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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_single.h"
  48 #include "kernelutil_x86_avx_256_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_256_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrE,jnrF,jnrG,jnrH;
  76     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  77     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  84     real             scratch[4*DIM];
  85     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 101     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 119     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 120     __m256i          vfitab;
 121     __m128i          vfitab_lo,vfitab_hi;
 122     __m128i          ifour       = _mm_set1_epi32(4);
 123     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 124     real             *vftab;
 125     __m256           dummy_mask,cutoff_mask;
 126     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 127     __m256           one     = _mm256_set1_ps(1.0);
 128     __m256           two     = _mm256_set1_ps(2.0);
 129     x                = xx[0];
 130     f                = ff[0];
 131
 132     nri              = nlist->nri;
 133     iinr             = nlist->iinr;
 134     jindex           = nlist->jindex;
 135     jjnr             = nlist->jjnr;
 136     shiftidx         = nlist->shift;
 137     gid              = nlist->gid;
 138     shiftvec         = fr->shift_vec[0];
 139     fshift           = fr->fshift[0];
 140     facel            = _mm256_set1_ps(fr->epsfac);
 141     charge           = mdatoms->chargeA;
 142     nvdwtype         = fr->ntype;
 143     vdwparam         = fr->nbfp;
 144     vdwtype          = mdatoms->typeA;
 145
 146     vftab            = kernel_data->table_vdw->data;
 147     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 148
 149     /* Setup water-specific parameters */
 150     inr              = nlist->iinr[0];
 151     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 152     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 153     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 154     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 155
 156     jq1              = _mm256_set1_ps(charge[inr+1]);
 157     jq2              = _mm256_set1_ps(charge[inr+2]);
 158     jq3              = _mm256_set1_ps(charge[inr+3]);
 159     vdwjidx0A        = 2*vdwtype[inr+0];
 160     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 161     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 162     qq11             = _mm256_mul_ps(iq1,jq1);
 163     qq12             = _mm256_mul_ps(iq1,jq2);
 164     qq13             = _mm256_mul_ps(iq1,jq3);
 165     qq21             = _mm256_mul_ps(iq2,jq1);
 166     qq22             = _mm256_mul_ps(iq2,jq2);
 167     qq23             = _mm256_mul_ps(iq2,jq3);
 168     qq31             = _mm256_mul_ps(iq3,jq1);
 169     qq32             = _mm256_mul_ps(iq3,jq2);
 170     qq33             = _mm256_mul_ps(iq3,jq3);
 171
 172     /* Avoid stupid compiler warnings */
 173     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 174     j_coord_offsetA = 0;
 175     j_coord_offsetB = 0;
 176     j_coord_offsetC = 0;
 177     j_coord_offsetD = 0;
 178     j_coord_offsetE = 0;
 179     j_coord_offsetF = 0;
 180     j_coord_offsetG = 0;
 181     j_coord_offsetH = 0;
 182
 183     outeriter        = 0;
 184     inneriter        = 0;
 185
 186     for(iidx=0;iidx<4*DIM;iidx++)
 187     {
 188         scratch[iidx] = 0.0;
 189     }
 190
 191     /* Start outer loop over neighborlists */
 192     for(iidx=0; iidx<nri; iidx++)
 193     {
 194         /* Load shift vector for this list */
 195         i_shift_offset   = DIM*shiftidx[iidx];
 196
 197         /* Load limits for loop over neighbors */
 198         j_index_start    = jindex[iidx];
 199         j_index_end      = jindex[iidx+1];
 200
 201         /* Get outer coordinate index */
 202         inr              = iinr[iidx];
 203         i_coord_offset   = DIM*inr;
 204
 205         /* Load i particle coords and add shift vector */
 206         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 207                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 208
 209         fix0             = _mm256_setzero_ps();
 210         fiy0             = _mm256_setzero_ps();
 211         fiz0             = _mm256_setzero_ps();
 212         fix1             = _mm256_setzero_ps();
 213         fiy1             = _mm256_setzero_ps();
 214         fiz1             = _mm256_setzero_ps();
 215         fix2             = _mm256_setzero_ps();
 216         fiy2             = _mm256_setzero_ps();
 217         fiz2             = _mm256_setzero_ps();
 218         fix3             = _mm256_setzero_ps();
 219         fiy3             = _mm256_setzero_ps();
 220         fiz3             = _mm256_setzero_ps();
