src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwLJ_GeomW3W3_avx_256_single.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3W3_VF_avx_256_single
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwLJ_GeomW3W3_VF_avx_256_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrE,jnrF,jnrG,jnrH;
  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              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  83     real             scratch[4*DIM];
  84     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     real *           vdwioffsetptr0;
  86     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     real *           vdwioffsetptr1;
  88     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     real *           vdwioffsetptr2;
  90     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  92     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  94     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  95     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  96     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  97     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  98     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  99     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 100     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 101     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 104     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 113     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 114     __m256           dummy_mask,cutoff_mask;
 115     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 116     __m256           one     = _mm256_set1_ps(1.0);
 117     __m256           two     = _mm256_set1_ps(2.0);
 118     x                = xx[0];
 119     f                = ff[0];
 120
 121     nri              = nlist->nri;
 122     iinr             = nlist->iinr;
 123     jindex           = nlist->jindex;
 124     jjnr             = nlist->jjnr;
 125     shiftidx         = nlist->shift;
 126     gid              = nlist->gid;
 127     shiftvec         = fr->shift_vec[0];
 128     fshift           = fr->fshift[0];
 129     facel            = _mm256_set1_ps(fr->ic->epsfac);
 130     charge           = mdatoms->chargeA;
 131     nvdwtype         = fr->ntype;
 132     vdwparam         = fr->nbfp;
 133     vdwtype          = mdatoms->typeA;
 134
 135     /* Setup water-specific parameters */
 136     inr              = nlist->iinr[0];
 137     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 138     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 139     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 140     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 141
 142     jq0              = _mm256_set1_ps(charge[inr+0]);
 143     jq1              = _mm256_set1_ps(charge[inr+1]);
 144     jq2              = _mm256_set1_ps(charge[inr+2]);
 145     vdwjidx0A        = 2*vdwtype[inr+0];
 146     qq00             = _mm256_mul_ps(iq0,jq0);
 147     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 148     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 149     qq01             = _mm256_mul_ps(iq0,jq1);
 150     qq02             = _mm256_mul_ps(iq0,jq2);
 151     qq10             = _mm256_mul_ps(iq1,jq0);
 152     qq11             = _mm256_mul_ps(iq1,jq1);
 153     qq12             = _mm256_mul_ps(iq1,jq2);
 154     qq20             = _mm256_mul_ps(iq2,jq0);
 155     qq21             = _mm256_mul_ps(iq2,jq1);
 156     qq22             = _mm256_mul_ps(iq2,jq2);
 157
 158     /* Avoid stupid compiler warnings */
 159     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 160     j_coord_offsetA = 0;
 161     j_coord_offsetB = 0;
 162     j_coord_offsetC = 0;
 163     j_coord_offsetD = 0;
 164     j_coord_offsetE = 0;
 165     j_coord_offsetF = 0;
 166     j_coord_offsetG = 0;
 167     j_coord_offsetH = 0;
 168
 169     outeriter        = 0;
 170     inneriter        = 0;
 171
 172     for(iidx=0;iidx<4*DIM;iidx++)
 173     {
 174         scratch[iidx] = 0.0;
 175     }
 176
 177     /* Start outer loop over neighborlists */
 178     for(iidx=0; iidx<nri; iidx++)
 179     {
 180         /* Load shift vector for this list */
 181         i_shift_offset   = DIM*shiftidx[iidx];
 182
 183         /* Load limits for loop over neighbors */
 184         j_index_start    = jindex[iidx];
 185         j_index_end      = jindex[iidx+1];
 186
 187         /* Get outer coordinate index */
 188         inr              = iinr[iidx];
 189         i_coord_offset   = DIM*inr;
 190
 191         /* Load i particle coords and add shift vector */
 192         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 193                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 194
 195         fix0             = _mm256_setzero_ps();
 196         fiy0             = _mm256_setzero_ps();
 197         fiz0             = _mm256_setzero_ps();
 198         fix1             = _mm256_setzero_ps();
 199         fiy1             = _mm256_setzero_ps();
 200         fiz1             = _mm256_setzero_ps();
 201         fix2             = _mm256_setzero_ps();
 202         fiy2             = _mm256_setzero_ps();
 203         fiz2             = _mm256_setzero_ps();
 204
 205         /* Reset potential sums */
 206         velecsum         = _mm256_setzero_ps();
 207         vvdwsum          = _mm256_setzero_ps();
 208
 209         /* Start inner kernel loop */
 210         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 211         {
 212
 213             /* Get j neighbor index, and coordinate index */
 214             jnrA             = jjnr[jidx];
 215             jnrB             = jjnr[jidx+1];
 216             jnrC             = jjnr[jidx+2];
 217             jnrD             = jjnr[jidx+3];
 218             jnrE             = jjnr[jidx+4];
 219             jnrF             = jjnr[jidx+5];
 220             jnrG             = jjnr[jidx+6];
 221             jnrH             = jjnr[jidx+7];
 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             j_coord_offsetE  = DIM*jnrE;
 227             j_coord_offsetF  = DIM*jnrF;
 228             j_coord_offsetG  = DIM*jnrG;
 229             j_coord_offsetH  = DIM*jnrH;
 230
 231             /* load j atom coordinates */
 232             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 233                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 234                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 235                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 236                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 237
 238             /* Calculate displacement vector */
 239             dx00             = _mm256_sub_ps(ix0,jx0);
 240             dy00             = _mm256_sub_ps(iy0,jy0);
 241             dz00             = _mm256_sub_ps(iz0,jz0);
 242             dx01             = _mm256_sub_ps(ix0,jx1);
 243             dy01             = _mm256_sub_ps(iy0,jy1);
 244             dz01             = _mm256_sub_ps(iz0,jz1);
 245             dx02             = _mm256_sub_ps(ix0,jx2);
 246             dy02             = _mm256_sub_ps(iy0,jy2);
 247             dz02             = _mm256_sub_ps(iz0,jz2);
 248             dx10             = _mm256_sub_ps(ix1,jx0);
 249             dy10             = _mm256_sub_ps(iy1,jy0);
 250             dz10             = _mm256_sub_ps(iz1,jz0);
 251             dx11             = _mm256_sub_ps(ix1,jx1);
 252             dy11             = _mm256_sub_ps(iy1,jy1);
 253             dz11             = _mm256_sub_ps(iz1,jz1);
 254             dx12             = _mm256_sub_ps(ix1,jx2);
 255             dy12             = _mm256_sub_ps(iy1,jy2);
 256             dz12             = _mm256_sub_ps(iz1,jz2);
 257             dx20             = _mm256_sub_ps(ix2,jx0);
 258             dy20             = _mm256_sub_ps(iy2,jy0);
 259             dz20             = _mm256_sub_ps(iz2,jz0);
 260             dx21             = _mm256_sub_ps(ix2,jx1);
 261             dy21             = _mm256_sub_ps(iy2,jy1);
 262             dz21             = _mm256_sub_ps(iz2,jz1);
 263             dx22             = _mm256_sub_ps(ix2,jx2);
 264             dy22             = _mm256_sub_ps(iy2,jy2);
 265             dz22             = _mm256_sub_ps(iz2,jz2);
 266
 267             /* Calculate squared distance and things based on it */
 268             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 269             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 270             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 271             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 272             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 273             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 274             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 275             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 276             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 277
 278             rinv00           = avx256_invsqrt_f(rsq00);
 279             rinv01           = avx256_invsqrt_f(rsq01);
 280             rinv02           = avx256_invsqrt_f(rsq02);
 281             rinv10           = avx256_invsqrt_f(rsq10);
 282             rinv11           = avx256_invsqrt_f(rsq11);
 283             rinv12           = avx256_invsqrt_f(rsq12);
 284             rinv20           = avx256_invsqrt_f(rsq20);
 285             rinv21           = avx256_invsqrt_f(rsq21);
 286             rinv22           = avx256_invsqrt_f(rsq22);
 287
 288             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 289             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 290             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 291             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 292             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 293             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 294             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 295             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 296             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 297
 298             fjx0             = _mm256_setzero_ps();
 299             fjy0             = _mm256_setzero_ps();
 300             fjz0             = _mm256_setzero_ps();
 301             fjx1             = _mm256_setzero_ps();
 302             fjy1             = _mm256_setzero_ps();
 303             fjz1             = _mm256_setzero_ps();
 304             fjx2             = _mm256_setzero_ps();
 305             fjy2             = _mm256_setzero_ps();
 306             fjz2             = _mm256_setzero_ps();
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             /* COULOMB ELECTROSTATICS */
 313             velec            = _mm256_mul_ps(qq00,rinv00);
 314             felec            = _mm256_mul_ps(velec,rinvsq00);
 315
 316             /* LENNARD-JONES DISPERSION/REPULSION */
 317
 318             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 319             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 320             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 321             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 322             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 323
