src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_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/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_256_single
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_256_single
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrE,jnrF,jnrG,jnrH;
  78     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  79     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  80     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  81     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  86     real             scratch[4*DIM];
  87     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     real *           vdwioffsetptr0;
  89     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  90     real *           vdwioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  94     real *           vdwioffsetptr3;
  95     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  96     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  97     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  98     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  99     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
 100     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
 101     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 102     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 103     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 104     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 105     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 106     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 107     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 108     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 109     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 110     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 111     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 112     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 113     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 114     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 115     real             *charge;
 116     int              nvdwtype;
 117     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 118     int              *vdwtype;
 119     real             *vdwparam;
 120     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 121     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 122     __m256i          vfitab;
 123     __m128i          vfitab_lo,vfitab_hi;
 124     __m128i          ifour       = _mm_set1_epi32(4);
 125     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 126     real             *vftab;
 127     __m256           dummy_mask,cutoff_mask;
 128     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 129     __m256           one     = _mm256_set1_ps(1.0);
 130     __m256           two     = _mm256_set1_ps(2.0);
 131     x                = xx[0];
 132     f                = ff[0];
 133
 134     nri              = nlist->nri;
 135     iinr             = nlist->iinr;
 136     jindex           = nlist->jindex;
 137     jjnr             = nlist->jjnr;
 138     shiftidx         = nlist->shift;
 139     gid              = nlist->gid;
 140     shiftvec         = fr->shift_vec[0];
 141     fshift           = fr->fshift[0];
 142     facel            = _mm256_set1_ps(fr->epsfac);
 143     charge           = mdatoms->chargeA;
 144     nvdwtype         = fr->ntype;
 145     vdwparam         = fr->nbfp;
 146     vdwtype          = mdatoms->typeA;
 147
 148     vftab            = kernel_data->table_vdw->data;
 149     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 150
 151     /* Setup water-specific parameters */
 152     inr              = nlist->iinr[0];
 153     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 154     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 155     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 156     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 157
 158     jq1              = _mm256_set1_ps(charge[inr+1]);
 159     jq2              = _mm256_set1_ps(charge[inr+2]);
 160     jq3              = _mm256_set1_ps(charge[inr+3]);
 161     vdwjidx0A        = 2*vdwtype[inr+0];
 162     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 163     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 164     qq11             = _mm256_mul_ps(iq1,jq1);
 165     qq12             = _mm256_mul_ps(iq1,jq2);
 166     qq13             = _mm256_mul_ps(iq1,jq3);
 167     qq21             = _mm256_mul_ps(iq2,jq1);
 168     qq22             = _mm256_mul_ps(iq2,jq2);
 169     qq23             = _mm256_mul_ps(iq2,jq3);
 170     qq31             = _mm256_mul_ps(iq3,jq1);
 171     qq32             = _mm256_mul_ps(iq3,jq2);
 172     qq33             = _mm256_mul_ps(iq3,jq3);
 173
 174     /* Avoid stupid compiler warnings */
 175     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 176     j_coord_offsetA = 0;
 177     j_coord_offsetB = 0;
 178     j_coord_offsetC = 0;
 179     j_coord_offsetD = 0;
 180     j_coord_offsetE = 0;
 181     j_coord_offsetF = 0;
 182     j_coord_offsetG = 0;
 183     j_coord_offsetH = 0;
 184
 185     outeriter        = 0;
 186     inneriter        = 0;
 187
 188     for(iidx=0;iidx<4*DIM;iidx++)
 189     {
 190         scratch[iidx] = 0.0;
 191     }
 192
 193     /* Start outer loop over neighborlists */
 194     for(iidx=0; iidx<nri; iidx++)
 195     {
 196         /* Load shift vector for this list */
 197         i_shift_offset   = DIM*shiftidx[iidx];
 198
 199         /* Load limits for loop over neighbors */
 200         j_index_start    = jindex[iidx];
 201         j_index_end      = jindex[iidx+1];
 202
 203         /* Get outer coordinate index */
 204         inr              = iinr[iidx];
 205         i_coord_offset   = DIM*inr;
 206
 207         /* Load i particle coords and add shift vector */
 208         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 209                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 210
 211         fix0             = _mm256_setzero_ps();
 212         fiy0             = _mm256_setzero_ps();
 213         fiz0             = _mm256_setzero_ps();
 214         fix1             = _mm256_setzero_ps();
 215         fiy1             = _mm256_setzero_ps();
 216         fiz1             = _mm256_setzero_ps();
 217         fix2             = _mm256_setzero_ps();
 218         fiy2             = _mm256_setzero_ps();
 219         fiz2             = _mm256_setzero_ps();
 220         fix3             = _mm256_setzero_ps();
 221         fiy3             = _mm256_setzero_ps();
 222         fiz3             = _mm256_setzero_ps();
 223
 224         /* Reset potential sums */
 225         velecsum         = _mm256_setzero_ps();
 226         vvdwsum          = _mm256_setzero_ps();
 227
 228         /* Start inner kernel loop */
 229         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 230         {
 231
 232             /* Get j neighbor index, and coordinate index */
 233             jnrA             = jjnr[jidx];
 234             jnrB             = jjnr[jidx+1];
 235             jnrC             = jjnr[jidx+2];
 236             jnrD             = jjnr[jidx+3];
 237             jnrE             = jjnr[jidx+4];
 238             jnrF             = jjnr[jidx+5];
 239             jnrG             = jjnr[jidx+6];
 240             jnrH             = jjnr[jidx+7];
 241             j_coord_offsetA  = DIM*jnrA;
 242             j_coord_offsetB  = DIM*jnrB;
 243             j_coord_offsetC  = DIM*jnrC;
 244             j_coord_offsetD  = DIM*jnrD;
 245             j_coord_offsetE  = DIM*jnrE;
 246             j_coord_offsetF  = DIM*jnrF;
 247             j_coord_offsetG  = DIM*jnrG;
 248             j_coord_offsetH  = DIM*jnrH;
 249
 250             /* load j atom coordinates */
 251             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 252                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 253                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 254                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 255                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 256                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 257
 258             /* Calculate displacement vector */
 259             dx00             = _mm256_sub_ps(ix0,jx0);
 260             dy00             = _mm256_sub_ps(iy0,jy0);
 261             dz00             = _mm256_sub_ps(iz0,jz0);
 262             dx11             = _mm256_sub_ps(ix1,jx1);
 263             dy11             = _mm256_sub_ps(iy1,jy1);
 264             dz11             = _mm256_sub_ps(iz1,jz1);
 265             dx12             = _mm256_sub_ps(ix1,jx2);
 266             dy12             = _mm256_sub_ps(iy1,jy2);
 267             dz12             = _mm256_sub_ps(iz1,jz2);
 268             dx13             = _mm256_sub_ps(ix1,jx3);
 269             dy13             = _mm256_sub_ps(iy1,jy3);
 270             dz13             = _mm256_sub_ps(iz1,jz3);
 271             dx21             = _mm256_sub_ps(ix2,jx1);
 272             dy21             = _mm256_sub_ps(iy2,jy1);
 273             dz21             = _mm256_sub_ps(iz2,jz1);
 274             dx22             = _mm256_sub_ps(ix2,jx2);
 275             dy22             = _mm256_sub_ps(iy2,jy2);
 276             dz22             = _mm256_sub_ps(iz2,jz2);
 277             dx23             = _mm256_sub_ps(ix2,jx3);
 278             dy23             = _mm256_sub_ps(iy2,jy3);
 279             dz23             = _mm256_sub_ps(iz2,jz3);
 280             dx31             = _mm256_sub_ps(ix3,jx1);
 281             dy31             = _mm256_sub_ps(iy3,jy1);
 282             dz31             = _mm256_sub_ps(iz3,jz1);
 283             dx32             = _mm256_sub_ps(ix3,jx2);
 284             dy32             = _mm256_sub_ps(iy3,jy2);
 285             dz32             = _mm256_sub_ps(iz3,jz2);
 286             dx33             = _mm256_sub_ps(ix3,jx3);
 287             dy33             = _mm256_sub_ps(iy3,jy3);
 288             dz33             = _mm256_sub_ps(iz3,jz3);
 289
 290             /* Calculate squared distance and things based on it */
 291             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 292             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 293             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 294             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 295             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 296             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 297             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 298             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 299             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 300             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 301
 302             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 303             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 304             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 305             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 306             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 307             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 308             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 309             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 310             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 311             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 312
 313             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 314             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 315             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 316             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 317             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 318             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 319             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 320             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 321             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 322
 323             fjx0             = _mm256_setzero_ps();
 324             fjy0             = _mm256_setzero_ps();
 325             fjz0             = _mm256_setzero_ps();
 326             fjx1             = _mm256_setzero_ps();
 327             fjy1             = _mm256_setzero_ps();
 328             fjz1             = _mm256_setzero_ps();
 329             fjx2             = _mm256_setzero_ps();
 330             fjy2             = _mm256_setzero_ps();
 331             fjz2             = _mm256_setzero_ps();
 332             fjx3             = _mm256_setzero_ps();
 333             fjy3             = _mm256_setzero_ps();
 334             fjz3             = _mm256_setzero_ps();
 335
 336             /**************************
 337              * CALCULATE INTERACTIONS *
 338              **************************/
 339
 340             r00              = _mm256_mul_ps(rsq00,rinv00);
 341
 342             /* Calculate table index by multiplying r with table scale and truncate to integer */
 343             rt               = _mm256_mul_ps(r00,vftabscale);
 344             vfitab           = _mm256_cvttps_epi32(rt);
 345             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 346             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 347             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 348             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 349             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 350             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 351
 352             /* CUBIC SPLINE TABLE DISPERSION */
 353             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 354                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 355             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 356                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 357             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 358                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 359             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 360                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 361             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 362             Heps             = _mm256_mul_ps(vfeps,H);