 221
 222         /* Reset potential sums */
 223         velecsum         = _mm256_setzero_ps();
 224         vvdwsum          = _mm256_setzero_ps();
 225
 226         /* Start inner kernel loop */
 227         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 228         {
 229
 230             /* Get j neighbor index, and coordinate index */
 231             jnrA             = jjnr[jidx];
 232             jnrB             = jjnr[jidx+1];
 233             jnrC             = jjnr[jidx+2];
 234             jnrD             = jjnr[jidx+3];
 235             jnrE             = jjnr[jidx+4];
 236             jnrF             = jjnr[jidx+5];
 237             jnrG             = jjnr[jidx+6];
 238             jnrH             = jjnr[jidx+7];
 239             j_coord_offsetA  = DIM*jnrA;
 240             j_coord_offsetB  = DIM*jnrB;
 241             j_coord_offsetC  = DIM*jnrC;
 242             j_coord_offsetD  = DIM*jnrD;
 243             j_coord_offsetE  = DIM*jnrE;
 244             j_coord_offsetF  = DIM*jnrF;
 245             j_coord_offsetG  = DIM*jnrG;
 246             j_coord_offsetH  = DIM*jnrH;
 247
 248             /* load j atom coordinates */
 249             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 250                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 251                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 252                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 253                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 254                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 255
 256             /* Calculate displacement vector */
 257             dx00             = _mm256_sub_ps(ix0,jx0);
 258             dy00             = _mm256_sub_ps(iy0,jy0);
 259             dz00             = _mm256_sub_ps(iz0,jz0);
 260             dx11             = _mm256_sub_ps(ix1,jx1);
 261             dy11             = _mm256_sub_ps(iy1,jy1);
 262             dz11             = _mm256_sub_ps(iz1,jz1);
 263             dx12             = _mm256_sub_ps(ix1,jx2);
 264             dy12             = _mm256_sub_ps(iy1,jy2);
 265             dz12             = _mm256_sub_ps(iz1,jz2);
 266             dx13             = _mm256_sub_ps(ix1,jx3);
 267             dy13             = _mm256_sub_ps(iy1,jy3);
 268             dz13             = _mm256_sub_ps(iz1,jz3);
 269             dx21             = _mm256_sub_ps(ix2,jx1);
 270             dy21             = _mm256_sub_ps(iy2,jy1);
 271             dz21             = _mm256_sub_ps(iz2,jz1);
 272             dx22             = _mm256_sub_ps(ix2,jx2);
 273             dy22             = _mm256_sub_ps(iy2,jy2);
 274             dz22             = _mm256_sub_ps(iz2,jz2);
 275             dx23             = _mm256_sub_ps(ix2,jx3);
 276             dy23             = _mm256_sub_ps(iy2,jy3);
 277             dz23             = _mm256_sub_ps(iz2,jz3);
 278             dx31             = _mm256_sub_ps(ix3,jx1);
 279             dy31             = _mm256_sub_ps(iy3,jy1);
 280             dz31             = _mm256_sub_ps(iz3,jz1);
 281             dx32             = _mm256_sub_ps(ix3,jx2);
 282             dy32             = _mm256_sub_ps(iy3,jy2);
 283             dz32             = _mm256_sub_ps(iz3,jz2);
 284             dx33             = _mm256_sub_ps(ix3,jx3);
 285             dy33             = _mm256_sub_ps(iy3,jy3);
 286             dz33             = _mm256_sub_ps(iz3,jz3);
 287
 288             /* Calculate squared distance and things based on it */
 289             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 290             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 291             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 292             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 293             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 294             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 295             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 296             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 297             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 298             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 299
 300             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 301             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 302             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 303             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 304             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 305             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 306             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 307             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 308             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 309             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 310