 324             /* Update potential sum for this i atom from the interaction with this j atom. */
 325             velecsum         = _mm256_add_ps(velecsum,velec);
 326             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 327
 328             fscal            = _mm256_add_ps(felec,fvdw);
 329
 330             /* Calculate temporary vectorial force */
 331             tx               = _mm256_mul_ps(fscal,dx00);
 332             ty               = _mm256_mul_ps(fscal,dy00);
 333             tz               = _mm256_mul_ps(fscal,dz00);
 334
 335             /* Update vectorial force */
 336             fix0             = _mm256_add_ps(fix0,tx);
 337             fiy0             = _mm256_add_ps(fiy0,ty);
 338             fiz0             = _mm256_add_ps(fiz0,tz);
 339
 340             fjx0             = _mm256_add_ps(fjx0,tx);
 341             fjy0             = _mm256_add_ps(fjy0,ty);
 342             fjz0             = _mm256_add_ps(fjz0,tz);
 343
 344             /**************************
 345              * CALCULATE INTERACTIONS *
 346              **************************/
 347
 348             /* COULOMB ELECTROSTATICS */
 349             velec            = _mm256_mul_ps(qq01,rinv01);
 350             felec            = _mm256_mul_ps(velec,rinvsq01);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             velecsum         = _mm256_add_ps(velecsum,velec);
 354
 355             fscal            = felec;
 356
 357             /* Calculate temporary vectorial force */
 358             tx               = _mm256_mul_ps(fscal,dx01);
 359             ty               = _mm256_mul_ps(fscal,dy01);
 360             tz               = _mm256_mul_ps(fscal,dz01);
 361
 362             /* Update vectorial force */
 363             fix0             = _mm256_add_ps(fix0,tx);
 364             fiy0             = _mm256_add_ps(fiy0,ty);
 365             fiz0             = _mm256_add_ps(fiz0,tz);
 366
 367             fjx1             = _mm256_add_ps(fjx1,tx);
 368             fjy1             = _mm256_add_ps(fjy1,ty);
 369             fjz1             = _mm256_add_ps(fjz1,tz);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             /* COULOMB ELECTROSTATICS */
 376             velec            = _mm256_mul_ps(qq02,rinv02);
 377             felec            = _mm256_mul_ps(velec,rinvsq02);
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             velecsum         = _mm256_add_ps(velecsum,velec);
 381
 382             fscal            = felec;
 383
 384             /* Calculate temporary vectorial force */
 385             tx               = _mm256_mul_ps(fscal,dx02);
 386             ty               = _mm256_mul_ps(fscal,dy02);
 387             tz               = _mm256_mul_ps(fscal,dz02);
 388
 389             /* Update vectorial force */
 390             fix0             = _mm256_add_ps(fix0,tx);
 391             fiy0             = _mm256_add_ps(fiy0,ty);
 392             fiz0             = _mm256_add_ps(fiz0,tz);
 393
 394             fjx2             = _mm256_add_ps(fjx2,tx);
 395             fjy2             = _mm256_add_ps(fjy2,ty);
 396             fjz2             = _mm256_add_ps(fjz2,tz);
 397
 398             /**************************
 399              * CALCULATE INTERACTIONS *
 400              **************************/
 401
 402             /* COULOMB ELECTROSTATICS */
 403             velec            = _mm256_mul_ps(qq10,rinv10);
 404             felec            = _mm256_mul_ps(velec,rinvsq10);
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm256_add_ps(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm256_mul_ps(fscal,dx10);
 413             ty               = _mm256_mul_ps(fscal,dy10);
 414             tz               = _mm256_mul_ps(fscal,dz10);
 415
 416             /* Update vectorial force */
 417             fix1             = _mm256_add_ps(fix1,tx);
 418             fiy1             = _mm256_add_ps(fiy1,ty);
 419             fiz1             = _mm256_add_ps(fiz1,tz);
 420
 421             fjx0             = _mm256_add_ps(fjx0,tx);
 422             fjy0             = _mm256_add_ps(fjy0,ty);
 423             fjz0             = _mm256_add_ps(fjz0,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             /* COULOMB ELECTROSTATICS */
 430             velec            = _mm256_mul_ps(qq11,rinv11);
 431             felec            = _mm256_mul_ps(velec,rinvsq11);
 432
 433             /* Update potential sum for this i atom from the interaction with this j atom. */
 434             velecsum         = _mm256_add_ps(velecsum,velec);
 435
 436             fscal            = felec;
 437
 438             /* Calculate temporary vectorial force */
 439             tx               = _mm256_mul_ps(fscal,dx11);
 440             ty               = _mm256_mul_ps(fscal,dy11);
 441             tz               = _mm256_mul_ps(fscal,dz11);
 442
 443             /* Update vectorial force */
 444             fix1             = _mm256_add_ps(fix1,tx);
 445             fiy1             = _mm256_add_ps(fiy1,ty);
 446             fiz1             = _mm256_add_ps(fiz1,tz);
 447
 448             fjx1             = _mm256_add_ps(fjx1,tx);
 449             fjy1             = _mm256_add_ps(fjy1,ty);
 450             fjz1             = _mm256_add_ps(fjz1,tz);
 451
 452             /**************************
 453              * CALCULATE INTERACTIONS *
 454              **************************/
 455
 456             /* COULOMB ELECTROSTATICS */
 457             velec            = _mm256_mul_ps(qq12,rinv12);
 458             felec            = _mm256_mul_ps(velec,rinvsq12);
 459
 460             /* Update potential sum for this i atom from the interaction with this j atom. */
 461             velecsum         = _mm256_add_ps(velecsum,velec);
 462
 463             fscal            = felec;
 464
 465             /* Calculate temporary vectorial force */
 466             tx               = _mm256_mul_ps(fscal,dx12);
 467             ty               = _mm256_mul_ps(fscal,dy12);
 468             tz               = _mm256_mul_ps(fscal,dz12);
 469
 470             /* Update vectorial force */
 471             fix1             = _mm256_add_ps(fix1,tx);
 472             fiy1             = _mm256_add_ps(fiy1,ty);
 473             fiz1             = _mm256_add_ps(fiz1,tz);
 474
 475             fjx2             = _mm256_add_ps(fjx2,tx);
 476             fjy2             = _mm256_add_ps(fjy2,ty);
 477             fjz2             = _mm256_add_ps(fjz2,tz);
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             /* COULOMB ELECTROSTATICS */
 484             velec            = _mm256_mul_ps(qq20,rinv20);
 485             felec            = _mm256_mul_ps(velec,rinvsq20);
 486
 487             /* Update potential sum for this i atom from the interaction with this j atom. */
 488             velecsum         = _mm256_add_ps(velecsum,velec);
 489
 490             fscal            = felec;
 491
 492             /* Calculate temporary vectorial force */
 493             tx               = _mm256_mul_ps(fscal,dx20);
 494             ty               = _mm256_mul_ps(fscal,dy20);
 495             tz               = _mm256_mul_ps(fscal,dz20);
 496
 497             /* Update vectorial force */
 498             fix2             = _mm256_add_ps(fix2,tx);
 499             fiy2             = _mm256_add_ps(fiy2,ty);
 500             fiz2             = _mm256_add_ps(fiz2,tz);
 501
 502             fjx0             = _mm256_add_ps(fjx0,tx);
 503             fjy0             = _mm256_add_ps(fjy0,ty);
 504             fjz0             = _mm256_add_ps(fjz0,tz);
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             /* COULOMB ELECTROSTATICS */
 511             velec            = _mm256_mul_ps(qq21,rinv21);
 512             felec            = _mm256_mul_ps(velec,rinvsq21);
 513
 514             /* Update potential sum for this i atom from the interaction with this j atom. */
 515             velecsum         = _mm256_add_ps(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519             /* Calculate temporary vectorial force */
 520             tx               = _mm256_mul_ps(fscal,dx21);
 521             ty               = _mm256_mul_ps(fscal,dy21);
 522             tz               = _mm256_mul_ps(fscal,dz21);
 523
 524             /* Update vectorial force */
 525             fix2             = _mm256_add_ps(fix2,tx);
 526             fiy2             = _mm256_add_ps(fiy2,ty);
 527             fiz2             = _mm256_add_ps(fiz2,tz);
 528
 529             fjx1             = _mm256_add_ps(fjx1,tx);
 530             fjy1             = _mm256_add_ps(fjy1,ty);
 531             fjz1             = _mm256_add_ps(fjz1,tz);
 532
 533             /**************************
 534              * CALCULATE INTERACTIONS *
 535              **************************/
 536
 537             /* COULOMB ELECTROSTATICS */
 538             velec            = _mm256_mul_ps(qq22,rinv22);
 539             felec            = _mm256_mul_ps(velec,rinvsq22);
 540
 541             /* Update potential sum for this i atom from the interaction with this j atom. */
 542             velecsum         = _mm256_add_ps(velecsum,velec);
 543
 544             fscal            = felec;
 545
 546             /* Calculate temporary vectorial force */
 547             tx               = _mm256_mul_ps(fscal,dx22);
 548             ty               = _mm256_mul_ps(fscal,dy22);
 549             tz               = _mm256_mul_ps(fscal,dz22);
 550
 551             /* Update vectorial force */
 552             fix2             = _mm256_add_ps(fix2,tx);
 553             fiy2             = _mm256_add_ps(fiy2,ty);
 554             fiz2             = _mm256_add_ps(fiz2,tz);
 555
 556             fjx2             = _mm256_add_ps(fjx2,tx);
 557             fjy2             = _mm256_add_ps(fjy2,ty);
 558             fjz2             = _mm256_add_ps(fjz2,tz);
 559
 560             fjptrA             = f+j_coord_offsetA;
 561             fjptrB             = f+j_coord_offsetB;
 562             fjptrC             = f+j_coord_offsetC;
 563             fjptrD             = f+j_coord_offsetD;
 564             fjptrE             = f+j_coord_offsetE;
 565             fjptrF             = f+j_coord_offsetF;
 566             fjptrG             = f+j_coord_offsetG;
 567             fjptrH             = f+j_coord_offsetH;
 568
 569             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 570                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 571
 572             /* Inner loop uses 255 flops */
 573         }
 574
 575         if(jidx<j_index_end)
 576         {
 577
 578             /* Get j neighbor index, and coordinate index */
 579             jnrlistA         = jjnr[jidx];
 580             jnrlistB         = jjnr[jidx+1];
 581             jnrlistC         = jjnr[jidx+2];
 582             jnrlistD         = jjnr[jidx+3];
 583             jnrlistE         = jjnr[jidx+4];
 584             jnrlistF         = jjnr[jidx+5];
 585             jnrlistG         = jjnr[jidx+6];
 586             jnrlistH         = jjnr[jidx+7];
 587             /* Sign of each element will be negative for non-real atoms.
