1628             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1629             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1630             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1631             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1632             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1633
1634             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1635             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1636             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1637             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1638             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1639             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1640             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1641             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1642             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1643
1644             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1645             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1646             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1647             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1648             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1649             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1650             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1651             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1652             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1653             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1654
1655             fjx0             = _mm256_setzero_pd();
1656             fjy0             = _mm256_setzero_pd();
1657             fjz0             = _mm256_setzero_pd();
1658             fjx1             = _mm256_setzero_pd();
1659             fjy1             = _mm256_setzero_pd();
1660             fjz1             = _mm256_setzero_pd();
1661             fjx2             = _mm256_setzero_pd();
1662             fjy2             = _mm256_setzero_pd();
1663             fjz2             = _mm256_setzero_pd();
1664             fjx3             = _mm256_setzero_pd();
1665             fjy3             = _mm256_setzero_pd();
1666             fjz3             = _mm256_setzero_pd();
1667
1668             /**************************
1669              * CALCULATE INTERACTIONS *
1670              **************************/
1671
1672             /* LENNARD-JONES DISPERSION/REPULSION */
1673
1674             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1675             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1676
1677             fscal            = fvdw;
1678
1679             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1680
1681             /* Calculate temporary vectorial force */
1682             tx               = _mm256_mul_pd(fscal,dx00);
1683             ty               = _mm256_mul_pd(fscal,dy00);
1684             tz               = _mm256_mul_pd(fscal,dz00);
1685
1686             /* Update vectorial force */
1687             fix0             = _mm256_add_pd(fix0,tx);
1688             fiy0             = _mm256_add_pd(fiy0,ty);
1689             fiz0             = _mm256_add_pd(fiz0,tz);
1690
1691             fjx0             = _mm256_add_pd(fjx0,tx);
1692             fjy0             = _mm256_add_pd(fjy0,ty);
1693             fjz0             = _mm256_add_pd(fjz0,tz);
1694
1695             /**************************
1696              * CALCULATE INTERACTIONS *
1697              **************************/
1698
1699             /* COULOMB ELECTROSTATICS */
1700             velec            = _mm256_mul_pd(qq11,rinv11);
1701             felec            = _mm256_mul_pd(velec,rinvsq11);
1702
1703             fscal            = felec;
1704
1705             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1706
1707             /* Calculate temporary vectorial force */
1708             tx               = _mm256_mul_pd(fscal,dx11);
1709             ty               = _mm256_mul_pd(fscal,dy11);
1710             tz               = _mm256_mul_pd(fscal,dz11);
1711
1712             /* Update vectorial force */
1713             fix1             = _mm256_add_pd(fix1,tx);
1714             fiy1             = _mm256_add_pd(fiy1,ty);
1715             fiz1             = _mm256_add_pd(fiz1,tz);
1716
1717             fjx1             = _mm256_add_pd(fjx1,tx);
1718             fjy1             = _mm256_add_pd(fjy1,ty);
1719             fjz1             = _mm256_add_pd(fjz1,tz);
1720
1721             /**************************
1722              * CALCULATE INTERACTIONS *
1723              **************************/
1724
1725             /* COULOMB ELECTROSTATICS */
1726             velec            = _mm256_mul_pd(qq12,rinv12);