 311             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 312             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 313             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 314             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 315             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 316             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 317             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 318             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 319             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 320
 321             fjx0             = _mm256_setzero_ps();
 322             fjy0             = _mm256_setzero_ps();
 323             fjz0             = _mm256_setzero_ps();
 324             fjx1             = _mm256_setzero_ps();
 325             fjy1             = _mm256_setzero_ps();
 326             fjz1             = _mm256_setzero_ps();
 327             fjx2             = _mm256_setzero_ps();
 328             fjy2             = _mm256_setzero_ps();
 329             fjz2             = _mm256_setzero_ps();
 330             fjx3             = _mm256_setzero_ps();
 331             fjy3             = _mm256_setzero_ps();
 332             fjz3             = _mm256_setzero_ps();
 333
 334             /**************************
 335              * CALCULATE INTERACTIONS *
 336              **************************/
 337
 338             r00              = _mm256_mul_ps(rsq00,rinv00);
 339
 340             /* Calculate table index by multiplying r with table scale and truncate to integer */
 341             rt               = _mm256_mul_ps(r00,vftabscale);
 342             vfitab           = _mm256_cvttps_epi32(rt);
 343             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 344             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 345             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 346             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 347             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 348             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 349
 350             /* CUBIC SPLINE TABLE DISPERSION */
 351             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 352                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 353             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 354                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 355             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 356                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 357             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 358                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 359             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 360             Heps             = _mm256_mul_ps(vfeps,H);
 361             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 362             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 363             vvdw6            = _mm256_mul_ps(c6_00,VV);
 364             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 365             fvdw6            = _mm256_mul_ps(c6_00,FF);
 366
 367             /* CUBIC SPLINE TABLE REPULSION */
 368             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 369             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 370             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 371                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 372             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 373                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 374             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 375                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 376             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 377                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 378             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 379             Heps             = _mm256_mul_ps(vfeps,H);
 380             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 381             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 382             vvdw12           = _mm256_mul_ps(c12_00,VV);
 383             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 384             fvdw12           = _mm256_mul_ps(c12_00,FF);
 385             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 386             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 387
 388             /* Update potential sum for this i atom from the interaction with this j atom. */
 389             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 390