 588              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 589              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 590              */
 591             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 592                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 593
 594             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 595             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 596             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 597             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 598             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 599             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 600             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 601             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 602             j_coord_offsetA  = DIM*jnrA;
 603             j_coord_offsetB  = DIM*jnrB;
 604             j_coord_offsetC  = DIM*jnrC;
 605             j_coord_offsetD  = DIM*jnrD;
 606             j_coord_offsetE  = DIM*jnrE;
 607             j_coord_offsetF  = DIM*jnrF;
 608             j_coord_offsetG  = DIM*jnrG;
 609             j_coord_offsetH  = DIM*jnrH;
 610
 611             /* load j atom coordinates */
 612             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 613                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 614                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 615                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 616                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 617
 618             /* Calculate displacement vector */
 619             dx00             = _mm256_sub_ps(ix0,jx0);
 620             dy00             = _mm256_sub_ps(iy0,jy0);
 621             dz00             = _mm256_sub_ps(iz0,jz0);
 622             dx01             = _mm256_sub_ps(ix0,jx1);
 623             dy01             = _mm256_sub_ps(iy0,jy1);
 624             dz01             = _mm256_sub_ps(iz0,jz1);
 625             dx02             = _mm256_sub_ps(ix0,jx2);
 626             dy02             = _mm256_sub_ps(iy0,jy2);
 627             dz02             = _mm256_sub_ps(iz0,jz2);
 628             dx10             = _mm256_sub_ps(ix1,jx0);
 629             dy10             = _mm256_sub_ps(iy1,jy0);
 630             dz10             = _mm256_sub_ps(iz1,jz0);
 631             dx11             = _mm256_sub_ps(ix1,jx1);
 632             dy11             = _mm256_sub_ps(iy1,jy1);
 633             dz11             = _mm256_sub_ps(iz1,jz1);
 634             dx12             = _mm256_sub_ps(ix1,jx2);
 635             dy12             = _mm256_sub_ps(iy1,jy2);
 636             dz12             = _mm256_sub_ps(iz1,jz2);
 637             dx20             = _mm256_sub_ps(ix2,jx0);
 638             dy20             = _mm256_sub_ps(iy2,jy0);
 639             dz20             = _mm256_sub_ps(iz2,jz0);
 640             dx21             = _mm256_sub_ps(ix2,jx1);
 641             dy21             = _mm256_sub_ps(iy2,jy1);
 642             dz21             = _mm256_sub_ps(iz2,jz1);
 643             dx22             = _mm256_sub_ps(ix2,jx2);
 644             dy22             = _mm256_sub_ps(iy2,jy2);
 645             dz22             = _mm256_sub_ps(iz2,jz2);
 646
 647             /* Calculate squared distance and things based on it */
 648             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 649             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 650             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 651             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 652             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 653             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 654             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 655             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 656             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 657
 658             rinv00           = avx256_invsqrt_f(rsq00);
 659             rinv01           = avx256_invsqrt_f(rsq01);
 660             rinv02           = avx256_invsqrt_f(rsq02);
 661             rinv10           = avx256_invsqrt_f(rsq10);
 662             rinv11           = avx256_invsqrt_f(rsq11);
 663             rinv12           = avx256_invsqrt_f(rsq12);
 664             rinv20           = avx256_invsqrt_f(rsq20);
 665             rinv21           = avx256_invsqrt_f(rsq21);
 666             rinv22           = avx256_invsqrt_f(rsq22);
 667
 668             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 669             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 670             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 671             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 672             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 673             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 674             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 675             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 676             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 677
 678             fjx0             = _mm256_setzero_ps();
 679             fjy0             = _mm256_setzero_ps();
 680             fjz0             = _mm256_setzero_ps();
 681             fjx1             = _mm256_setzero_ps();
 682             fjy1             = _mm256_setzero_ps();
 683             fjz1             = _mm256_setzero_ps();
 684             fjx2             = _mm256_setzero_ps();
 685             fjy2             = _mm256_setzero_ps();
 686             fjz2             = _mm256_setzero_ps();
 687
 688             /**************************
 689              * CALCULATE INTERACTIONS *
 690              **************************/
 691
 692             /* COULOMB ELECTROSTATICS */
 693             velec            = _mm256_mul_ps(qq00,rinv00);
 694             felec            = _mm256_mul_ps(velec,rinvsq00);
 695
 696             /* LENNARD-JONES DISPERSION/REPULSION */
 697
 698             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 699             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 700             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 701             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 702             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 703
 704             /* Update potential sum for this i atom from the interaction with this j atom. */
 705             velec            = _mm256_andnot_ps(dummy_mask,velec);
 706             velecsum         = _mm256_add_ps(velecsum,velec);
 707             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 708             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 709
 710             fscal            = _mm256_add_ps(felec,fvdw);
 711
 712             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 713
 714             /* Calculate temporary vectorial force */
 715             tx               = _mm256_mul_ps(fscal,dx00);
 716             ty               = _mm256_mul_ps(fscal,dy00);
 717             tz               = _mm256_mul_ps(fscal,dz00);
 718
 719             /* Update vectorial force */
 720             fix0             = _mm256_add_ps(fix0,tx);
 721             fiy0             = _mm256_add_ps(fiy0,ty);
 722             fiz0             = _mm256_add_ps(fiz0,tz);
 723
 724             fjx0             = _mm256_add_ps(fjx0,tx);
 725             fjy0             = _mm256_add_ps(fjy0,ty);
 726             fjz0             = _mm256_add_ps(fjz0,tz);
 727
 728             /**************************
 729              * CALCULATE INTERACTIONS *
 730              **************************/
 731
 732             /* COULOMB ELECTROSTATICS */
 733             velec            = _mm256_mul_ps(qq01,rinv01);
 734             felec            = _mm256_mul_ps(velec,rinvsq01);
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             velec            = _mm256_andnot_ps(dummy_mask,velec);
 738             velecsum         = _mm256_add_ps(velecsum,velec);
 739
 740             fscal            = felec;
 741
 742             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 743
 744             /* Calculate temporary vectorial force */
 745             tx               = _mm256_mul_ps(fscal,dx01);
 746             ty               = _mm256_mul_ps(fscal,dy01);
 747             tz               = _mm256_mul_ps(fscal,dz01);
 748
 749             /* Update vectorial force */
 750             fix0             = _mm256_add_ps(fix0,tx);
 751             fiy0             = _mm256_add_ps(fiy0,ty);
 752             fiz0             = _mm256_add_ps(fiz0,tz);
 753
 754             fjx1             = _mm256_add_ps(fjx1,tx);
 755             fjy1             = _mm256_add_ps(fjy1,ty);
 756             fjz1             = _mm256_add_ps(fjz1,tz);
 757
 758             /**************************
 759              * CALCULATE INTERACTIONS *
 760              **************************/
 761
 762             /* COULOMB ELECTROSTATICS */
 763             velec            = _mm256_mul_ps(qq02,rinv02);
 764             felec            = _mm256_mul_ps(velec,rinvsq02);
 765
 766             /* Update potential sum for this i atom from the interaction with this j atom. */
 767             velec            = _mm256_andnot_ps(dummy_mask,velec);
 768             velecsum         = _mm256_add_ps(velecsum,velec);
 769
 770             fscal            = felec;
 771
 772             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm256_mul_ps(fscal,dx02);
 776             ty               = _mm256_mul_ps(fscal,dy02);
 777             tz               = _mm256_mul_ps(fscal,dz02);
 778
 779             /* Update vectorial force */
 780             fix0             = _mm256_add_ps(fix0,tx);
 781             fiy0             = _mm256_add_ps(fiy0,ty);
 782             fiz0             = _mm256_add_ps(fiz0,tz);
 783
 784             fjx2             = _mm256_add_ps(fjx2,tx);
 785             fjy2             = _mm256_add_ps(fjy2,ty);
 786             fjz2             = _mm256_add_ps(fjz2,tz);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm256_mul_ps(qq10,rinv10);