1727             felec            = _mm256_mul_pd(velec,rinvsq12);
1728
1729             fscal            = felec;
1730
1731             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1732
1733             /* Calculate temporary vectorial force */
1734             tx               = _mm256_mul_pd(fscal,dx12);
1735             ty               = _mm256_mul_pd(fscal,dy12);
1736             tz               = _mm256_mul_pd(fscal,dz12);
1737
1738             /* Update vectorial force */
1739             fix1             = _mm256_add_pd(fix1,tx);
1740             fiy1             = _mm256_add_pd(fiy1,ty);
1741             fiz1             = _mm256_add_pd(fiz1,tz);
1742
1743             fjx2             = _mm256_add_pd(fjx2,tx);
1744             fjy2             = _mm256_add_pd(fjy2,ty);
1745             fjz2             = _mm256_add_pd(fjz2,tz);
1746
1747             /**************************
1748              * CALCULATE INTERACTIONS *
1749              **************************/
1750
1751             /* COULOMB ELECTROSTATICS */
1752             velec            = _mm256_mul_pd(qq13,rinv13);
1753             felec            = _mm256_mul_pd(velec,rinvsq13);
1754
1755             fscal            = felec;
1756
1757             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1758
1759             /* Calculate temporary vectorial force */
1760             tx               = _mm256_mul_pd(fscal,dx13);
1761             ty               = _mm256_mul_pd(fscal,dy13);
1762             tz               = _mm256_mul_pd(fscal,dz13);
1763
1764             /* Update vectorial force */
1765             fix1             = _mm256_add_pd(fix1,tx);
1766             fiy1             = _mm256_add_pd(fiy1,ty);
1767             fiz1             = _mm256_add_pd(fiz1,tz);
1768
1769             fjx3             = _mm256_add_pd(fjx3,tx);
1770             fjy3             = _mm256_add_pd(fjy3,ty);
1771             fjz3             = _mm256_add_pd(fjz3,tz);
1772
1773             /**************************
1774              * CALCULATE INTERACTIONS *
1775              **************************/
1776
1777             /* COULOMB ELECTROSTATICS */
1778             velec            = _mm256_mul_pd(qq21,rinv21);
1779             felec            = _mm256_mul_pd(velec,rinvsq21);
1780
1781             fscal            = felec;
1782
1783             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1784
1785             /* Calculate temporary vectorial force */
1786             tx               = _mm256_mul_pd(fscal,dx21);
1787             ty               = _mm256_mul_pd(fscal,dy21);
1788             tz               = _mm256_mul_pd(fscal,dz21);
1789
1790             /* Update vectorial force */
1791             fix2             = _mm256_add_pd(fix2,tx);
1792             fiy2             = _mm256_add_pd(fiy2,ty);
1793             fiz2             = _mm256_add_pd(fiz2,tz);
1794
1795             fjx1             = _mm256_add_pd(fjx1,tx);
1796             fjy1             = _mm256_add_pd(fjy1,ty);
1797             fjz1             = _mm256_add_pd(fjz1,tz);
1798
1799             /**************************
1800              * CALCULATE INTERACTIONS *
1801              **************************/
1802
1803             /* COULOMB ELECTROSTATICS */
1804             velec            = _mm256_mul_pd(qq22,rinv22);
1805             felec            = _mm256_mul_pd(velec,rinvsq22);
1806
1807             fscal            = felec;
1808
1809             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1810
1811             /* Calculate temporary vectorial force */
1812             tx               = _mm256_mul_pd(fscal,dx22);
1813             ty               = _mm256_mul_pd(fscal,dy22);
1814             tz               = _mm256_mul_pd(fscal,dz22);
1815
1816             /* Update vectorial force */
1817             fix2             = _mm256_add_pd(fix2,tx);
1818             fiy2             = _mm256_add_pd(fiy2,ty);
1819             fiz2             = _mm256_add_pd(fiz2,tz);
1820
1821             fjx2             = _mm256_add_pd(fjx2,tx);
1822             fjy2             = _mm256_add_pd(fjy2,ty);
1823             fjz2             = _mm256_add_pd(fjz2,tz);
1824
1825             /**************************
1826              * CALCULATE INTERACTIONS *
1827              **************************/
1828
1829             /* COULOMB ELECTROSTATICS */
1830             velec            = _mm256_mul_pd(qq23,rinv23);
1831             felec            = _mm256_mul_pd(velec,rinvsq23);
1832
1833             fscal            = felec;
1834
1835             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1836
1837             /* Calculate temporary vectorial force */
1838             tx               = _mm256_mul_pd(fscal,dx23);
1839             ty               = _mm256_mul_pd(fscal,dy23);
1840             tz               = _mm256_mul_pd(fscal,dz23);
1841
1842             /* Update vectorial force */