 363             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 364             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 365             vvdw6            = _mm256_mul_ps(c6_00,VV);
 366             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 367             fvdw6            = _mm256_mul_ps(c6_00,FF);
 368
 369             /* CUBIC SPLINE TABLE REPULSION */
 370             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 371             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 372             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 373                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 374             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 375                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 376             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 377                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 378             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 379                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 380             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 381             Heps             = _mm256_mul_ps(vfeps,H);
 382             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 383             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 384             vvdw12           = _mm256_mul_ps(c12_00,VV);
 385             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 386             fvdw12           = _mm256_mul_ps(c12_00,FF);
 387             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 388             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 392
 393             fscal            = fvdw;
 394
 395             /* Calculate temporary vectorial force */
 396             tx               = _mm256_mul_ps(fscal,dx00);
 397             ty               = _mm256_mul_ps(fscal,dy00);
 398             tz               = _mm256_mul_ps(fscal,dz00);
 399
 400             /* Update vectorial force */
 401             fix0             = _mm256_add_ps(fix0,tx);
 402             fiy0             = _mm256_add_ps(fiy0,ty);
 403             fiz0             = _mm256_add_ps(fiz0,tz);
 404
 405             fjx0             = _mm256_add_ps(fjx0,tx);
 406             fjy0             = _mm256_add_ps(fjy0,ty);
 407             fjz0             = _mm256_add_ps(fjz0,tz);
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             /* COULOMB ELECTROSTATICS */
 414             velec            = _mm256_mul_ps(qq11,rinv11);
 415             felec            = _mm256_mul_ps(velec,rinvsq11);
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velecsum         = _mm256_add_ps(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             /* Calculate temporary vectorial force */
 423             tx               = _mm256_mul_ps(fscal,dx11);
 424             ty               = _mm256_mul_ps(fscal,dy11);
 425             tz               = _mm256_mul_ps(fscal,dz11);
 426
 427             /* Update vectorial force */
 428             fix1             = _mm256_add_ps(fix1,tx);
 429             fiy1             = _mm256_add_ps(fiy1,ty);
 430             fiz1             = _mm256_add_ps(fiz1,tz);
 431
 432             fjx1             = _mm256_add_ps(fjx1,tx);
 433             fjy1             = _mm256_add_ps(fjy1,ty);
 434             fjz1             = _mm256_add_ps(fjz1,tz);
 435
 436             /**************************
 437              * CALCULATE INTERACTIONS *
 438              **************************/
 439
 440             /* COULOMB ELECTROSTATICS */
 441             velec            = _mm256_mul_ps(qq12,rinv12);
 442             felec            = _mm256_mul_ps(velec,rinvsq12);
 443
 444             /* Update potential sum for this i atom from the interaction with this j atom. */
 445             velecsum         = _mm256_add_ps(velecsum,velec);
 446
 447             fscal            = felec;
 448
 449             /* Calculate temporary vectorial force */
 450             tx               = _mm256_mul_ps(fscal,dx12);
 451             ty               = _mm256_mul_ps(fscal,dy12);
 452             tz               = _mm256_mul_ps(fscal,dz12);
 453
 454             /* Update vectorial force */
 455             fix1             = _mm256_add_ps(fix1,tx);
 456             fiy1             = _mm256_add_ps(fiy1,ty);
 457             fiz1             = _mm256_add_ps(fiz1,tz);
 458
 459             fjx2             = _mm256_add_ps(fjx2,tx);
 460             fjy2             = _mm256_add_ps(fjy2,ty);
 461             fjz2             = _mm256_add_ps(fjz2,tz);
 462
 463             /**************************
 464              * CALCULATE INTERACTIONS *
 465              **************************/
 466
 467             /* COULOMB ELECTROSTATICS */
 468             velec            = _mm256_mul_ps(qq13,rinv13);
 469             felec            = _mm256_mul_ps(velec,rinvsq13);
 470
 471             /* Update potential sum for this i atom from the interaction with this j atom. */
 472             velecsum         = _mm256_add_ps(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             /* Calculate temporary vectorial force */
 477             tx               = _mm256_mul_ps(fscal,dx13);
 478             ty               = _mm256_mul_ps(fscal,dy13);
 479             tz               = _mm256_mul_ps(fscal,dz13);
 480
 481             /* Update vectorial force */
 482             fix1             = _mm256_add_ps(fix1,tx);
 483             fiy1             = _mm256_add_ps(fiy1,ty);
 484             fiz1             = _mm256_add_ps(fiz1,tz);
 485
 486             fjx3             = _mm256_add_ps(fjx3,tx);
 487             fjy3             = _mm256_add_ps(fjy3,ty);
 488             fjz3             = _mm256_add_ps(fjz3,tz);
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             /* COULOMB ELECTROSTATICS */
 495             velec            = _mm256_mul_ps(qq21,rinv21);
 496             felec            = _mm256_mul_ps(velec,rinvsq21);
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velecsum         = _mm256_add_ps(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503             /* Calculate temporary vectorial force */
 504             tx               = _mm256_mul_ps(fscal,dx21);
 505             ty               = _mm256_mul_ps(fscal,dy21);
 506             tz               = _mm256_mul_ps(fscal,dz21);
 507
 508             /* Update vectorial force */
 509             fix2             = _mm256_add_ps(fix2,tx);
 510             fiy2             = _mm256_add_ps(fiy2,ty);
 511             fiz2             = _mm256_add_ps(fiz2,tz);
 512
 513             fjx1             = _mm256_add_ps(fjx1,tx);
 514             fjy1             = _mm256_add_ps(fjy1,ty);
 515             fjz1             = _mm256_add_ps(fjz1,tz);
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             /* COULOMB ELECTROSTATICS */
 522             velec            = _mm256_mul_ps(qq22,rinv22);
 523             felec            = _mm256_mul_ps(velec,rinvsq22);
 524
 525             /* Update potential sum for this i atom from the interaction with this j atom. */
 526             velecsum         = _mm256_add_ps(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             /* Calculate temporary vectorial force */
 531             tx               = _mm256_mul_ps(fscal,dx22);
 532             ty               = _mm256_mul_ps(fscal,dy22);
 533             tz               = _mm256_mul_ps(fscal,dz22);
 534
 535             /* Update vectorial force */
 536             fix2             = _mm256_add_ps(fix2,tx);
 537             fiy2             = _mm256_add_ps(fiy2,ty);
 538             fiz2             = _mm256_add_ps(fiz2,tz);
 539
 540             fjx2             = _mm256_add_ps(fjx2,tx);
 541             fjy2             = _mm256_add_ps(fjy2,ty);
 542             fjz2             = _mm256_add_ps(fjz2,tz);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             /* COULOMB ELECTROSTATICS */
 549             velec            = _mm256_mul_ps(qq23,rinv23);
 550             felec            = _mm256_mul_ps(velec,rinvsq23);
 551
 552             /* Update potential sum for this i atom from the interaction with this j atom. */
 553             velecsum         = _mm256_add_ps(velecsum,velec);
 554
 555             fscal            = felec;
 556
 557             /* Calculate temporary vectorial force */
 558             tx               = _mm256_mul_ps(fscal,dx23);
 559             ty               = _mm256_mul_ps(fscal,dy23);
 560             tz               = _mm256_mul_ps(fscal,dz23);
 561
 562             /* Update vectorial force */
 563             fix2             = _mm256_add_ps(fix2,tx);
 564             fiy2             = _mm256_add_ps(fiy2,ty);
 565             fiz2             = _mm256_add_ps(fiz2,tz);
 566
 567             fjx3             = _mm256_add_ps(fjx3,tx);
 568             fjy3             = _mm256_add_ps(fjy3,ty);
 569             fjz3             = _mm256_add_ps(fjz3,tz);
 570
 571             /**************************
 572              * CALCULATE INTERACTIONS *
 573              **************************/
 574
 575             /* COULOMB ELECTROSTATICS */
 576             velec            = _mm256_mul_ps(qq31,rinv31);
 577             felec            = _mm256_mul_ps(velec,rinvsq31);
 578
 579             /* Update potential sum for this i atom from the interaction with this j atom. */
 580             velecsum         = _mm256_add_ps(velecsum,velec);
 581
 582             fscal            = felec;
 583
 584             /* Calculate temporary vectorial force */
 585             tx               = _mm256_mul_ps(fscal,dx31);
 586             ty               = _mm256_mul_ps(fscal,dy31);
 587             tz               = _mm256_mul_ps(fscal,dz31);
 588
 589             /* Update vectorial force */
 590             fix3             = _mm256_add_ps(fix3,tx);
 591             fiy3             = _mm256_add_ps(fiy3,ty);
 592             fiz3             = _mm256_add_ps(fiz3,tz);
 593
 594             fjx1             = _mm256_add_ps(fjx1,tx);
 595             fjy1             = _mm256_add_ps(fjy1,ty);
 596             fjz1             = _mm256_add_ps(fjz1,tz);
 597
 598             /**************************
 599              * CALCULATE INTERACTIONS *
 600              **************************/
 601
 602             /* COULOMB ELECTROSTATICS */
 603             velec            = _mm256_mul_ps(qq32,rinv32);
 604             felec            = _mm256_mul_ps(velec,rinvsq32);
 605
 606             /* Update potential sum for this i atom from the interaction with this j atom. */
 607             velecsum         = _mm256_add_ps(velecsum,velec);
 608
 609             fscal            = felec;
 610
 611             /* Calculate temporary vectorial force */
 612             tx               = _mm256_mul_ps(fscal,dx32);
 613             ty               = _mm256_mul_ps(fscal,dy32);
 614             tz               = _mm256_mul_ps(fscal,dz32);
 615
 616             /* Update vectorial force */
 617             fix3             = _mm256_add_ps(fix3,tx);
 618             fiy3             = _mm256_add_ps(fiy3,ty);
 619             fiz3             = _mm256_add_ps(fiz3,tz);
 620
 621             fjx2             = _mm256_add_ps(fjx2,tx);
 622             fjy2             = _mm256_add_ps(fjy2,ty);
 623             fjz2             = _mm256_add_ps(fjz2,tz);
 624
 625             /**************************
 626              * CALCULATE INTERACTIONS *
 627              **************************/
 628
 629             /* COULOMB ELECTROSTATICS */
 630             velec            = _mm256_mul_ps(qq33,rinv33);
 631             felec            = _mm256_mul_ps(velec,rinvsq33);
 632
 633             /* Update potential sum for this i atom from the interaction with this j atom. */
 634             velecsum         = _mm256_add_ps(velecsum,velec);
 635
 636             fscal            = felec;
 637
 638             /* Calculate temporary vectorial force */
 639             tx               = _mm256_mul_ps(fscal,dx33);
 640             ty               = _mm256_mul_ps(fscal,dy33);
 641             tz               = _mm256_mul_ps(fscal,dz33);
 642
 643             /* Update vectorial force */
 644             fix3             = _mm256_add_ps(fix3,tx);
 645             fiy3             = _mm256_add_ps(fiy3,ty);
 646             fiz3             = _mm256_add_ps(fiz3,tz);
 647
 648             fjx3             = _mm256_add_ps(fjx3,tx);
 649             fjy3             = _mm256_add_ps(fjy3,ty);
 650             fjz3             = _mm256_add_ps(fjz3,tz);
 651
 652             fjptrA             = f+j_coord_offsetA;
 653             fjptrB             = f+j_coord_offsetB;
 654             fjptrC             = f+j_coord_offsetC;
 655             fjptrD             = f+j_coord_offsetD;
 656             fjptrE             = f+j_coord_offsetE;
 657             fjptrF             = f+j_coord_offsetF;
 658             fjptrG             = f+j_coord_offsetG;
 659             fjptrH             = f+j_coord_offsetH;
 660
 661             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 662                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 663                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 664
 665             /* Inner loop uses 302 flops */
 666         }
 667
 668         if(jidx<j_index_end)
 669         {
 670
 671             /* Get j neighbor index, and coordinate index */
 672             jnrlistA         = jjnr[jidx];
 673             jnrlistB         = jjnr[jidx+1];
 674             jnrlistC         = jjnr[jidx+2];
 675             jnrlistD         = jjnr[jidx+3];
 676             jnrlistE         = jjnr[jidx+4];
 677             jnrlistF         = jjnr[jidx+5];
 678             jnrlistG         = jjnr[jidx+6];
 679             jnrlistH         = jjnr[jidx+7];
 680             /* Sign of each element will be negative for non-real atoms.