 794             felec            = _mm256_mul_ps(velec,rinvsq10);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm256_andnot_ps(dummy_mask,velec);
 798             velecsum         = _mm256_add_ps(velecsum,velec);
 799
 800             fscal            = felec;
 801
 802             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 803
 804             /* Calculate temporary vectorial force */
 805             tx               = _mm256_mul_ps(fscal,dx10);
 806             ty               = _mm256_mul_ps(fscal,dy10);
 807             tz               = _mm256_mul_ps(fscal,dz10);
 808
 809             /* Update vectorial force */
 810             fix1             = _mm256_add_ps(fix1,tx);
 811             fiy1             = _mm256_add_ps(fiy1,ty);
 812             fiz1             = _mm256_add_ps(fiz1,tz);
 813
 814             fjx0             = _mm256_add_ps(fjx0,tx);
 815             fjy0             = _mm256_add_ps(fjy0,ty);
 816             fjz0             = _mm256_add_ps(fjz0,tz);
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             /* COULOMB ELECTROSTATICS */
 823             velec            = _mm256_mul_ps(qq11,rinv11);
 824             felec            = _mm256_mul_ps(velec,rinvsq11);
 825
 826             /* Update potential sum for this i atom from the interaction with this j atom. */
 827             velec            = _mm256_andnot_ps(dummy_mask,velec);
 828             velecsum         = _mm256_add_ps(velecsum,velec);
 829
 830             fscal            = felec;
 831
 832             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 833
 834             /* Calculate temporary vectorial force */
 835             tx               = _mm256_mul_ps(fscal,dx11);
 836             ty               = _mm256_mul_ps(fscal,dy11);
 837             tz               = _mm256_mul_ps(fscal,dz11);
 838
 839             /* Update vectorial force */
 840             fix1             = _mm256_add_ps(fix1,tx);
 841             fiy1             = _mm256_add_ps(fiy1,ty);
 842             fiz1             = _mm256_add_ps(fiz1,tz);
 843
 844             fjx1             = _mm256_add_ps(fjx1,tx);
 845             fjy1             = _mm256_add_ps(fjy1,ty);
 846             fjz1             = _mm256_add_ps(fjz1,tz);
 847
 848             /**************************
 849              * CALCULATE INTERACTIONS *
 850              **************************/
 851
 852             /* COULOMB ELECTROSTATICS */
 853             velec            = _mm256_mul_ps(qq12,rinv12);
 854             felec            = _mm256_mul_ps(velec,rinvsq12);
 855
 856             /* Update potential sum for this i atom from the interaction with this j atom. */
 857             velec            = _mm256_andnot_ps(dummy_mask,velec);
 858             velecsum         = _mm256_add_ps(velecsum,velec);
 859
 860             fscal            = felec;
 861
 862             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 863
 864             /* Calculate temporary vectorial force */
 865             tx               = _mm256_mul_ps(fscal,dx12);
 866             ty               = _mm256_mul_ps(fscal,dy12);
 867             tz               = _mm256_mul_ps(fscal,dz12);
 868
 869             /* Update vectorial force */
 870             fix1             = _mm256_add_ps(fix1,tx);
 871             fiy1             = _mm256_add_ps(fiy1,ty);
 872             fiz1             = _mm256_add_ps(fiz1,tz);
 873
 874             fjx2             = _mm256_add_ps(fjx2,tx);
 875             fjy2             = _mm256_add_ps(fjy2,ty);
 876             fjz2             = _mm256_add_ps(fjz2,tz);
 877
 878             /**************************
 879              * CALCULATE INTERACTIONS *
 880              **************************/
 881
 882             /* COULOMB ELECTROSTATICS */
 883             velec            = _mm256_mul_ps(qq20,rinv20);
 884             felec            = _mm256_mul_ps(velec,rinvsq20);
 885
 886             /* Update potential sum for this i atom from the interaction with this j atom. */
 887             velec            = _mm256_andnot_ps(dummy_mask,velec);
 888             velecsum         = _mm256_add_ps(velecsum,velec);
 889
 890             fscal            = felec;
 891
 892             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 893
 894             /* Calculate temporary vectorial force */
 895             tx               = _mm256_mul_ps(fscal,dx20);
 896             ty               = _mm256_mul_ps(fscal,dy20);
 897             tz               = _mm256_mul_ps(fscal,dz20);
 898
 899             /* Update vectorial force */
 900             fix2             = _mm256_add_ps(fix2,tx);
 901             fiy2             = _mm256_add_ps(fiy2,ty);
 902             fiz2             = _mm256_add_ps(fiz2,tz);
 903
 904             fjx0             = _mm256_add_ps(fjx0,tx);
 905             fjy0             = _mm256_add_ps(fjy0,ty);
 906             fjz0             = _mm256_add_ps(fjz0,tz);
 907
 908             /**************************
 909              * CALCULATE INTERACTIONS *
 910              **************************/
 911
 912             /* COULOMB ELECTROSTATICS */
 913             velec            = _mm256_mul_ps(qq21,rinv21);
 914             felec            = _mm256_mul_ps(velec,rinvsq21);
 915
 916             /* Update potential sum for this i atom from the interaction with this j atom. */
 917             velec            = _mm256_andnot_ps(dummy_mask,velec);
 918             velecsum         = _mm256_add_ps(velecsum,velec);
 919
 920             fscal            = felec;
 921
 922             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 923
 924             /* Calculate temporary vectorial force */
 925             tx               = _mm256_mul_ps(fscal,dx21);
 926             ty               = _mm256_mul_ps(fscal,dy21);
 927             tz               = _mm256_mul_ps(fscal,dz21);
 928
 929             /* Update vectorial force */
 930             fix2             = _mm256_add_ps(fix2,tx);
 931             fiy2             = _mm256_add_ps(fiy2,ty);
 932             fiz2             = _mm256_add_ps(fiz2,tz);
 933
 934             fjx1             = _mm256_add_ps(fjx1,tx);
 935             fjy1             = _mm256_add_ps(fjy1,ty);
 936             fjz1             = _mm256_add_ps(fjz1,tz);
 937
 938             /**************************
 939              * CALCULATE INTERACTIONS *
 940              **************************/
 941
 942             /* COULOMB ELECTROSTATICS */
 943             velec            = _mm256_mul_ps(qq22,rinv22);
 944             felec            = _mm256_mul_ps(velec,rinvsq22);
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm256_andnot_ps(dummy_mask,velec);
 948             velecsum         = _mm256_add_ps(velecsum,velec);
 949
 950             fscal            = felec;
 951
 952             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 953
 954             /* Calculate temporary vectorial force */
 955             tx               = _mm256_mul_ps(fscal,dx22);
 956             ty               = _mm256_mul_ps(fscal,dy22);
 957             tz               = _mm256_mul_ps(fscal,dz22);
 958
 959             /* Update vectorial force */
 960             fix2             = _mm256_add_ps(fix2,tx);
 961             fiy2             = _mm256_add_ps(fiy2,ty);
 962             fiz2             = _mm256_add_ps(fiz2,tz);
 963
 964             fjx2             = _mm256_add_ps(fjx2,tx);
 965             fjy2             = _mm256_add_ps(fjy2,ty);
 966             fjz2             = _mm256_add_ps(fjz2,tz);
 967
 968             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 969             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 970             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 971             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 972             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 973             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 974             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 975             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 976
 977             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 978                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 979
 980             /* Inner loop uses 255 flops */
 981         }
 982
 983         /* End of innermost loop */
 984
 985         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 986                                                  f+i_coord_offset,fshift+i_shift_offset);
 987
 988         ggid                        = gid[iidx];
 989         /* Update potential energies */
 990         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 991         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 992
 993         /* Increment number of inner iterations */
 994         inneriter                  += j_index_end - j_index_start;
 995
 996         /* Outer loop uses 20 flops */
 997     }
 998
 999     /* Increment number of outer iterations */
1000     outeriter        += nri;
1001
1002     /* Update outer/inner flops */
1003
1004     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*255);
1005 }
1006 /*
1007  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3W3_F_avx_256_single
1008  * Electrostatics interaction: Coulomb
1009  * VdW interaction:            LennardJones
1010  * Geometry:                   Water3-Water3
1011  * Calculate force/pot:        Force
1012  */
1013 void
1014 nb_kernel_ElecCoul_VdwLJ_GeomW3W3_F_avx_256_single
1015                     (t_nblist                    * gmx_restrict       nlist,
1016                      rvec                        * gmx_restrict          xx,
1017                      rvec                        * gmx_restrict          ff,
1018                      struct t_forcerec           * gmx_restrict          fr,
1019                      t_mdatoms                   * gmx_restrict     mdatoms,
1020                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1021                      t_nrnb                      * gmx_restrict        nrnb)
1022 {
1023     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1024      * just 0 for non-waters.