1843             fix2             = _mm256_add_pd(fix2,tx);
1844             fiy2             = _mm256_add_pd(fiy2,ty);
1845             fiz2             = _mm256_add_pd(fiz2,tz);
1846
1847             fjx3             = _mm256_add_pd(fjx3,tx);
1848             fjy3             = _mm256_add_pd(fjy3,ty);
1849             fjz3             = _mm256_add_pd(fjz3,tz);
1850
1851             /**************************
1852              * CALCULATE INTERACTIONS *
1853              **************************/
1854
1855             /* COULOMB ELECTROSTATICS */
1856             velec            = _mm256_mul_pd(qq31,rinv31);
1857             felec            = _mm256_mul_pd(velec,rinvsq31);
1858
1859             fscal            = felec;
1860
1861             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1862
1863             /* Calculate temporary vectorial force */
1864             tx               = _mm256_mul_pd(fscal,dx31);
1865             ty               = _mm256_mul_pd(fscal,dy31);
1866             tz               = _mm256_mul_pd(fscal,dz31);
1867
1868             /* Update vectorial force */
1869             fix3             = _mm256_add_pd(fix3,tx);
1870             fiy3             = _mm256_add_pd(fiy3,ty);
1871             fiz3             = _mm256_add_pd(fiz3,tz);
1872
1873             fjx1             = _mm256_add_pd(fjx1,tx);
1874             fjy1             = _mm256_add_pd(fjy1,ty);
1875             fjz1             = _mm256_add_pd(fjz1,tz);
1876
1877             /**************************
1878              * CALCULATE INTERACTIONS *
1879              **************************/
1880
1881             /* COULOMB ELECTROSTATICS */
1882             velec            = _mm256_mul_pd(qq32,rinv32);
1883             felec            = _mm256_mul_pd(velec,rinvsq32);
1884
1885             fscal            = felec;
1886
1887             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1888
1889             /* Calculate temporary vectorial force */
1890             tx               = _mm256_mul_pd(fscal,dx32);
1891             ty               = _mm256_mul_pd(fscal,dy32);
1892             tz               = _mm256_mul_pd(fscal,dz32);
1893
1894             /* Update vectorial force */
1895             fix3             = _mm256_add_pd(fix3,tx);
1896             fiy3             = _mm256_add_pd(fiy3,ty);
1897             fiz3             = _mm256_add_pd(fiz3,tz);
1898
1899             fjx2             = _mm256_add_pd(fjx2,tx);
1900             fjy2             = _mm256_add_pd(fjy2,ty);
1901             fjz2             = _mm256_add_pd(fjz2,tz);
1902
1903             /**************************
1904              * CALCULATE INTERACTIONS *
1905              **************************/
1906
1907             /* COULOMB ELECTROSTATICS */
1908             velec            = _mm256_mul_pd(qq33,rinv33);
1909             felec            = _mm256_mul_pd(velec,rinvsq33);
1910
1911             fscal            = felec;
1912
1913             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1914
1915             /* Calculate temporary vectorial force */
1916             tx               = _mm256_mul_pd(fscal,dx33);
1917             ty               = _mm256_mul_pd(fscal,dy33);
1918             tz               = _mm256_mul_pd(fscal,dz33);
1919
1920             /* Update vectorial force */
1921             fix3             = _mm256_add_pd(fix3,tx);
1922             fiy3             = _mm256_add_pd(fiy3,ty);
1923             fiz3             = _mm256_add_pd(fiz3,tz);
1924
1925             fjx3             = _mm256_add_pd(fjx3,tx);
1926             fjy3             = _mm256_add_pd(fjy3,ty);
1927             fjz3             = _mm256_add_pd(fjz3,tz);
1928
1929             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1930             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1931             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1932             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1933
1934             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1935                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1936                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1937
1938             /* Inner loop uses 264 flops */
1939         }
1940
1941         /* End of innermost loop */
1942
1943         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1944                                                  f+i_coord_offset,fshift+i_shift_offset);
1945
1946         /* Increment number of inner iterations */
1947         inneriter                  += j_index_end - j_index_start;
1948
1949         /* Outer loop uses 24 flops */
1950     }
1951
1952     /* Increment number of outer iterations */
1953     outeriter        += nri;
1954
1955     /* Update outer/inner flops */
1956
1957     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
1958 }