 681              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 682              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 683              */
 684             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 685                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 686
 687             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 688             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 689             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 690             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 691             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 692             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 693             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 694             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 695             j_coord_offsetA  = DIM*jnrA;
 696             j_coord_offsetB  = DIM*jnrB;
 697             j_coord_offsetC  = DIM*jnrC;
 698             j_coord_offsetD  = DIM*jnrD;
 699             j_coord_offsetE  = DIM*jnrE;
 700             j_coord_offsetF  = DIM*jnrF;
 701             j_coord_offsetG  = DIM*jnrG;
 702             j_coord_offsetH  = DIM*jnrH;
 703
 704             /* load j atom coordinates */
 705             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 706                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 707                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 708                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 709                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 710                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 711
 712             /* Calculate displacement vector */
 713             dx00             = _mm256_sub_ps(ix0,jx0);
 714             dy00             = _mm256_sub_ps(iy0,jy0);
 715             dz00             = _mm256_sub_ps(iz0,jz0);
 716             dx11             = _mm256_sub_ps(ix1,jx1);
 717             dy11             = _mm256_sub_ps(iy1,jy1);
 718             dz11             = _mm256_sub_ps(iz1,jz1);
 719             dx12             = _mm256_sub_ps(ix1,jx2);
 720             dy12             = _mm256_sub_ps(iy1,jy2);
 721             dz12             = _mm256_sub_ps(iz1,jz2);
 722             dx13             = _mm256_sub_ps(ix1,jx3);
 723             dy13             = _mm256_sub_ps(iy1,jy3);
 724             dz13             = _mm256_sub_ps(iz1,jz3);
 725             dx21             = _mm256_sub_ps(ix2,jx1);
 726             dy21             = _mm256_sub_ps(iy2,jy1);
 727             dz21             = _mm256_sub_ps(iz2,jz1);
 728             dx22             = _mm256_sub_ps(ix2,jx2);
 729             dy22             = _mm256_sub_ps(iy2,jy2);
 730             dz22             = _mm256_sub_ps(iz2,jz2);
 731             dx23             = _mm256_sub_ps(ix2,jx3);
 732             dy23             = _mm256_sub_ps(iy2,jy3);
 733             dz23             = _mm256_sub_ps(iz2,jz3);
 734             dx31             = _mm256_sub_ps(ix3,jx1);
 735             dy31             = _mm256_sub_ps(iy3,jy1);
 736             dz31             = _mm256_sub_ps(iz3,jz1);
 737             dx32             = _mm256_sub_ps(ix3,jx2);
 738             dy32             = _mm256_sub_ps(iy3,jy2);
 739             dz32             = _mm256_sub_ps(iz3,jz2);
 740             dx33             = _mm256_sub_ps(ix3,jx3);
 741             dy33             = _mm256_sub_ps(iy3,jy3);
 742             dz33             = _mm256_sub_ps(iz3,jz3);
 743
 744             /* Calculate squared distance and things based on it */
 745             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 746             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 747             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 748             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 749             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 750             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 751             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 752             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 753             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 754             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 755
 756             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 757             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 758             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 759             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 760             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 761             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 762             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 763             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 764             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 765             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 766
 767             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 768             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 769             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 770             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 771             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 772             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 773             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 774             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 775             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 776
 777             fjx0             = _mm256_setzero_ps();
 778             fjy0             = _mm256_setzero_ps();
 779             fjz0             = _mm256_setzero_ps();
 780             fjx1             = _mm256_setzero_ps();
 781             fjy1             = _mm256_setzero_ps();
 782             fjz1             = _mm256_setzero_ps();
 783             fjx2             = _mm256_setzero_ps();
 784             fjy2             = _mm256_setzero_ps();
 785             fjz2             = _mm256_setzero_ps();
 786             fjx3             = _mm256_setzero_ps();
 787             fjy3             = _mm256_setzero_ps();
 788             fjz3             = _mm256_setzero_ps();
 789
 790             /**************************
 791              * CALCULATE INTERACTIONS *
 792              **************************/
 793
 794             r00              = _mm256_mul_ps(rsq00,rinv00);
 795             r00              = _mm256_andnot_ps(dummy_mask,r00);
 796
 797             /* Calculate table index by multiplying r with table scale and truncate to integer */
 798             rt               = _mm256_mul_ps(r00,vftabscale);
 799             vfitab           = _mm256_cvttps_epi32(rt);
 800             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 801             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 802             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 803             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 804             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 805             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 806
 807             /* CUBIC SPLINE TABLE DISPERSION */
 808             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 809                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 810             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 811                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 812             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 813                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 814             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 815                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 816             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 817             Heps             = _mm256_mul_ps(vfeps,H);
 818             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 819             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 820             vvdw6            = _mm256_mul_ps(c6_00,VV);
 821             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 822             fvdw6            = _mm256_mul_ps(c6_00,FF);
 823
 824             /* CUBIC SPLINE TABLE REPULSION */
 825             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 826             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 827             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 828                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 829             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 830                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 831             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 832                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 833             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 834                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 835             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 836             Heps             = _mm256_mul_ps(vfeps,H);
 837             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 838             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 839             vvdw12           = _mm256_mul_ps(c12_00,VV);
 840             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 841             fvdw12           = _mm256_mul_ps(c12_00,FF);
 842             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 843             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 844
 845             /* Update potential sum for this i atom from the interaction with this j atom. */
 846             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 847             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 848
 849             fscal            = fvdw;
 850
 851             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 852
 853             /* Calculate temporary vectorial force */
 854             tx               = _mm256_mul_ps(fscal,dx00);
 855             ty               = _mm256_mul_ps(fscal,dy00);
 856             tz               = _mm256_mul_ps(fscal,dz00);
 857
 858             /* Update vectorial force */
 859             fix0             = _mm256_add_ps(fix0,tx);
 860             fiy0             = _mm256_add_ps(fiy0,ty);
 861             fiz0             = _mm256_add_ps(fiz0,tz);
 862
 863             fjx0             = _mm256_add_ps(fjx0,tx);
 864             fjy0             = _mm256_add_ps(fjy0,ty);
 865             fjz0             = _mm256_add_ps(fjz0,tz);
 866
 867             /**************************
 868              * CALCULATE INTERACTIONS *
 869              **************************/
 870
 871             /* COULOMB ELECTROSTATICS */
 872             velec            = _mm256_mul_ps(qq11,rinv11);
 873             felec            = _mm256_mul_ps(velec,rinvsq11);
 874
 875             /* Update potential sum for this i atom from the interaction with this j atom. */
 876             velec            = _mm256_andnot_ps(dummy_mask,velec);
 877             velecsum         = _mm256_add_ps(velecsum,velec);
 878
 879             fscal            = felec;
 880
 881             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 882
 883             /* Calculate temporary vectorial force */
 884             tx               = _mm256_mul_ps(fscal,dx11);
 885             ty               = _mm256_mul_ps(fscal,dy11);
 886             tz               = _mm256_mul_ps(fscal,dz11);
 887
 888             /* Update vectorial force */
 889             fix1             = _mm256_add_ps(fix1,tx);
 890             fiy1             = _mm256_add_ps(fiy1,ty);
 891             fiz1             = _mm256_add_ps(fiz1,tz);
 892
 893             fjx1             = _mm256_add_ps(fjx1,tx);
 894             fjy1             = _mm256_add_ps(fjy1,ty);
 895             fjz1             = _mm256_add_ps(fjz1,tz);
 896
 897             /**************************
 898              * CALCULATE INTERACTIONS *
 899              **************************/
 900
 901             /* COULOMB ELECTROSTATICS */
 902             velec            = _mm256_mul_ps(qq12,rinv12);
 903             felec            = _mm256_mul_ps(velec,rinvsq12);
 904
 905             /* Update potential sum for this i atom from the interaction with this j atom. */
 906             velec            = _mm256_andnot_ps(dummy_mask,velec);
 907             velecsum         = _mm256_add_ps(velecsum,velec);
 908
 909             fscal            = felec;
 910