1025      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1026      * jnr indices corresponding to data put in the four positions in the SIMD register.
1027      */
1028     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1029     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1030     int              jnrA,jnrB,jnrC,jnrD;
1031     int              jnrE,jnrF,jnrG,jnrH;
1032     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1033     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1034     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1035     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1036     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1037     real             rcutoff_scalar;
1038     real             *shiftvec,*fshift,*x,*f;
1039     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1040     real             scratch[4*DIM];
1041     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1042     real *           vdwioffsetptr0;
1043     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1044     real *           vdwioffsetptr1;
1045     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1046     real *           vdwioffsetptr2;
1047     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1048     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1049     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1050     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1051     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1052     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1053     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1054     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1055     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1056     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1057     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1058     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1059     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1060     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1061     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1062     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1063     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1064     real             *charge;
1065     int              nvdwtype;
1066     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1067     int              *vdwtype;
1068     real             *vdwparam;
1069     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1070     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1071     __m256           dummy_mask,cutoff_mask;
1072     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1073     __m256           one     = _mm256_set1_ps(1.0);
1074     __m256           two     = _mm256_set1_ps(2.0);
1075     x                = xx[0];
1076     f                = ff[0];
1077
1078     nri              = nlist->nri;
1079     iinr             = nlist->iinr;
1080     jindex           = nlist->jindex;
1081     jjnr             = nlist->jjnr;
1082     shiftidx         = nlist->shift;
1083     gid              = nlist->gid;
1084     shiftvec         = fr->shift_vec[0];
1085     fshift           = fr->fshift[0];
1086     facel            = _mm256_set1_ps(fr->ic->epsfac);
1087     charge           = mdatoms->chargeA;
1088     nvdwtype         = fr->ntype;
1089     vdwparam         = fr->nbfp;
1090     vdwtype          = mdatoms->typeA;
1091
1092     /* Setup water-specific parameters */
1093     inr              = nlist->iinr[0];
1094     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1095     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1096     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1097     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1098
1099     jq0              = _mm256_set1_ps(charge[inr+0]);
1100     jq1              = _mm256_set1_ps(charge[inr+1]);
1101     jq2              = _mm256_set1_ps(charge[inr+2]);
1102     vdwjidx0A        = 2*vdwtype[inr+0];
1103     qq00             = _mm256_mul_ps(iq0,jq0);
1104     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1105     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1106     qq01             = _mm256_mul_ps(iq0,jq1);
1107     qq02             = _mm256_mul_ps(iq0,jq2);
1108     qq10             = _mm256_mul_ps(iq1,jq0);
1109     qq11             = _mm256_mul_ps(iq1,jq1);
1110     qq12             = _mm256_mul_ps(iq1,jq2);
1111     qq20             = _mm256_mul_ps(iq2,jq0);
1112     qq21             = _mm256_mul_ps(iq2,jq1);
1113     qq22             = _mm256_mul_ps(iq2,jq2);
1114
1115     /* Avoid stupid compiler warnings */
1116     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1117     j_coord_offsetA = 0;
1118     j_coord_offsetB = 0;
1119     j_coord_offsetC = 0;
1120     j_coord_offsetD = 0;
1121     j_coord_offsetE = 0;
1122     j_coord_offsetF = 0;
1123     j_coord_offsetG = 0;
1124     j_coord_offsetH = 0;
1125
1126     outeriter        = 0;
1127     inneriter        = 0;
1128
1129     for(iidx=0;iidx<4*DIM;iidx++)
1130     {
1131         scratch[iidx] = 0.0;
1132     }
1133
1134     /* Start outer loop over neighborlists */
1135     for(iidx=0; iidx<nri; iidx++)
1136     {
1137         /* Load shift vector for this list */
1138         i_shift_offset   = DIM*shiftidx[iidx];
1139
1140         /* Load limits for loop over neighbors */
1141         j_index_start    = jindex[iidx];
1142         j_index_end      = jindex[iidx+1];
1143
1144         /* Get outer coordinate index */
1145         inr              = iinr[iidx];
1146         i_coord_offset   = DIM*inr;
1147
1148         /* Load i particle coords and add shift vector */
1149         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1150                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1151
1152         fix0             = _mm256_setzero_ps();
1153         fiy0             = _mm256_setzero_ps();
1154         fiz0             = _mm256_setzero_ps();
1155         fix1             = _mm256_setzero_ps();
1156         fiy1             = _mm256_setzero_ps();
1157         fiz1             = _mm256_setzero_ps();
1158         fix2             = _mm256_setzero_ps();
1159         fiy2             = _mm256_setzero_ps();
1160         fiz2             = _mm256_setzero_ps();
1161
1162         /* Start inner kernel loop */
1163         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1164         {
1165
1166             /* Get j neighbor index, and coordinate index */
1167             jnrA             = jjnr[jidx];
1168             jnrB             = jjnr[jidx+1];
1169             jnrC             = jjnr[jidx+2];
1170             jnrD             = jjnr[jidx+3];
1171             jnrE             = jjnr[jidx+4];
1172             jnrF             = jjnr[jidx+5];
1173             jnrG             = jjnr[jidx+6];
1174             jnrH             = jjnr[jidx+7];
1175             j_coord_offsetA  = DIM*jnrA;
1176             j_coord_offsetB  = DIM*jnrB;
1177             j_coord_offsetC  = DIM*jnrC;
1178             j_coord_offsetD  = DIM*jnrD;
1179             j_coord_offsetE  = DIM*jnrE;
1180             j_coord_offsetF  = DIM*jnrF;
1181             j_coord_offsetG  = DIM*jnrG;
1182             j_coord_offsetH  = DIM*jnrH;
1183
1184             /* load j atom coordinates */
1185             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1186                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1187                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1188                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1189                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1190
1191             /* Calculate displacement vector */
1192             dx00             = _mm256_sub_ps(ix0,jx0);
1193             dy00             = _mm256_sub_ps(iy0,jy0);
1194             dz00             = _mm256_sub_ps(iz0,jz0);
1195             dx01             = _mm256_sub_ps(ix0,jx1);
1196             dy01             = _mm256_sub_ps(iy0,jy1);
1197             dz01             = _mm256_sub_ps(iz0,jz1);
1198             dx02             = _mm256_sub_ps(ix0,jx2);
1199             dy02             = _mm256_sub_ps(iy0,jy2);
1200             dz02             = _mm256_sub_ps(iz0,jz2);
1201             dx10             = _mm256_sub_ps(ix1,jx0);
1202             dy10             = _mm256_sub_ps(iy1,jy0);
1203             dz10             = _mm256_sub_ps(iz1,jz0);
1204             dx11             = _mm256_sub_ps(ix1,jx1);
1205             dy11             = _mm256_sub_ps(iy1,jy1);
1206             dz11             = _mm256_sub_ps(iz1,jz1);
1207             dx12             = _mm256_sub_ps(ix1,jx2);
1208             dy12             = _mm256_sub_ps(iy1,jy2);
1209             dz12             = _mm256_sub_ps(iz1,jz2);
1210             dx20             = _mm256_sub_ps(ix2,jx0);
1211             dy20             = _mm256_sub_ps(iy2,jy0);
1212             dz20             = _mm256_sub_ps(iz2,jz0);
1213             dx21             = _mm256_sub_ps(ix2,jx1);
1214             dy21             = _mm256_sub_ps(iy2,jy1);
1215             dz21             = _mm256_sub_ps(iz2,jz1);
1216             dx22             = _mm256_sub_ps(ix2,jx2);
1217             dy22             = _mm256_sub_ps(iy2,jy2);
1218             dz22             = _mm256_sub_ps(iz2,jz2);
1219
1220             /* Calculate squared distance and things based on it */
1221             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1222             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1223             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1224             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1225             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1226             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1227             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1228             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1229             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1230