 391             fscal            = fvdw;
 392
 393             /* Calculate temporary vectorial force */
 394             tx               = _mm256_mul_ps(fscal,dx00);
 395             ty               = _mm256_mul_ps(fscal,dy00);
 396             tz               = _mm256_mul_ps(fscal,dz00);
 397
 398             /* Update vectorial force */
 399             fix0             = _mm256_add_ps(fix0,tx);
 400             fiy0             = _mm256_add_ps(fiy0,ty);
 401             fiz0             = _mm256_add_ps(fiz0,tz);
 402
 403             fjx0             = _mm256_add_ps(fjx0,tx);
 404             fjy0             = _mm256_add_ps(fjy0,ty);
 405             fjz0             = _mm256_add_ps(fjz0,tz);
 406
 407             /**************************
 408              * CALCULATE INTERACTIONS *
 409              **************************/
 410
 411             /* COULOMB ELECTROSTATICS */
 412             velec            = _mm256_mul_ps(qq11,rinv11);
 413             felec            = _mm256_mul_ps(velec,rinvsq11);
 414
 415             /* Update potential sum for this i atom from the interaction with this j atom. */
 416             velecsum         = _mm256_add_ps(velecsum,velec);
 417
 418             fscal            = felec;
 419
 420             /* Calculate temporary vectorial force */
 421             tx               = _mm256_mul_ps(fscal,dx11);
 422             ty               = _mm256_mul_ps(fscal,dy11);
 423             tz               = _mm256_mul_ps(fscal,dz11);
 424
 425             /* Update vectorial force */
 426             fix1             = _mm256_add_ps(fix1,tx);
 427             fiy1             = _mm256_add_ps(fiy1,ty);
 428             fiz1             = _mm256_add_ps(fiz1,tz);
 429
 430             fjx1             = _mm256_add_ps(fjx1,tx);
 431             fjy1             = _mm256_add_ps(fjy1,ty);
 432             fjz1             = _mm256_add_ps(fjz1,tz);
 433
 434             /**************************
 435              * CALCULATE INTERACTIONS *
 436              **************************/
 437
 438             /* COULOMB ELECTROSTATICS */
 439             velec            = _mm256_mul_ps(qq12,rinv12);
 440             felec            = _mm256_mul_ps(velec,rinvsq12);
 441
 442             /* Update potential sum for this i atom from the interaction with this j atom. */
 443             velecsum         = _mm256_add_ps(velecsum,velec);
 444
 445             fscal            = felec;
 446
 447             /* Calculate temporary vectorial force */
 448             tx               = _mm256_mul_ps(fscal,dx12);
 449             ty               = _mm256_mul_ps(fscal,dy12);
 450             tz               = _mm256_mul_ps(fscal,dz12);
 451
 452             /* Update vectorial force */
 453             fix1             = _mm256_add_ps(fix1,tx);
 454             fiy1             = _mm256_add_ps(fiy1,ty);
 455             fiz1             = _mm256_add_ps(fiz1,tz);
 456
 457             fjx2             = _mm256_add_ps(fjx2,tx);
 458             fjy2             = _mm256_add_ps(fjy2,ty);
 459             fjz2             = _mm256_add_ps(fjz2,tz);
 460
 461             /**************************
 462              * CALCULATE INTERACTIONS *
 463              **************************/
 464
 465             /* COULOMB ELECTROSTATICS */
 466             velec            = _mm256_mul_ps(qq13,rinv13);
 467             felec            = _mm256_mul_ps(velec,rinvsq13);
 468
 469             /* Update potential sum for this i atom from the interaction with this j atom. */
 470             velecsum         = _mm256_add_ps(velecsum,velec);
 471
 472             fscal            = felec;
 473
 474             /* Calculate temporary vectorial force */
 475             tx               = _mm256_mul_ps(fscal,dx13);
 476             ty               = _mm256_mul_ps(fscal,dy13);
 477             tz               = _mm256_mul_ps(fscal,dz13);
 478
 479             /* Update vectorial force */
 480             fix1             = _mm256_add_ps(fix1,tx);
 481             fiy1             = _mm256_add_ps(fiy1,ty);
 482             fiz1             = _mm256_add_ps(fiz1,tz);
 483
 484             fjx3             = _mm256_add_ps(fjx3,tx);
 485             fjy3             = _mm256_add_ps(fjy3,ty);
 486             fjz3             = _mm256_add_ps(fjz3,tz);
 487
 488             /**************************
 489              * CALCULATE INTERACTIONS *
 490              **************************/
 491
 492             /* COULOMB ELECTROSTATICS */
 493             velec            = _mm256_mul_ps(qq21,rinv21);
 494             felec            = _mm256_mul_ps(velec,rinvsq21);
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             velecsum         = _mm256_add_ps(velecsum,velec);
 498
 499             fscal            = felec;
 500