 911             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 912
 913             /* Calculate temporary vectorial force */
 914             tx               = _mm256_mul_ps(fscal,dx12);
 915             ty               = _mm256_mul_ps(fscal,dy12);
 916             tz               = _mm256_mul_ps(fscal,dz12);
 917
 918             /* Update vectorial force */
 919             fix1             = _mm256_add_ps(fix1,tx);
 920             fiy1             = _mm256_add_ps(fiy1,ty);
 921             fiz1             = _mm256_add_ps(fiz1,tz);
 922
 923             fjx2             = _mm256_add_ps(fjx2,tx);
 924             fjy2             = _mm256_add_ps(fjy2,ty);
 925             fjz2             = _mm256_add_ps(fjz2,tz);
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             /* COULOMB ELECTROSTATICS */
 932             velec            = _mm256_mul_ps(qq13,rinv13);
 933             felec            = _mm256_mul_ps(velec,rinvsq13);
 934
 935             /* Update potential sum for this i atom from the interaction with this j atom. */
 936             velec            = _mm256_andnot_ps(dummy_mask,velec);
 937             velecsum         = _mm256_add_ps(velecsum,velec);
 938
 939             fscal            = felec;
 940
 941             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 942
 943             /* Calculate temporary vectorial force */
 944             tx               = _mm256_mul_ps(fscal,dx13);
 945             ty               = _mm256_mul_ps(fscal,dy13);
 946             tz               = _mm256_mul_ps(fscal,dz13);
 947
 948             /* Update vectorial force */
 949             fix1             = _mm256_add_ps(fix1,tx);
 950             fiy1             = _mm256_add_ps(fiy1,ty);
 951             fiz1             = _mm256_add_ps(fiz1,tz);
 952
 953             fjx3             = _mm256_add_ps(fjx3,tx);
 954             fjy3             = _mm256_add_ps(fjy3,ty);
 955             fjz3             = _mm256_add_ps(fjz3,tz);
 956
 957             /**************************
 958              * CALCULATE INTERACTIONS *
 959              **************************/
 960
 961             /* COULOMB ELECTROSTATICS */
 962             velec            = _mm256_mul_ps(qq21,rinv21);
 963             felec            = _mm256_mul_ps(velec,rinvsq21);
 964
 965             /* Update potential sum for this i atom from the interaction with this j atom. */
 966             velec            = _mm256_andnot_ps(dummy_mask,velec);
 967             velecsum         = _mm256_add_ps(velecsum,velec);
 968
 969             fscal            = felec;
 970
 971             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 972
 973             /* Calculate temporary vectorial force */
 974             tx               = _mm256_mul_ps(fscal,dx21);
 975             ty               = _mm256_mul_ps(fscal,dy21);
 976             tz               = _mm256_mul_ps(fscal,dz21);
 977
 978             /* Update vectorial force */
 979             fix2             = _mm256_add_ps(fix2,tx);
 980             fiy2             = _mm256_add_ps(fiy2,ty);
 981             fiz2             = _mm256_add_ps(fiz2,tz);
 982
 983             fjx1             = _mm256_add_ps(fjx1,tx);
 984             fjy1             = _mm256_add_ps(fjy1,ty);
 985             fjz1             = _mm256_add_ps(fjz1,tz);
 986
 987             /**************************
 988              * CALCULATE INTERACTIONS *
 989              **************************/
 990
 991             /* COULOMB ELECTROSTATICS */
 992             velec            = _mm256_mul_ps(qq22,rinv22);
 993             felec            = _mm256_mul_ps(velec,rinvsq22);
 994
 995             /* Update potential sum for this i atom from the interaction with this j atom. */
 996             velec            = _mm256_andnot_ps(dummy_mask,velec);
 997             velecsum         = _mm256_add_ps(velecsum,velec);
 998
 999             fscal            = felec;
1000
1001             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1002
1003             /* Calculate temporary vectorial force */
1004             tx               = _mm256_mul_ps(fscal,dx22);
1005             ty               = _mm256_mul_ps(fscal,dy22);
1006             tz               = _mm256_mul_ps(fscal,dz22);
1007
1008             /* Update vectorial force */
1009             fix2             = _mm256_add_ps(fix2,tx);
1010             fiy2             = _mm256_add_ps(fiy2,ty);
1011             fiz2             = _mm256_add_ps(fiz2,tz);
1012
1013             fjx2             = _mm256_add_ps(fjx2,tx);
1014             fjy2             = _mm256_add_ps(fjy2,ty);
1015             fjz2             = _mm256_add_ps(fjz2,tz);
1016
1017             /**************************
1018              * CALCULATE INTERACTIONS *
1019              **************************/
1020
1021             /* COULOMB ELECTROSTATICS */
1022             velec            = _mm256_mul_ps(qq23,rinv23);
1023             felec            = _mm256_mul_ps(velec,rinvsq23);
1024
1025             /* Update potential sum for this i atom from the interaction with this j atom. */
1026             velec            = _mm256_andnot_ps(dummy_mask,velec);
1027             velecsum         = _mm256_add_ps(velecsum,velec);
1028
1029             fscal            = felec;
1030
1031             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1032
1033             /* Calculate temporary vectorial force */
1034             tx               = _mm256_mul_ps(fscal,dx23);
1035             ty               = _mm256_mul_ps(fscal,dy23);
1036             tz               = _mm256_mul_ps(fscal,dz23);
1037
1038             /* Update vectorial force */
1039             fix2             = _mm256_add_ps(fix2,tx);
1040             fiy2             = _mm256_add_ps(fiy2,ty);
1041             fiz2             = _mm256_add_ps(fiz2,tz);
1042
1043             fjx3             = _mm256_add_ps(fjx3,tx);
1044             fjy3             = _mm256_add_ps(fjy3,ty);
1045             fjz3             = _mm256_add_ps(fjz3,tz);
1046
1047             /**************************
1048              * CALCULATE INTERACTIONS *
1049              **************************/
1050
1051             /* COULOMB ELECTROSTATICS */
1052             velec            = _mm256_mul_ps(qq31,rinv31);
1053             felec            = _mm256_mul_ps(velec,rinvsq31);
1054
1055             /* Update potential sum for this i atom from the interaction with this j atom. */
1056             velec            = _mm256_andnot_ps(dummy_mask,velec);
1057             velecsum         = _mm256_add_ps(velecsum,velec);
1058
1059             fscal            = felec;
1060
1061             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1062
1063             /* Calculate temporary vectorial force */
1064             tx               = _mm256_mul_ps(fscal,dx31);
1065             ty               = _mm256_mul_ps(fscal,dy31);
1066             tz               = _mm256_mul_ps(fscal,dz31);
1067
1068             /* Update vectorial force */
1069             fix3             = _mm256_add_ps(fix3,tx);
1070             fiy3             = _mm256_add_ps(fiy3,ty);
1071             fiz3             = _mm256_add_ps(fiz3,tz);
1072
1073             fjx1             = _mm256_add_ps(fjx1,tx);
1074             fjy1             = _mm256_add_ps(fjy1,ty);
1075             fjz1             = _mm256_add_ps(fjz1,tz);
1076
1077             /**************************
1078              * CALCULATE INTERACTIONS *
1079              **************************/
1080
1081             /* COULOMB ELECTROSTATICS */
1082             velec            = _mm256_mul_ps(qq32,rinv32);
1083             felec            = _mm256_mul_ps(velec,rinvsq32);
1084
1085             /* Update potential sum for this i atom from the interaction with this j atom. */
1086             velec            = _mm256_andnot_ps(dummy_mask,velec);
1087             velecsum         = _mm256_add_ps(velecsum,velec);
1088
1089             fscal            = felec;
1090
1091             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1092
1093             /* Calculate temporary vectorial force */
1094             tx               = _mm256_mul_ps(fscal,dx32);
1095             ty               = _mm256_mul_ps(fscal,dy32);
1096             tz               = _mm256_mul_ps(fscal,dz32);
1097
1098             /* Update vectorial force */
1099             fix3             = _mm256_add_ps(fix3,tx);
1100             fiy3             = _mm256_add_ps(fiy3,ty);
1101             fiz3             = _mm256_add_ps(fiz3,tz);
1102
1103             fjx2             = _mm256_add_ps(fjx2,tx);
1104             fjy2             = _mm256_add_ps(fjy2,ty);
1105             fjz2             = _mm256_add_ps(fjz2,tz);
1106
1107             /**************************
1108              * CALCULATE INTERACTIONS *
1109              **************************/
1110
1111             /* COULOMB ELECTROSTATICS */
1112             velec            = _mm256_mul_ps(qq33,rinv33);
1113             felec            = _mm256_mul_ps(velec,rinvsq33);
1114
1115             /* Update potential sum for this i atom from the interaction with this j atom. */
1116             velec            = _mm256_andnot_ps(dummy_mask,velec);
1117             velecsum         = _mm256_add_ps(velecsum,velec);
1118
1119             fscal            = felec;
1120
1121             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1122
1123             /* Calculate temporary vectorial force */
1124             tx               = _mm256_mul_ps(fscal,dx33);
1125             ty               = _mm256_mul_ps(fscal,dy33);
1126             tz               = _mm256_mul_ps(fscal,dz33);
1127
1128             /* Update vectorial force */
1129             fix3             = _mm256_add_ps(fix3,tx);
1130             fiy3             = _mm256_add_ps(fiy3,ty);
1131             fiz3             = _mm256_add_ps(fiz3,tz);
1132
1133             fjx3             = _mm256_add_ps(fjx3,tx);
1134             fjy3             = _mm256_add_ps(fjy3,ty);
1135             fjz3             = _mm256_add_ps(fjz3,tz);
1136
1137             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1138             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1139             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1140             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1141             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1142             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1143             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1144             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1145
1146             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1147                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1148                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1149
1150             /* Inner loop uses 303 flops */
1151         }
1152
1153         /* End of innermost loop */
1154
1155         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1156                                                  f+i_coord_offset,fshift+i_shift_offset);
1157
1158         ggid                        = gid[iidx];
1159         /* Update potential energies */
1160         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1161         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1162
1163         /* Increment number of inner iterations */
1164         inneriter                  += j_index_end - j_index_start;
1165
1166         /* Outer loop uses 26 flops */
1167     }
1168
1169     /* Increment number of outer iterations */
1170     outeriter        += nri;
1171
1172     /* Update outer/inner flops */
1173
1174     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*303);
1175 }
1176 /*
1177  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_single
1178  * Electrostatics interaction: Coulomb
1179  * VdW interaction:            CubicSplineTable
1180  * Geometry:                   Water4-Water4
1181  * Calculate force/pot:        Force
1182  */
1183 void
1184 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_single
1185                     (t_nblist                    * gmx_restrict       nlist,
1186                      rvec                        * gmx_restrict          xx,
1187                      rvec                        * gmx_restrict          ff,
1188                      t_forcerec                  * gmx_restrict          fr,
1189                      t_mdatoms                   * gmx_restrict     mdatoms,
1190                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1191                      t_nrnb                      * gmx_restrict        nrnb)
1192 {
1193     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1194      * just 0 for non-waters.