1231             rinv00           = avx256_invsqrt_f(rsq00);
1232             rinv01           = avx256_invsqrt_f(rsq01);
1233             rinv02           = avx256_invsqrt_f(rsq02);
1234             rinv10           = avx256_invsqrt_f(rsq10);
1235             rinv11           = avx256_invsqrt_f(rsq11);
1236             rinv12           = avx256_invsqrt_f(rsq12);
1237             rinv20           = avx256_invsqrt_f(rsq20);
1238             rinv21           = avx256_invsqrt_f(rsq21);
1239             rinv22           = avx256_invsqrt_f(rsq22);
1240
1241             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1242             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1243             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1244             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1245             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1246             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1247             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1248             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1249             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1250
1251             fjx0             = _mm256_setzero_ps();
1252             fjy0             = _mm256_setzero_ps();
1253             fjz0             = _mm256_setzero_ps();
1254             fjx1             = _mm256_setzero_ps();
1255             fjy1             = _mm256_setzero_ps();
1256             fjz1             = _mm256_setzero_ps();
1257             fjx2             = _mm256_setzero_ps();
1258             fjy2             = _mm256_setzero_ps();
1259             fjz2             = _mm256_setzero_ps();
1260
1261             /**************************
1262              * CALCULATE INTERACTIONS *
1263              **************************/
1264
1265             /* COULOMB ELECTROSTATICS */
1266             velec            = _mm256_mul_ps(qq00,rinv00);
1267             felec            = _mm256_mul_ps(velec,rinvsq00);
1268
1269             /* LENNARD-JONES DISPERSION/REPULSION */
1270
1271             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1272             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1273
1274             fscal            = _mm256_add_ps(felec,fvdw);
1275
1276             /* Calculate temporary vectorial force */
1277             tx               = _mm256_mul_ps(fscal,dx00);
1278             ty               = _mm256_mul_ps(fscal,dy00);
1279             tz               = _mm256_mul_ps(fscal,dz00);
1280
1281             /* Update vectorial force */
1282             fix0             = _mm256_add_ps(fix0,tx);
1283             fiy0             = _mm256_add_ps(fiy0,ty);
1284             fiz0             = _mm256_add_ps(fiz0,tz);
1285
1286             fjx0             = _mm256_add_ps(fjx0,tx);
1287             fjy0             = _mm256_add_ps(fjy0,ty);
1288             fjz0             = _mm256_add_ps(fjz0,tz);
1289
1290             /**************************
1291              * CALCULATE INTERACTIONS *
1292              **************************/
1293
1294             /* COULOMB ELECTROSTATICS */
1295             velec            = _mm256_mul_ps(qq01,rinv01);
1296             felec            = _mm256_mul_ps(velec,rinvsq01);
1297
1298             fscal            = felec;
1299
1300             /* Calculate temporary vectorial force */
1301             tx               = _mm256_mul_ps(fscal,dx01);
1302             ty               = _mm256_mul_ps(fscal,dy01);
1303             tz               = _mm256_mul_ps(fscal,dz01);
1304
1305             /* Update vectorial force */
1306             fix0             = _mm256_add_ps(fix0,tx);
1307             fiy0             = _mm256_add_ps(fiy0,ty);
1308             fiz0             = _mm256_add_ps(fiz0,tz);
1309
1310             fjx1             = _mm256_add_ps(fjx1,tx);
1311             fjy1             = _mm256_add_ps(fjy1,ty);
1312             fjz1             = _mm256_add_ps(fjz1,tz);
1313
1314             /**************************
1315              * CALCULATE INTERACTIONS *
1316              **************************/
1317
1318             /* COULOMB ELECTROSTATICS */
1319             velec            = _mm256_mul_ps(qq02,rinv02);
1320             felec            = _mm256_mul_ps(velec,rinvsq02);
1321
1322             fscal            = felec;
1323
1324             /* Calculate temporary vectorial force */
1325             tx               = _mm256_mul_ps(fscal,dx02);
1326             ty               = _mm256_mul_ps(fscal,dy02);
1327             tz               = _mm256_mul_ps(fscal,dz02);
1328
1329             /* Update vectorial force */
1330             fix0             = _mm256_add_ps(fix0,tx);
1331             fiy0             = _mm256_add_ps(fiy0,ty);
1332             fiz0             = _mm256_add_ps(fiz0,tz);
1333
1334             fjx2             = _mm256_add_ps(fjx2,tx);
1335             fjy2             = _mm256_add_ps(fjy2,ty);
1336             fjz2             = _mm256_add_ps(fjz2,tz);
1337
1338             /**************************
1339              * CALCULATE INTERACTIONS *
1340              **************************/
1341
1342             /* COULOMB ELECTROSTATICS */
1343             velec            = _mm256_mul_ps(qq10,rinv10);
1344             felec            = _mm256_mul_ps(velec,rinvsq10);
1345
1346             fscal            = felec;
1347
1348             /* Calculate temporary vectorial force */
1349             tx               = _mm256_mul_ps(fscal,dx10);
1350             ty               = _mm256_mul_ps(fscal,dy10);
1351             tz               = _mm256_mul_ps(fscal,dz10);
1352
1353             /* Update vectorial force */
1354             fix1             = _mm256_add_ps(fix1,tx);
1355             fiy1             = _mm256_add_ps(fiy1,ty);
1356             fiz1             = _mm256_add_ps(fiz1,tz);
1357
1358             fjx0             = _mm256_add_ps(fjx0,tx);
1359             fjy0             = _mm256_add_ps(fjy0,ty);
1360             fjz0             = _mm256_add_ps(fjz0,tz);
1361
1362             /**************************
1363              * CALCULATE INTERACTIONS *
1364              **************************/
1365
1366             /* COULOMB ELECTROSTATICS */
1367             velec            = _mm256_mul_ps(qq11,rinv11);
1368             felec            = _mm256_mul_ps(velec,rinvsq11);
1369
1370             fscal            = felec;
1371
1372             /* Calculate temporary vectorial force */
1373             tx               = _mm256_mul_ps(fscal,dx11);
1374             ty               = _mm256_mul_ps(fscal,dy11);
1375             tz               = _mm256_mul_ps(fscal,dz11);
1376
1377             /* Update vectorial force */
1378             fix1             = _mm256_add_ps(fix1,tx);
1379             fiy1             = _mm256_add_ps(fiy1,ty);
1380             fiz1             = _mm256_add_ps(fiz1,tz);
1381
1382             fjx1             = _mm256_add_ps(fjx1,tx);
1383             fjy1             = _mm256_add_ps(fjy1,ty);
1384             fjz1             = _mm256_add_ps(fjz1,tz);
1385
1386             /**************************
1387              * CALCULATE INTERACTIONS *
1388              **************************/
1389
1390             /* COULOMB ELECTROSTATICS */
1391             velec            = _mm256_mul_ps(qq12,rinv12);
1392             felec            = _mm256_mul_ps(velec,rinvsq12);
1393
1394             fscal            = felec;
1395
1396             /* Calculate temporary vectorial force */
1397             tx               = _mm256_mul_ps(fscal,dx12);
1398             ty               = _mm256_mul_ps(fscal,dy12);
1399             tz               = _mm256_mul_ps(fscal,dz12);
1400
1401             /* Update vectorial force */
1402             fix1             = _mm256_add_ps(fix1,tx);
1403             fiy1             = _mm256_add_ps(fiy1,ty);
1404             fiz1             = _mm256_add_ps(fiz1,tz);
1405
1406             fjx2             = _mm256_add_ps(fjx2,tx);
1407             fjy2             = _mm256_add_ps(fjy2,ty);
1408             fjz2             = _mm256_add_ps(fjz2,tz);
1409
1410             /**************************
1411              * CALCULATE INTERACTIONS *
1412              **************************/
1413
1414             /* COULOMB ELECTROSTATICS */
1415             velec            = _mm256_mul_ps(qq20,rinv20);
1416             felec            = _mm256_mul_ps(velec,rinvsq20);
1417
1418             fscal            = felec;
1419
1420             /* Calculate temporary vectorial force */
1421             tx               = _mm256_mul_ps(fscal,dx20);
1422             ty               = _mm256_mul_ps(fscal,dy20);
1423             tz               = _mm256_mul_ps(fscal,dz20);
1424
1425             /* Update vectorial force */
1426             fix2             = _mm256_add_ps(fix2,tx);
1427             fiy2             = _mm256_add_ps(fiy2,ty);
1428             fiz2             = _mm256_add_ps(fiz2,tz);
1429
1430             fjx0             = _mm256_add_ps(fjx0,tx);
1431             fjy0             = _mm256_add_ps(fjy0,ty);
1432             fjz0             = _mm256_add_ps(fjz0,tz);
1433
1434             /**************************
1435              * CALCULATE INTERACTIONS *
1436              **************************/
1437
1438             /* COULOMB ELECTROSTATICS */
1439             velec            = _mm256_mul_ps(qq21,rinv21);
1440             felec            = _mm256_mul_ps(velec,rinvsq21);
1441
1442             fscal            = felec;
1443
1444             /* Calculate temporary vectorial force */
1445             tx               = _mm256_mul_ps(fscal,dx21);
1446             ty               = _mm256_mul_ps(fscal,dy21);
1447             tz               = _mm256_mul_ps(fscal,dz21);
1448
1449             /* Update vectorial force */
1450             fix2             = _mm256_add_ps(fix2,tx);