 501             /* Calculate temporary vectorial force */
 502             tx               = _mm256_mul_ps(fscal,dx21);
 503             ty               = _mm256_mul_ps(fscal,dy21);
 504             tz               = _mm256_mul_ps(fscal,dz21);
 505
 506             /* Update vectorial force */
 507             fix2             = _mm256_add_ps(fix2,tx);
 508             fiy2             = _mm256_add_ps(fiy2,ty);
 509             fiz2             = _mm256_add_ps(fiz2,tz);
 510
 511             fjx1             = _mm256_add_ps(fjx1,tx);
 512             fjy1             = _mm256_add_ps(fjy1,ty);
 513             fjz1             = _mm256_add_ps(fjz1,tz);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             /* COULOMB ELECTROSTATICS */
 520             velec            = _mm256_mul_ps(qq22,rinv22);
 521             felec            = _mm256_mul_ps(velec,rinvsq22);
 522
 523             /* Update potential sum for this i atom from the interaction with this j atom. */
 524             velecsum         = _mm256_add_ps(velecsum,velec);
 525
 526             fscal            = felec;
 527
 528             /* Calculate temporary vectorial force */
 529             tx               = _mm256_mul_ps(fscal,dx22);
 530             ty               = _mm256_mul_ps(fscal,dy22);
 531             tz               = _mm256_mul_ps(fscal,dz22);
 532
 533             /* Update vectorial force */
 534             fix2             = _mm256_add_ps(fix2,tx);
 535             fiy2             = _mm256_add_ps(fiy2,ty);
 536             fiz2             = _mm256_add_ps(fiz2,tz);
 537
 538             fjx2             = _mm256_add_ps(fjx2,tx);
 539             fjy2             = _mm256_add_ps(fjy2,ty);
 540             fjz2             = _mm256_add_ps(fjz2,tz);
 541
 542             /**************************
 543              * CALCULATE INTERACTIONS *
 544              **************************/
 545
 546             /* COULOMB ELECTROSTATICS */
 547             velec            = _mm256_mul_ps(qq23,rinv23);
 548             felec            = _mm256_mul_ps(velec,rinvsq23);
 549
 550             /* Update potential sum for this i atom from the interaction with this j atom. */
 551             velecsum         = _mm256_add_ps(velecsum,velec);
 552
 553             fscal            = felec;
 554
 555             /* Calculate temporary vectorial force */
 556             tx               = _mm256_mul_ps(fscal,dx23);
 557             ty               = _mm256_mul_ps(fscal,dy23);
 558             tz               = _mm256_mul_ps(fscal,dz23);
 559
 560             /* Update vectorial force */
 561             fix2             = _mm256_add_ps(fix2,tx);
 562             fiy2             = _mm256_add_ps(fiy2,ty);
 563             fiz2             = _mm256_add_ps(fiz2,tz);
 564
 565             fjx3             = _mm256_add_ps(fjx3,tx);
 566             fjy3             = _mm256_add_ps(fjy3,ty);
 567             fjz3             = _mm256_add_ps(fjz3,tz);
 568
 569             /**************************
 570              * CALCULATE INTERACTIONS *
 571              **************************/
 572
 573             /* COULOMB ELECTROSTATICS */
 574             velec            = _mm256_mul_ps(qq31,rinv31);
 575             felec            = _mm256_mul_ps(velec,rinvsq31);
 576
 577             /* Update potential sum for this i atom from the interaction with this j atom. */
 578             velecsum         = _mm256_add_ps(velecsum,velec);
 579
 580             fscal            = felec;
 581
 582             /* Calculate temporary vectorial force */
 583             tx               = _mm256_mul_ps(fscal,dx31);
 584             ty               = _mm256_mul_ps(fscal,dy31);
 585             tz               = _mm256_mul_ps(fscal,dz31);
 586
 587             /* Update vectorial force */
 588             fix3             = _mm256_add_ps(fix3,tx);
 589             fiy3             = _mm256_add_ps(fiy3,ty);
 590             fiz3             = _mm256_add_ps(fiz3,tz);
 591
 592             fjx1             = _mm256_add_ps(fjx1,tx);
 593             fjy1             = _mm256_add_ps(fjy1,ty);
 594             fjz1             = _mm256_add_ps(fjz1,tz);
 595
 596             /**************************
 597              * CALCULATE INTERACTIONS *
 598              **************************/
 599
 600             /* COULOMB ELECTROSTATICS */
 601             velec            = _mm256_mul_ps(qq32,rinv32);
 602             felec            = _mm256_mul_ps(velec,rinvsq32);
 603
 604             /* Update potential sum for this i atom from the interaction with this j atom. */
 605             velecsum         = _mm256_add_ps(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             /* Calculate temporary vectorial force */
 610             tx               = _mm256_mul_ps(fscal,dx32);
 611             ty               = _mm256_mul_ps(fscal,dy32);