1195      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1196      * jnr indices corresponding to data put in the four positions in the SIMD register.
1197      */
1198     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1199     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1200     int              jnrA,jnrB,jnrC,jnrD;
1201     int              jnrE,jnrF,jnrG,jnrH;
1202     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1203     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1204     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1205     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1206     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1207     real             rcutoff_scalar;
1208     real             *shiftvec,*fshift,*x,*f;
1209     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1210     real             scratch[4*DIM];
1211     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1212     real *           vdwioffsetptr0;
1213     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1214     real *           vdwioffsetptr1;
1215     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1216     real *           vdwioffsetptr2;
1217     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1218     real *           vdwioffsetptr3;
1219     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1220     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1221     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1222     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1223     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1224     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1225     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1226     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1227     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1228     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1229     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1230     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1231     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1232     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1233     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1234     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1235     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1236     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1237     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1238     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1239     real             *charge;
1240     int              nvdwtype;
1241     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1242     int              *vdwtype;
1243     real             *vdwparam;
1244     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1245     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1246     __m256i          vfitab;
1247     __m128i          vfitab_lo,vfitab_hi;
1248     __m128i          ifour       = _mm_set1_epi32(4);
1249     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1250     real             *vftab;
1251     __m256           dummy_mask,cutoff_mask;
1252     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1253     __m256           one     = _mm256_set1_ps(1.0);
1254     __m256           two     = _mm256_set1_ps(2.0);
1255     x                = xx[0];
1256     f                = ff[0];
1257
1258     nri              = nlist->nri;
1259     iinr             = nlist->iinr;
1260     jindex           = nlist->jindex;
1261     jjnr             = nlist->jjnr;
1262     shiftidx         = nlist->shift;
1263     gid              = nlist->gid;
1264     shiftvec         = fr->shift_vec[0];
1265     fshift           = fr->fshift[0];
1266     facel            = _mm256_set1_ps(fr->epsfac);
1267     charge           = mdatoms->chargeA;
1268     nvdwtype         = fr->ntype;
1269     vdwparam         = fr->nbfp;
1270     vdwtype          = mdatoms->typeA;
1271
1272     vftab            = kernel_data->table_vdw->data;
1273     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
1274
1275     /* Setup water-specific parameters */
1276     inr              = nlist->iinr[0];
1277     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1278     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1279     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1280     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1281
1282     jq1              = _mm256_set1_ps(charge[inr+1]);
1283     jq2              = _mm256_set1_ps(charge[inr+2]);
1284     jq3              = _mm256_set1_ps(charge[inr+3]);
1285     vdwjidx0A        = 2*vdwtype[inr+0];
1286     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1287     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1288     qq11             = _mm256_mul_ps(iq1,jq1);
1289     qq12             = _mm256_mul_ps(iq1,jq2);
1290     qq13             = _mm256_mul_ps(iq1,jq3);
1291     qq21             = _mm256_mul_ps(iq2,jq1);
1292     qq22             = _mm256_mul_ps(iq2,jq2);
1293     qq23             = _mm256_mul_ps(iq2,jq3);
1294     qq31             = _mm256_mul_ps(iq3,jq1);
1295     qq32             = _mm256_mul_ps(iq3,jq2);
1296     qq33             = _mm256_mul_ps(iq3,jq3);
1297
1298     /* Avoid stupid compiler warnings */
1299     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1300     j_coord_offsetA = 0;
1301     j_coord_offsetB = 0;
1302     j_coord_offsetC = 0;
1303     j_coord_offsetD = 0;
1304     j_coord_offsetE = 0;
1305     j_coord_offsetF = 0;
1306     j_coord_offsetG = 0;
1307     j_coord_offsetH = 0;
1308
1309     outeriter        = 0;
1310     inneriter        = 0;
1311
1312     for(iidx=0;iidx<4*DIM;iidx++)
1313     {
1314         scratch[iidx] = 0.0;
1315     }
1316
1317     /* Start outer loop over neighborlists */
1318     for(iidx=0; iidx<nri; iidx++)
1319     {
1320         /* Load shift vector for this list */
1321         i_shift_offset   = DIM*shiftidx[iidx];
1322
1323         /* Load limits for loop over neighbors */
1324         j_index_start    = jindex[iidx];
1325         j_index_end      = jindex[iidx+1];
1326
1327         /* Get outer coordinate index */
1328         inr              = iinr[iidx];
1329         i_coord_offset   = DIM*inr;
1330
1331         /* Load i particle coords and add shift vector */
1332         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1333                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1334
1335         fix0             = _mm256_setzero_ps();
1336         fiy0             = _mm256_setzero_ps();
1337         fiz0             = _mm256_setzero_ps();
1338         fix1             = _mm256_setzero_ps();
1339         fiy1             = _mm256_setzero_ps();
1340         fiz1             = _mm256_setzero_ps();
1341         fix2             = _mm256_setzero_ps();
1342         fiy2             = _mm256_setzero_ps();
1343         fiz2             = _mm256_setzero_ps();
1344         fix3             = _mm256_setzero_ps();
1345         fiy3             = _mm256_setzero_ps();
1346         fiz3             = _mm256_setzero_ps();
1347
1348         /* Start inner kernel loop */
1349         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1350         {
1351
1352             /* Get j neighbor index, and coordinate index */
1353             jnrA             = jjnr[jidx];
1354             jnrB             = jjnr[jidx+1];
1355             jnrC             = jjnr[jidx+2];
1356             jnrD             = jjnr[jidx+3];
1357             jnrE             = jjnr[jidx+4];
1358             jnrF             = jjnr[jidx+5];
1359             jnrG             = jjnr[jidx+6];
1360             jnrH             = jjnr[jidx+7];
1361             j_coord_offsetA  = DIM*jnrA;
1362             j_coord_offsetB  = DIM*jnrB;
1363             j_coord_offsetC  = DIM*jnrC;
1364             j_coord_offsetD  = DIM*jnrD;
1365             j_coord_offsetE  = DIM*jnrE;
1366             j_coord_offsetF  = DIM*jnrF;
1367             j_coord_offsetG  = DIM*jnrG;
1368             j_coord_offsetH  = DIM*jnrH;
1369
1370             /* load j atom coordinates */
1371             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1372                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1373                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1374                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1375                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1376                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1377
1378             /* Calculate displacement vector */
1379             dx00             = _mm256_sub_ps(ix0,jx0);
1380             dy00             = _mm256_sub_ps(iy0,jy0);
1381             dz00             = _mm256_sub_ps(iz0,jz0);
1382             dx11             = _mm256_sub_ps(ix1,jx1);
1383             dy11             = _mm256_sub_ps(iy1,jy1);
1384             dz11             = _mm256_sub_ps(iz1,jz1);
1385             dx12             = _mm256_sub_ps(ix1,jx2);
1386             dy12             = _mm256_sub_ps(iy1,jy2);
1387             dz12             = _mm256_sub_ps(iz1,jz2);
1388             dx13             = _mm256_sub_ps(ix1,jx3);
1389             dy13             = _mm256_sub_ps(iy1,jy3);
1390             dz13             = _mm256_sub_ps(iz1,jz3);
1391             dx21             = _mm256_sub_ps(ix2,jx1);
1392             dy21             = _mm256_sub_ps(iy2,jy1);
1393             dz21             = _mm256_sub_ps(iz2,jz1);
1394             dx22             = _mm256_sub_ps(ix2,jx2);
1395             dy22             = _mm256_sub_ps(iy2,jy2);
1396             dz22             = _mm256_sub_ps(iz2,jz2);
1397             dx23             = _mm256_sub_ps(ix2,jx3);
1398             dy23             = _mm256_sub_ps(iy2,jy3);
1399             dz23             = _mm256_sub_ps(iz2,jz3);
1400             dx31             = _mm256_sub_ps(ix3,jx1);
1401             dy31             = _mm256_sub_ps(iy3,jy1);
1402             dz31             = _mm256_sub_ps(iz3,jz1);
1403             dx32             = _mm256_sub_ps(ix3,jx2);
1404             dy32             = _mm256_sub_ps(iy3,jy2);
1405             dz32             = _mm256_sub_ps(iz3,jz2);
1406             dx33             = _mm256_sub_ps(ix3,jx3);
1407             dy33             = _mm256_sub_ps(iy3,jy3);
1408             dz33             = _mm256_sub_ps(iz3,jz3);
1409
1410             /* Calculate squared distance and things based on it */
1411             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1412             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1413             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1414             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1415             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1416             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1417             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1418             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1419             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1420             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1421
1422             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1423             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1424             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1425             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1426             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1427             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1428             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1429             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1430             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1431             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1432
1433             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1434             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1435             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1436             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1437             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1438             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1439             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1440             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1441             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1442