1451             fiy2             = _mm256_add_ps(fiy2,ty);
1452             fiz2             = _mm256_add_ps(fiz2,tz);
1453
1454             fjx1             = _mm256_add_ps(fjx1,tx);
1455             fjy1             = _mm256_add_ps(fjy1,ty);
1456             fjz1             = _mm256_add_ps(fjz1,tz);
1457
1458             /**************************
1459              * CALCULATE INTERACTIONS *
1460              **************************/
1461
1462             /* COULOMB ELECTROSTATICS */
1463             velec            = _mm256_mul_ps(qq22,rinv22);
1464             felec            = _mm256_mul_ps(velec,rinvsq22);
1465
1466             fscal            = felec;
1467
1468             /* Calculate temporary vectorial force */
1469             tx               = _mm256_mul_ps(fscal,dx22);
1470             ty               = _mm256_mul_ps(fscal,dy22);
1471             tz               = _mm256_mul_ps(fscal,dz22);
1472
1473             /* Update vectorial force */
1474             fix2             = _mm256_add_ps(fix2,tx);
1475             fiy2             = _mm256_add_ps(fiy2,ty);
1476             fiz2             = _mm256_add_ps(fiz2,tz);
1477
1478             fjx2             = _mm256_add_ps(fjx2,tx);
1479             fjy2             = _mm256_add_ps(fjy2,ty);
1480             fjz2             = _mm256_add_ps(fjz2,tz);
1481
1482             fjptrA             = f+j_coord_offsetA;
1483             fjptrB             = f+j_coord_offsetB;
1484             fjptrC             = f+j_coord_offsetC;
1485             fjptrD             = f+j_coord_offsetD;
1486             fjptrE             = f+j_coord_offsetE;
1487             fjptrF             = f+j_coord_offsetF;
1488             fjptrG             = f+j_coord_offsetG;
1489             fjptrH             = f+j_coord_offsetH;
1490
1491             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1492                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1493
1494             /* Inner loop uses 241 flops */
1495         }
1496
1497         if(jidx<j_index_end)
1498         {
1499
1500             /* Get j neighbor index, and coordinate index */
1501             jnrlistA         = jjnr[jidx];
1502             jnrlistB         = jjnr[jidx+1];
1503             jnrlistC         = jjnr[jidx+2];
1504             jnrlistD         = jjnr[jidx+3];
1505             jnrlistE         = jjnr[jidx+4];
1506             jnrlistF         = jjnr[jidx+5];
1507             jnrlistG         = jjnr[jidx+6];
1508             jnrlistH         = jjnr[jidx+7];
1509             /* Sign of each element will be negative for non-real atoms.
1510              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1511              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1512              */
1513             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1514                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1515
1516             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1517             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1518             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1519             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1520             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1521             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1522             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1523             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1524             j_coord_offsetA  = DIM*jnrA;
1525             j_coord_offsetB  = DIM*jnrB;
1526             j_coord_offsetC  = DIM*jnrC;
1527             j_coord_offsetD  = DIM*jnrD;
1528             j_coord_offsetE  = DIM*jnrE;
1529             j_coord_offsetF  = DIM*jnrF;
1530             j_coord_offsetG  = DIM*jnrG;
1531             j_coord_offsetH  = DIM*jnrH;
1532
1533             /* load j atom coordinates */
1534             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1535                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1536                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1537                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1538                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1539
1540             /* Calculate displacement vector */
1541             dx00             = _mm256_sub_ps(ix0,jx0);
1542             dy00             = _mm256_sub_ps(iy0,jy0);
1543             dz00             = _mm256_sub_ps(iz0,jz0);
1544             dx01             = _mm256_sub_ps(ix0,jx1);
1545             dy01             = _mm256_sub_ps(iy0,jy1);
1546             dz01             = _mm256_sub_ps(iz0,jz1);
1547             dx02             = _mm256_sub_ps(ix0,jx2);
1548             dy02             = _mm256_sub_ps(iy0,jy2);
1549             dz02             = _mm256_sub_ps(iz0,jz2);
1550             dx10             = _mm256_sub_ps(ix1,jx0);
1551             dy10             = _mm256_sub_ps(iy1,jy0);
1552             dz10             = _mm256_sub_ps(iz1,jz0);
1553             dx11             = _mm256_sub_ps(ix1,jx1);
1554             dy11             = _mm256_sub_ps(iy1,jy1);
1555             dz11             = _mm256_sub_ps(iz1,jz1);
1556             dx12             = _mm256_sub_ps(ix1,jx2);
1557             dy12             = _mm256_sub_ps(iy1,jy2);
1558             dz12             = _mm256_sub_ps(iz1,jz2);
1559             dx20             = _mm256_sub_ps(ix2,jx0);
1560             dy20             = _mm256_sub_ps(iy2,jy0);
1561             dz20             = _mm256_sub_ps(iz2,jz0);
1562             dx21             = _mm256_sub_ps(ix2,jx1);
1563             dy21             = _mm256_sub_ps(iy2,jy1);
1564             dz21             = _mm256_sub_ps(iz2,jz1);
1565             dx22             = _mm256_sub_ps(ix2,jx2);
1566             dy22             = _mm256_sub_ps(iy2,jy2);
1567             dz22             = _mm256_sub_ps(iz2,jz2);
1568
1569             /* Calculate squared distance and things based on it */
1570             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1571             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1572             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1573             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1574             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1575             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1576             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1577             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1578             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1579
1580             rinv00           = avx256_invsqrt_f(rsq00);
1581             rinv01           = avx256_invsqrt_f(rsq01);
1582             rinv02           = avx256_invsqrt_f(rsq02);
1583             rinv10           = avx256_invsqrt_f(rsq10);
1584             rinv11           = avx256_invsqrt_f(rsq11);
1585             rinv12           = avx256_invsqrt_f(rsq12);
1586             rinv20           = avx256_invsqrt_f(rsq20);
1587             rinv21           = avx256_invsqrt_f(rsq21);
1588             rinv22           = avx256_invsqrt_f(rsq22);
1589
1590             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1591             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1592             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1593             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1594             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1595             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1596             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1597             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1598             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1599
1600             fjx0             = _mm256_setzero_ps();
1601             fjy0             = _mm256_setzero_ps();
1602             fjz0             = _mm256_setzero_ps();
1603             fjx1             = _mm256_setzero_ps();
1604             fjy1             = _mm256_setzero_ps();
1605             fjz1             = _mm256_setzero_ps();
1606             fjx2             = _mm256_setzero_ps();
1607             fjy2             = _mm256_setzero_ps();
1608             fjz2             = _mm256_setzero_ps();
1609
1610             /**************************
1611              * CALCULATE INTERACTIONS *
1612              **************************/
1613
1614             /* COULOMB ELECTROSTATICS */
1615             velec            = _mm256_mul_ps(qq00,rinv00);
1616             felec            = _mm256_mul_ps(velec,rinvsq00);
1617
1618             /* LENNARD-JONES DISPERSION/REPULSION */
1619
1620             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1621             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1622
1623             fscal            = _mm256_add_ps(felec,fvdw);
1624
1625             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1626
1627             /* Calculate temporary vectorial force */
1628             tx               = _mm256_mul_ps(fscal,dx00);
1629             ty               = _mm256_mul_ps(fscal,dy00);
1630             tz               = _mm256_mul_ps(fscal,dz00);
1631
1632             /* Update vectorial force */
1633             fix0             = _mm256_add_ps(fix0,tx);
1634             fiy0             = _mm256_add_ps(fiy0,ty);
1635             fiz0             = _mm256_add_ps(fiz0,tz);
1636
1637             fjx0             = _mm256_add_ps(fjx0,tx);
1638             fjy0             = _mm256_add_ps(fjy0,ty);
1639             fjz0             = _mm256_add_ps(fjz0,tz);
1640
1641             /**************************
1642              * CALCULATE INTERACTIONS *
1643              **************************/
1644
1645             /* COULOMB ELECTROSTATICS */
1646             velec            = _mm256_mul_ps(qq01,rinv01);
1647             felec            = _mm256_mul_ps(velec,rinvsq01);
1648
1649             fscal            = felec;
1650
1651             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1652
1653             /* Calculate temporary vectorial force */
1654             tx               = _mm256_mul_ps(fscal,dx01);
1655             ty               = _mm256_mul_ps(fscal,dy01);
1656             tz               = _mm256_mul_ps(fscal,dz01);
1657
1658             /* Update vectorial force */
1659             fix0             = _mm256_add_ps(fix0,tx);
1660             fiy0             = _mm256_add_ps(fiy0,ty);
1661             fiz0             = _mm256_add_ps(fiz0,tz);
1662
1663             fjx1             = _mm256_add_ps(fjx1,tx);
1664             fjy1             = _mm256_add_ps(fjy1,ty);
1665             fjz1             = _mm256_add_ps(fjz1,tz);
1666
1667             /**************************
1668              * CALCULATE INTERACTIONS *
1669              **************************/
1670
1671             /* COULOMB ELECTROSTATICS */
1672             velec            = _mm256_mul_ps(qq02,rinv02);
1673             felec            = _mm256_mul_ps(velec,rinvsq02);
1674
1675             fscal            = felec;
1676
1677             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1678
1679             /* Calculate temporary vectorial force */
1680             tx               = _mm256_mul_ps(fscal,dx02);
1681             ty               = _mm256_mul_ps(fscal,dy02);
1682             tz               = _mm256_mul_ps(fscal,dz02);
1683
1684             /* Update vectorial force */
1685             fix0             = _mm256_add_ps(fix0,tx);
1686             fiy0             = _mm256_add_ps(fiy0,ty);
1687             fiz0             = _mm256_add_ps(fiz0,tz);
1688
1689             fjx2             = _mm256_add_ps(fjx2,tx);
1690             fjy2             = _mm256_add_ps(fjy2,ty);
1691             fjz2             = _mm256_add_ps(fjz2,tz);
1692
1693             /**************************
1694              * CALCULATE INTERACTIONS *
1695              **************************/
1696
1697             /* COULOMB ELECTROSTATICS */
1698             velec            = _mm256_mul_ps(qq10,rinv10);
1699             felec            = _mm256_mul_ps(velec,rinvsq10);
1700
1701             fscal            = felec;
1702
1703             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1704
1705             /* Calculate temporary vectorial force */
1706             tx               = _mm256_mul_ps(fscal,dx10);
1707             ty               = _mm256_mul_ps(fscal,dy10);
1708             tz               = _mm256_mul_ps(fscal,dz10);
1709
1710             /* Update vectorial force */
1711             fix1             = _mm256_add_ps(fix1,tx);
1712             fiy1             = _mm256_add_ps(fiy1,ty);
1713             fiz1             = _mm256_add_ps(fiz1,tz);
1714
1715             fjx0             = _mm256_add_ps(fjx0,tx);
1716             fjy0             = _mm256_add_ps(fjy0,ty);
1717             fjz0             = _mm256_add_ps(fjz0,tz);
1718
1719             /**************************
1720              * CALCULATE INTERACTIONS *
1721              **************************/
1722
1723             /* COULOMB ELECTROSTATICS */
1724             velec            = _mm256_mul_ps(qq11,rinv11);
1725             felec            = _mm256_mul_ps(velec,rinvsq11);
1726
1727             fscal            = felec;
1728
1729             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1730
1731             /* Calculate temporary vectorial force */
1732             tx               = _mm256_mul_ps(fscal,dx11);
1733             ty               = _mm256_mul_ps(fscal,dy11);
1734             tz               = _mm256_mul_ps(fscal,dz11);
1735
1736             /* Update vectorial force */
1737             fix1             = _mm256_add_ps(fix1,tx);
1738             fiy1             = _mm256_add_ps(fiy1,ty);
1739             fiz1             = _mm256_add_ps(fiz1,tz);
1740
1741             fjx1             = _mm256_add_ps(fjx1,tx);
1742             fjy1             = _mm256_add_ps(fjy1,ty);
1743             fjz1             = _mm256_add_ps(fjz1,tz);
1744
1745             /**************************
1746              * CALCULATE INTERACTIONS *
1747              **************************/
1748
1749             /* COULOMB ELECTROSTATICS */
1750             velec            = _mm256_mul_ps(qq12,rinv12);
1751             felec            = _mm256_mul_ps(velec,rinvsq12);
1752
1753             fscal            = felec;
1754
1755             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1756
1757             /* Calculate temporary vectorial force */
1758             tx               = _mm256_mul_ps(fscal,dx12);
1759             ty               = _mm256_mul_ps(fscal,dy12);
1760             tz               = _mm256_mul_ps(fscal,dz12);
1761
1762             /* Update vectorial force */
1763             fix1             = _mm256_add_ps(fix1,tx);
1764             fiy1             = _mm256_add_ps(fiy1,ty);
1765             fiz1             = _mm256_add_ps(fiz1,tz);
1766
1767             fjx2             = _mm256_add_ps(fjx2,tx);
1768             fjy2             = _mm256_add_ps(fjy2,ty);
1769             fjz2             = _mm256_add_ps(fjz2,tz);
1770
1771             /**************************
1772              * CALCULATE INTERACTIONS *
1773              **************************/
1774
1775             /* COULOMB ELECTROSTATICS */
1776             velec            = _mm256_mul_ps(qq20,rinv20);
1777             felec            = _mm256_mul_ps(velec,rinvsq20);
1778
1779             fscal            = felec;
1780
1781             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1782
1783             /* Calculate temporary vectorial force */
1784             tx               = _mm256_mul_ps(fscal,dx20);
1785             ty               = _mm256_mul_ps(fscal,dy20);
1786             tz               = _mm256_mul_ps(fscal,dz20);
1787
1788             /* Update vectorial force */
1789             fix2             = _mm256_add_ps(fix2,tx);
1790             fiy2             = _mm256_add_ps(fiy2,ty);
1791             fiz2             = _mm256_add_ps(fiz2,tz);
1792
1793             fjx0             = _mm256_add_ps(fjx0,tx);
1794             fjy0             = _mm256_add_ps(fjy0,ty);
1795             fjz0             = _mm256_add_ps(fjz0,tz);
1796
1797             /**************************
1798              * CALCULATE INTERACTIONS *
1799              **************************/
1800
1801             /* COULOMB ELECTROSTATICS */
1802             velec            = _mm256_mul_ps(qq21,rinv21);
1803             felec            = _mm256_mul_ps(velec,rinvsq21);
1804
1805             fscal            = felec;
1806
1807             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1808
1809             /* Calculate temporary vectorial force */
1810             tx               = _mm256_mul_ps(fscal,dx21);
1811             ty               = _mm256_mul_ps(fscal,dy21);
1812             tz               = _mm256_mul_ps(fscal,dz21);
1813
1814             /* Update vectorial force */
1815             fix2             = _mm256_add_ps(fix2,tx);
1816             fiy2             = _mm256_add_ps(fiy2,ty);
1817             fiz2             = _mm256_add_ps(fiz2,tz);
1818
1819             fjx1             = _mm256_add_ps(fjx1,tx);
1820             fjy1             = _mm256_add_ps(fjy1,ty);
1821             fjz1             = _mm256_add_ps(fjz1,tz);
1822
1823             /**************************
1824              * CALCULATE INTERACTIONS *
1825              **************************/
1826
1827             /* COULOMB ELECTROSTATICS */
1828             velec            = _mm256_mul_ps(qq22,rinv22);
1829             felec            = _mm256_mul_ps(velec,rinvsq22);
1830
1831             fscal            = felec;
1832
1833             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1834
1835             /* Calculate temporary vectorial force */
1836             tx               = _mm256_mul_ps(fscal,dx22);
1837             ty               = _mm256_mul_ps(fscal,dy22);
1838             tz               = _mm256_mul_ps(fscal,dz22);
1839
1840             /* Update vectorial force */
1841             fix2             = _mm256_add_ps(fix2,tx);
1842             fiy2             = _mm256_add_ps(fiy2,ty);
1843             fiz2             = _mm256_add_ps(fiz2,tz);
1844
1845             fjx2             = _mm256_add_ps(fjx2,tx);
1846             fjy2             = _mm256_add_ps(fjy2,ty);
1847             fjz2             = _mm256_add_ps(fjz2,tz);
1848
1849             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1850             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1851             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1852             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1853             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1854             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1855             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1856             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1857
1858             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1859                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1860
1861             /* Inner loop uses 241 flops */
1862         }
1863
1864         /* End of innermost loop */
1865
1866         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1867                                                  f+i_coord_offset,fshift+i_shift_offset);
1868
1869         /* Increment number of inner iterations */
1870         inneriter                  += j_index_end - j_index_start;
1871
1872         /* Outer loop uses 18 flops */
1873     }
1874
1875     /* Increment number of outer iterations */
1876     outeriter        += nri;
1877
1878     /* Update outer/inner flops */
1879
1880     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*241);
1881 }