1443             fjx0             = _mm256_setzero_ps();
1444             fjy0             = _mm256_setzero_ps();
1445             fjz0             = _mm256_setzero_ps();
1446             fjx1             = _mm256_setzero_ps();
1447             fjy1             = _mm256_setzero_ps();
1448             fjz1             = _mm256_setzero_ps();
1449             fjx2             = _mm256_setzero_ps();
1450             fjy2             = _mm256_setzero_ps();
1451             fjz2             = _mm256_setzero_ps();
1452             fjx3             = _mm256_setzero_ps();
1453             fjy3             = _mm256_setzero_ps();
1454             fjz3             = _mm256_setzero_ps();
1455
1456             /**************************
1457              * CALCULATE INTERACTIONS *
1458              **************************/
1459
1460             r00              = _mm256_mul_ps(rsq00,rinv00);
1461
1462             /* Calculate table index by multiplying r with table scale and truncate to integer */
1463             rt               = _mm256_mul_ps(r00,vftabscale);
1464             vfitab           = _mm256_cvttps_epi32(rt);
1465             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1466             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1467             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1468             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1469             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1470             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1471
1472             /* CUBIC SPLINE TABLE DISPERSION */
1473             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1474                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1475             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1476                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1477             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1478                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1479             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1480                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1481             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1482             Heps             = _mm256_mul_ps(vfeps,H);
1483             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1484             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1485             fvdw6            = _mm256_mul_ps(c6_00,FF);
1486
1487             /* CUBIC SPLINE TABLE REPULSION */
1488             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1489             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1490             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1491                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1492             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1493                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1494             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1495                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1496             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1497                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1498             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1499             Heps             = _mm256_mul_ps(vfeps,H);
1500             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1501             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1502             fvdw12           = _mm256_mul_ps(c12_00,FF);
1503             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1504
1505             fscal            = fvdw;
1506
1507             /* Calculate temporary vectorial force */
1508             tx               = _mm256_mul_ps(fscal,dx00);
1509             ty               = _mm256_mul_ps(fscal,dy00);
1510             tz               = _mm256_mul_ps(fscal,dz00);
1511
1512             /* Update vectorial force */
1513             fix0             = _mm256_add_ps(fix0,tx);
1514             fiy0             = _mm256_add_ps(fiy0,ty);
1515             fiz0             = _mm256_add_ps(fiz0,tz);
1516
1517             fjx0             = _mm256_add_ps(fjx0,tx);
1518             fjy0             = _mm256_add_ps(fjy0,ty);
1519             fjz0             = _mm256_add_ps(fjz0,tz);
1520
1521             /**************************
1522              * CALCULATE INTERACTIONS *
1523              **************************/
1524
1525             /* COULOMB ELECTROSTATICS */
1526             velec            = _mm256_mul_ps(qq11,rinv11);
1527             felec            = _mm256_mul_ps(velec,rinvsq11);
1528
1529             fscal            = felec;
1530
1531             /* Calculate temporary vectorial force */
1532             tx               = _mm256_mul_ps(fscal,dx11);
1533             ty               = _mm256_mul_ps(fscal,dy11);
1534             tz               = _mm256_mul_ps(fscal,dz11);
1535
1536             /* Update vectorial force */
1537             fix1             = _mm256_add_ps(fix1,tx);
1538             fiy1             = _mm256_add_ps(fiy1,ty);
1539             fiz1             = _mm256_add_ps(fiz1,tz);
1540
1541             fjx1             = _mm256_add_ps(fjx1,tx);
1542             fjy1             = _mm256_add_ps(fjy1,ty);
1543             fjz1             = _mm256_add_ps(fjz1,tz);
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* COULOMB ELECTROSTATICS */
1550             velec            = _mm256_mul_ps(qq12,rinv12);
1551             felec            = _mm256_mul_ps(velec,rinvsq12);
1552
1553             fscal            = felec;
1554
1555             /* Calculate temporary vectorial force */
1556             tx               = _mm256_mul_ps(fscal,dx12);
1557             ty               = _mm256_mul_ps(fscal,dy12);
1558             tz               = _mm256_mul_ps(fscal,dz12);
1559
1560             /* Update vectorial force */
1561             fix1             = _mm256_add_ps(fix1,tx);
1562             fiy1             = _mm256_add_ps(fiy1,ty);
1563             fiz1             = _mm256_add_ps(fiz1,tz);
1564
1565             fjx2             = _mm256_add_ps(fjx2,tx);
1566             fjy2             = _mm256_add_ps(fjy2,ty);
1567             fjz2             = _mm256_add_ps(fjz2,tz);
1568
1569             /**************************
1570              * CALCULATE INTERACTIONS *
1571              **************************/
1572
1573             /* COULOMB ELECTROSTATICS */
1574             velec            = _mm256_mul_ps(qq13,rinv13);
1575             felec            = _mm256_mul_ps(velec,rinvsq13);
1576
1577             fscal            = felec;
1578
1579             /* Calculate temporary vectorial force */
1580             tx               = _mm256_mul_ps(fscal,dx13);
1581             ty               = _mm256_mul_ps(fscal,dy13);
1582             tz               = _mm256_mul_ps(fscal,dz13);
1583
1584             /* Update vectorial force */
1585             fix1             = _mm256_add_ps(fix1,tx);
1586             fiy1             = _mm256_add_ps(fiy1,ty);
1587             fiz1             = _mm256_add_ps(fiz1,tz);
1588
1589             fjx3             = _mm256_add_ps(fjx3,tx);
1590             fjy3             = _mm256_add_ps(fjy3,ty);
1591             fjz3             = _mm256_add_ps(fjz3,tz);
1592
1593             /**************************
1594              * CALCULATE INTERACTIONS *
1595              **************************/
1596
1597             /* COULOMB ELECTROSTATICS */
1598             velec            = _mm256_mul_ps(qq21,rinv21);
1599             felec            = _mm256_mul_ps(velec,rinvsq21);
1600
1601             fscal            = felec;
1602
1603             /* Calculate temporary vectorial force */
1604             tx               = _mm256_mul_ps(fscal,dx21);
1605             ty               = _mm256_mul_ps(fscal,dy21);
1606             tz               = _mm256_mul_ps(fscal,dz21);
1607
1608             /* Update vectorial force */
1609             fix2             = _mm256_add_ps(fix2,tx);
1610             fiy2             = _mm256_add_ps(fiy2,ty);
1611             fiz2             = _mm256_add_ps(fiz2,tz);
1612
1613             fjx1             = _mm256_add_ps(fjx1,tx);
1614             fjy1             = _mm256_add_ps(fjy1,ty);
1615             fjz1             = _mm256_add_ps(fjz1,tz);
1616
1617             /**************************
1618              * CALCULATE INTERACTIONS *
1619              **************************/
1620
1621             /* COULOMB ELECTROSTATICS */
1622             velec            = _mm256_mul_ps(qq22,rinv22);
1623             felec            = _mm256_mul_ps(velec,rinvsq22);
1624
1625             fscal            = felec;
1626
1627             /* Calculate temporary vectorial force */
1628             tx               = _mm256_mul_ps(fscal,dx22);
1629             ty               = _mm256_mul_ps(fscal,dy22);
1630             tz               = _mm256_mul_ps(fscal,dz22);
1631
1632             /* Update vectorial force */
1633             fix2             = _mm256_add_ps(fix2,tx);
1634             fiy2             = _mm256_add_ps(fiy2,ty);
1635             fiz2             = _mm256_add_ps(fiz2,tz);
1636
1637             fjx2             = _mm256_add_ps(fjx2,tx);
1638             fjy2             = _mm256_add_ps(fjy2,ty);
1639             fjz2             = _mm256_add_ps(fjz2,tz);
1640
1641             /**************************
1642              * CALCULATE INTERACTIONS *
1643              **************************/
1644
1645             /* COULOMB ELECTROSTATICS */
1646             velec            = _mm256_mul_ps(qq23,rinv23);
1647             felec            = _mm256_mul_ps(velec,rinvsq23);
1648
1649             fscal            = felec;
1650
1651             /* Calculate temporary vectorial force */
1652             tx               = _mm256_mul_ps(fscal,dx23);
1653             ty               = _mm256_mul_ps(fscal,dy23);
1654             tz               = _mm256_mul_ps(fscal,dz23);
1655
1656             /* Update vectorial force */
1657             fix2             = _mm256_add_ps(fix2,tx);
1658             fiy2             = _mm256_add_ps(fiy2,ty);
1659             fiz2             = _mm256_add_ps(fiz2,tz);
1660
1661             fjx3             = _mm256_add_ps(fjx3,tx);
1662             fjy3             = _mm256_add_ps(fjy3,ty);
1663             fjz3             = _mm256_add_ps(fjz3,tz);
1664
1665             /**************************
1666              * CALCULATE INTERACTIONS *
1667              **************************/
1668
1669             /* COULOMB ELECTROSTATICS */
1670             velec            = _mm256_mul_ps(qq31,rinv31);
1671             felec            = _mm256_mul_ps(velec,rinvsq31);
1672
1673             fscal            = felec;
1674
1675             /* Calculate temporary vectorial force */
1676             tx               = _mm256_mul_ps(fscal,dx31);
1677             ty               = _mm256_mul_ps(fscal,dy31);
1678             tz               = _mm256_mul_ps(fscal,dz31);
1679
1680             /* Update vectorial force */
1681             fix3             = _mm256_add_ps(fix3,tx);
1682             fiy3             = _mm256_add_ps(fiy3,ty);
1683             fiz3             = _mm256_add_ps(fiz3,tz);
1684
1685             fjx1             = _mm256_add_ps(fjx1,tx);
1686             fjy1             = _mm256_add_ps(fjy1,ty);
1687             fjz1             = _mm256_add_ps(fjz1,tz);
1688
1689             /**************************
1690              * CALCULATE INTERACTIONS *
1691              **************************/
1692
1693             /* COULOMB ELECTROSTATICS */
1694             velec            = _mm256_mul_ps(qq32,rinv32);
1695             felec            = _mm256_mul_ps(velec,rinvsq32);
1696
1697             fscal            = felec;
1698
1699             /* Calculate temporary vectorial force */
1700             tx               = _mm256_mul_ps(fscal,dx32);
1701             ty               = _mm256_mul_ps(fscal,dy32);
1702             tz               = _mm256_mul_ps(fscal,dz32);
1703
1704             /* Update vectorial force */
1705             fix3             = _mm256_add_ps(fix3,tx);
1706             fiy3             = _mm256_add_ps(fiy3,ty);
1707             fiz3             = _mm256_add_ps(fiz3,tz);
1708
1709             fjx2             = _mm256_add_ps(fjx2,tx);
1710             fjy2             = _mm256_add_ps(fjy2,ty);
1711             fjz2             = _mm256_add_ps(fjz2,tz);
1712
1713             /**************************
1714              * CALCULATE INTERACTIONS *
1715              **************************/
1716
1717             /* COULOMB ELECTROSTATICS */
1718             velec            = _mm256_mul_ps(qq33,rinv33);
1719             felec            = _mm256_mul_ps(velec,rinvsq33);
1720
1721             fscal            = felec;
1722
1723             /* Calculate temporary vectorial force */
1724             tx               = _mm256_mul_ps(fscal,dx33);
1725             ty               = _mm256_mul_ps(fscal,dy33);
1726             tz               = _mm256_mul_ps(fscal,dz33);
1727
1728             /* Update vectorial force */
1729             fix3             = _mm256_add_ps(fix3,tx);
1730             fiy3             = _mm256_add_ps(fiy3,ty);
1731             fiz3             = _mm256_add_ps(fiz3,tz);
1732
1733             fjx3             = _mm256_add_ps(fjx3,tx);
1734             fjy3             = _mm256_add_ps(fjy3,ty);
1735             fjz3             = _mm256_add_ps(fjz3,tz);
1736
1737             fjptrA             = f+j_coord_offsetA;
1738             fjptrB             = f+j_coord_offsetB;
1739             fjptrC             = f+j_coord_offsetC;
1740             fjptrD             = f+j_coord_offsetD;
1741             fjptrE             = f+j_coord_offsetE;
1742             fjptrF             = f+j_coord_offsetF;
1743             fjptrG             = f+j_coord_offsetG;
1744             fjptrH             = f+j_coord_offsetH;
1745
1746             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1747                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1748                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1749
1750             /* Inner loop uses 285 flops */
1751         }
1752
1753         if(jidx<j_index_end)
1754         {
1755
1756             /* Get j neighbor index, and coordinate index */
1757             jnrlistA         = jjnr[jidx];
1758             jnrlistB         = jjnr[jidx+1];
1759             jnrlistC         = jjnr[jidx+2];
1760             jnrlistD         = jjnr[jidx+3];
1761             jnrlistE         = jjnr[jidx+4];
1762             jnrlistF         = jjnr[jidx+5];
1763             jnrlistG         = jjnr[jidx+6];
1764             jnrlistH         = jjnr[jidx+7];
1765             /* Sign of each element will be negative for non-real atoms.
1766              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1767              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1768              */
1769             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1770                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1771
1772             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1773             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1774             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1775             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1776             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1777             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1778             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1779             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1780             j_coord_offsetA  = DIM*jnrA;
1781             j_coord_offsetB  = DIM*jnrB;
1782             j_coord_offsetC  = DIM*jnrC;
1783             j_coord_offsetD  = DIM*jnrD;
1784             j_coord_offsetE  = DIM*jnrE;
1785             j_coord_offsetF  = DIM*jnrF;
1786             j_coord_offsetG  = DIM*jnrG;
1787             j_coord_offsetH  = DIM*jnrH;
1788
1789             /* load j atom coordinates */
1790             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1791                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1792                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1793                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1794                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1795                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1796
1797             /* Calculate displacement vector */
1798             dx00             = _mm256_sub_ps(ix0,jx0);
1799             dy00             = _mm256_sub_ps(iy0,jy0);
1800             dz00             = _mm256_sub_ps(iz0,jz0);
1801             dx11             = _mm256_sub_ps(ix1,jx1);
1802             dy11             = _mm256_sub_ps(iy1,jy1);
1803             dz11             = _mm256_sub_ps(iz1,jz1);
1804             dx12             = _mm256_sub_ps(ix1,jx2);
1805             dy12             = _mm256_sub_ps(iy1,jy2);
1806             dz12             = _mm256_sub_ps(iz1,jz2);
1807             dx13             = _mm256_sub_ps(ix1,jx3);
1808             dy13             = _mm256_sub_ps(iy1,jy3);
1809             dz13             = _mm256_sub_ps(iz1,jz3);
1810             dx21             = _mm256_sub_ps(ix2,jx1);
1811             dy21             = _mm256_sub_ps(iy2,jy1);
1812             dz21             = _mm256_sub_ps(iz2,jz1);
1813             dx22             = _mm256_sub_ps(ix2,jx2);
1814             dy22             = _mm256_sub_ps(iy2,jy2);
1815             dz22             = _mm256_sub_ps(iz2,jz2);
1816             dx23             = _mm256_sub_ps(ix2,jx3);
1817             dy23             = _mm256_sub_ps(iy2,jy3);
1818             dz23             = _mm256_sub_ps(iz2,jz3);
1819             dx31             = _mm256_sub_ps(ix3,jx1);
1820             dy31             = _mm256_sub_ps(iy3,jy1);
1821             dz31             = _mm256_sub_ps(iz3,jz1);
1822             dx32             = _mm256_sub_ps(ix3,jx2);
1823             dy32             = _mm256_sub_ps(iy3,jy2);
1824             dz32             = _mm256_sub_ps(iz3,jz2);
1825             dx33             = _mm256_sub_ps(ix3,jx3);
1826             dy33             = _mm256_sub_ps(iy3,jy3);
1827             dz33             = _mm256_sub_ps(iz3,jz3);
1828
1829             /* Calculate squared distance and things based on it */
1830             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1831             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1832             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1833             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1834             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1835             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1836             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1837             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1838             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1839             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1840
1841             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1842             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1843             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1844             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1845             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1846             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1847             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1848             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1849             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1850             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1851
1852             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1853             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1854             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1855             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1856             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1857             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1858             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1859             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1860             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1861
1862             fjx0             = _mm256_setzero_ps();
1863             fjy0             = _mm256_setzero_ps();
1864             fjz0             = _mm256_setzero_ps();
1865             fjx1             = _mm256_setzero_ps();
1866             fjy1             = _mm256_setzero_ps();
1867             fjz1             = _mm256_setzero_ps();
1868             fjx2             = _mm256_setzero_ps();
1869             fjy2             = _mm256_setzero_ps();
1870             fjz2             = _mm256_setzero_ps();
1871             fjx3             = _mm256_setzero_ps();
1872             fjy3             = _mm256_setzero_ps();
1873             fjz3             = _mm256_setzero_ps();
1874
1875             /**************************
1876              * CALCULATE INTERACTIONS *
1877              **************************/
1878
1879             r00              = _mm256_mul_ps(rsq00,rinv00);
1880             r00              = _mm256_andnot_ps(dummy_mask,r00);
1881
1882             /* Calculate table index by multiplying r with table scale and truncate to integer */
1883             rt               = _mm256_mul_ps(r00,vftabscale);
1884             vfitab           = _mm256_cvttps_epi32(rt);
1885             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1886             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1887             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1888             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1889             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1890             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1891
1892             /* CUBIC SPLINE TABLE DISPERSION */
1893             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1894                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1895             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1896                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1897             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1898                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1899             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1900                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1901             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1902             Heps             = _mm256_mul_ps(vfeps,H);
1903             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1904             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1905             fvdw6            = _mm256_mul_ps(c6_00,FF);
1906
1907             /* CUBIC SPLINE TABLE REPULSION */
1908             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1909             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1910             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1911                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1912             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1913                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1914             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1915                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1916             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1917                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1918             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1919             Heps             = _mm256_mul_ps(vfeps,H);
1920             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1921             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1922             fvdw12           = _mm256_mul_ps(c12_00,FF);
1923             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1924
1925             fscal            = fvdw;
1926
1927             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1928
1929             /* Calculate temporary vectorial force */
1930             tx               = _mm256_mul_ps(fscal,dx00);
1931             ty               = _mm256_mul_ps(fscal,dy00);
1932             tz               = _mm256_mul_ps(fscal,dz00);
1933
1934             /* Update vectorial force */
1935             fix0             = _mm256_add_ps(fix0,tx);
1936             fiy0             = _mm256_add_ps(fiy0,ty);
1937             fiz0             = _mm256_add_ps(fiz0,tz);
1938
1939             fjx0             = _mm256_add_ps(fjx0,tx);
1940             fjy0             = _mm256_add_ps(fjy0,ty);
1941             fjz0             = _mm256_add_ps(fjz0,tz);
1942
1943             /**************************
1944              * CALCULATE INTERACTIONS *
1945              **************************/
1946
1947             /* COULOMB ELECTROSTATICS */
1948             velec            = _mm256_mul_ps(qq11,rinv11);
1949             felec            = _mm256_mul_ps(velec,rinvsq11);
1950
1951             fscal            = felec;
1952
1953             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1954
1955             /* Calculate temporary vectorial force */
1956             tx               = _mm256_mul_ps(fscal,dx11);
1957             ty               = _mm256_mul_ps(fscal,dy11);
1958             tz               = _mm256_mul_ps(fscal,dz11);
1959
1960             /* Update vectorial force */
1961             fix1             = _mm256_add_ps(fix1,tx);
1962             fiy1             = _mm256_add_ps(fiy1,ty);
1963             fiz1             = _mm256_add_ps(fiz1,tz);
1964
1965             fjx1             = _mm256_add_ps(fjx1,tx);
1966             fjy1             = _mm256_add_ps(fjy1,ty);
1967             fjz1             = _mm256_add_ps(fjz1,tz);
1968
1969             /**************************
1970              * CALCULATE INTERACTIONS *
1971              **************************/
1972
1973             /* COULOMB ELECTROSTATICS */
1974             velec            = _mm256_mul_ps(qq12,rinv12);
1975             felec            = _mm256_mul_ps(velec,rinvsq12);
1976
1977             fscal            = felec;
1978
1979             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1980
1981             /* Calculate temporary vectorial force */
1982             tx               = _mm256_mul_ps(fscal,dx12);
1983             ty               = _mm256_mul_ps(fscal,dy12);
1984             tz               = _mm256_mul_ps(fscal,dz12);
1985
1986             /* Update vectorial force */
1987             fix1             = _mm256_add_ps(fix1,tx);
1988             fiy1             = _mm256_add_ps(fiy1,ty);
1989             fiz1             = _mm256_add_ps(fiz1,tz);
1990
1991             fjx2             = _mm256_add_ps(fjx2,tx);
1992             fjy2             = _mm256_add_ps(fjy2,ty);
1993             fjz2             = _mm256_add_ps(fjz2,tz);
1994
1995             /**************************
1996              * CALCULATE INTERACTIONS *
1997              **************************/
1998
1999             /* COULOMB ELECTROSTATICS */
2000             velec            = _mm256_mul_ps(qq13,rinv13);
2001             felec            = _mm256_mul_ps(velec,rinvsq13);
2002
2003             fscal            = felec;
2004
2005             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2006
2007             /* Calculate temporary vectorial force */
2008             tx               = _mm256_mul_ps(fscal,dx13);
2009             ty               = _mm256_mul_ps(fscal,dy13);
2010             tz               = _mm256_mul_ps(fscal,dz13);
2011
2012             /* Update vectorial force */
2013             fix1             = _mm256_add_ps(fix1,tx);
2014             fiy1             = _mm256_add_ps(fiy1,ty);
2015             fiz1             = _mm256_add_ps(fiz1,tz);
2016
2017             fjx3             = _mm256_add_ps(fjx3,tx);
2018             fjy3             = _mm256_add_ps(fjy3,ty);
2019             fjz3             = _mm256_add_ps(fjz3,tz);
2020
2021             /**************************
2022              * CALCULATE INTERACTIONS *
2023              **************************/
2024
2025             /* COULOMB ELECTROSTATICS */
2026             velec            = _mm256_mul_ps(qq21,rinv21);
2027             felec            = _mm256_mul_ps(velec,rinvsq21);
2028
2029             fscal            = felec;
2030
2031             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2032
2033             /* Calculate temporary vectorial force */
2034             tx               = _mm256_mul_ps(fscal,dx21);
2035             ty               = _mm256_mul_ps(fscal,dy21);
2036             tz               = _mm256_mul_ps(fscal,dz21);
2037
2038             /* Update vectorial force */
2039             fix2             = _mm256_add_ps(fix2,tx);
2040             fiy2             = _mm256_add_ps(fiy2,ty);
2041             fiz2             = _mm256_add_ps(fiz2,tz);
2042
2043             fjx1             = _mm256_add_ps(fjx1,tx);
2044             fjy1             = _mm256_add_ps(fjy1,ty);
2045             fjz1             = _mm256_add_ps(fjz1,tz);
2046
2047             /**************************
2048              * CALCULATE INTERACTIONS *
2049              **************************/
2050
2051             /* COULOMB ELECTROSTATICS */
2052             velec            = _mm256_mul_ps(qq22,rinv22);
2053             felec            = _mm256_mul_ps(velec,rinvsq22);
2054
2055             fscal            = felec;
2056
2057             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2058
2059             /* Calculate temporary vectorial force */
2060             tx               = _mm256_mul_ps(fscal,dx22);
2061             ty               = _mm256_mul_ps(fscal,dy22);
2062             tz               = _mm256_mul_ps(fscal,dz22);
2063
2064             /* Update vectorial force */
2065             fix2             = _mm256_add_ps(fix2,tx);
2066             fiy2             = _mm256_add_ps(fiy2,ty);
2067             fiz2             = _mm256_add_ps(fiz2,tz);
2068
2069             fjx2             = _mm256_add_ps(fjx2,tx);
2070             fjy2             = _mm256_add_ps(fjy2,ty);
2071             fjz2             = _mm256_add_ps(fjz2,tz);
2072
2073             /**************************
2074              * CALCULATE INTERACTIONS *
2075              **************************/
2076
2077             /* COULOMB ELECTROSTATICS */
2078             velec            = _mm256_mul_ps(qq23,rinv23);
2079             felec            = _mm256_mul_ps(velec,rinvsq23);
2080
2081             fscal            = felec;
2082
2083             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2084
2085             /* Calculate temporary vectorial force */
2086             tx               = _mm256_mul_ps(fscal,dx23);
2087             ty               = _mm256_mul_ps(fscal,dy23);
2088             tz               = _mm256_mul_ps(fscal,dz23);
2089
2090             /* Update vectorial force */
2091             fix2             = _mm256_add_ps(fix2,tx);
2092             fiy2             = _mm256_add_ps(fiy2,ty);
2093             fiz2             = _mm256_add_ps(fiz2,tz);
2094
2095             fjx3             = _mm256_add_ps(fjx3,tx);
2096             fjy3             = _mm256_add_ps(fjy3,ty);
2097             fjz3             = _mm256_add_ps(fjz3,tz);
2098
2099             /**************************
2100              * CALCULATE INTERACTIONS *
2101              **************************/
2102
2103             /* COULOMB ELECTROSTATICS */
2104             velec            = _mm256_mul_ps(qq31,rinv31);
2105             felec            = _mm256_mul_ps(velec,rinvsq31);
2106
2107             fscal            = felec;
2108
2109             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2110
2111             /* Calculate temporary vectorial force */
2112             tx               = _mm256_mul_ps(fscal,dx31);
2113             ty               = _mm256_mul_ps(fscal,dy31);
2114             tz               = _mm256_mul_ps(fscal,dz31);
2115
2116             /* Update vectorial force */
2117             fix3             = _mm256_add_ps(fix3,tx);
2118             fiy3             = _mm256_add_ps(fiy3,ty);
2119             fiz3             = _mm256_add_ps(fiz3,tz);
2120
2121             fjx1             = _mm256_add_ps(fjx1,tx);
2122             fjy1             = _mm256_add_ps(fjy1,ty);
2123             fjz1             = _mm256_add_ps(fjz1,tz);
2124
2125             /**************************
2126              * CALCULATE INTERACTIONS *
2127              **************************/
2128
2129             /* COULOMB ELECTROSTATICS */
2130             velec            = _mm256_mul_ps(qq32,rinv32);
2131             felec            = _mm256_mul_ps(velec,rinvsq32);
2132
2133             fscal            = felec;
2134
2135             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2136
2137             /* Calculate temporary vectorial force */
2138             tx               = _mm256_mul_ps(fscal,dx32);
2139             ty               = _mm256_mul_ps(fscal,dy32);
2140             tz               = _mm256_mul_ps(fscal,dz32);
2141
2142             /* Update vectorial force */
2143             fix3             = _mm256_add_ps(fix3,tx);
2144             fiy3             = _mm256_add_ps(fiy3,ty);
2145             fiz3             = _mm256_add_ps(fiz3,tz);
2146
2147             fjx2             = _mm256_add_ps(fjx2,tx);
2148             fjy2             = _mm256_add_ps(fjy2,ty);
2149             fjz2             = _mm256_add_ps(fjz2,tz);
2150
2151             /**************************
2152              * CALCULATE INTERACTIONS *
2153              **************************/
2154
2155             /* COULOMB ELECTROSTATICS */
2156             velec            = _mm256_mul_ps(qq33,rinv33);
2157             felec            = _mm256_mul_ps(velec,rinvsq33);
2158
2159             fscal            = felec;
2160
2161             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2162
2163             /* Calculate temporary vectorial force */
2164             tx               = _mm256_mul_ps(fscal,dx33);
2165             ty               = _mm256_mul_ps(fscal,dy33);
2166             tz               = _mm256_mul_ps(fscal,dz33);
2167
2168             /* Update vectorial force */
2169             fix3             = _mm256_add_ps(fix3,tx);
2170             fiy3             = _mm256_add_ps(fiy3,ty);
2171             fiz3             = _mm256_add_ps(fiz3,tz);
2172
2173             fjx3             = _mm256_add_ps(fjx3,tx);
2174             fjy3             = _mm256_add_ps(fjy3,ty);
2175             fjz3             = _mm256_add_ps(fjz3,tz);
2176
2177             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2178             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2179             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2180             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2181             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2182             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2183             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2184             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2185
2186             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2187                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2188                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2189
2190             /* Inner loop uses 286 flops */
2191         }
2192
2193         /* End of innermost loop */
2194
2195         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2196                                                  f+i_coord_offset,fshift+i_shift_offset);
2197
2198         /* Increment number of inner iterations */
2199         inneriter                  += j_index_end - j_index_start;
2200
2201         /* Outer loop uses 24 flops */
2202     }
2203
2204     /* Increment number of outer iterations */
2205     outeriter        += nri;
2206
2207     /* Update outer/inner flops */
2208
2209     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*286);
2210 }