 612             tz               = _mm256_mul_ps(fscal,dz32);
 613
 614             /* Update vectorial force */
 615             fix3             = _mm256_add_ps(fix3,tx);
 616             fiy3             = _mm256_add_ps(fiy3,ty);
 617             fiz3             = _mm256_add_ps(fiz3,tz);
 618
 619             fjx2             = _mm256_add_ps(fjx2,tx);
 620             fjy2             = _mm256_add_ps(fjy2,ty);
 621             fjz2             = _mm256_add_ps(fjz2,tz);
 622
 623             /**************************
 624              * CALCULATE INTERACTIONS *
 625              **************************/
 626
 627             /* COULOMB ELECTROSTATICS */
 628             velec            = _mm256_mul_ps(qq33,rinv33);
 629             felec            = _mm256_mul_ps(velec,rinvsq33);
 630
 631             /* Update potential sum for this i atom from the interaction with this j atom. */
 632             velecsum         = _mm256_add_ps(velecsum,velec);
 633
 634             fscal            = felec;
 635
 636             /* Calculate temporary vectorial force */
 637             tx               = _mm256_mul_ps(fscal,dx33);
 638             ty               = _mm256_mul_ps(fscal,dy33);
 639             tz               = _mm256_mul_ps(fscal,dz33);
 640
 641             /* Update vectorial force */
 642             fix3             = _mm256_add_ps(fix3,tx);
 643             fiy3             = _mm256_add_ps(fiy3,ty);
 644             fiz3             = _mm256_add_ps(fiz3,tz);
 645
 646             fjx3             = _mm256_add_ps(fjx3,tx);
 647             fjy3             = _mm256_add_ps(fjy3,ty);
 648             fjz3             = _mm256_add_ps(fjz3,tz);
 649
 650             fjptrA             = f+j_coord_offsetA;
 651             fjptrB             = f+j_coord_offsetB;
 652             fjptrC             = f+j_coord_offsetC;
 653             fjptrD             = f+j_coord_offsetD;
 654             fjptrE             = f+j_coord_offsetE;
 655             fjptrF             = f+j_coord_offsetF;
 656             fjptrG             = f+j_coord_offsetG;
 657             fjptrH             = f+j_coord_offsetH;
 658
 659             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 660                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 661                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 662
 663             /* Inner loop uses 302 flops */
 664         }
 665
 666         if(jidx<j_index_end)
 667         {
 668
 669             /* Get j neighbor index, and coordinate index */
 670             jnrlistA         = jjnr[jidx];
 671             jnrlistB         = jjnr[jidx+1];
 672             jnrlistC         = jjnr[jidx+2];
 673             jnrlistD         = jjnr[jidx+3];
 674             jnrlistE         = jjnr[jidx+4];
 675             jnrlistF         = jjnr[jidx+5];
 676             jnrlistG         = jjnr[jidx+6];
 677             jnrlistH         = jjnr[jidx+7];
 678             /* Sign of each element will be negative for non-real atoms.
 679              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 680              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 681              */
 682             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 683                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 684
 685             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 686             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 687             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 688             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 689             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 690             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 691             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 692             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 693             j_coord_offsetA  = DIM*jnrA;
 694             j_coord_offsetB  = DIM*jnrB;
 695             j_coord_offsetC  = DIM*jnrC;
 696             j_coord_offsetD  = DIM*jnrD;
 697             j_coord_offsetE  = DIM*jnrE;
 698             j_coord_offsetF  = DIM*jnrF;
 699             j_coord_offsetG  = DIM*jnrG;
 700             j_coord_offsetH  = DIM*jnrH;
 701
 702             /* load j atom coordinates */
 703             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 704                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 705                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 706                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 707                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 708                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 709
 710             /* Calculate displacement vector */
 711             dx00             = _mm256_sub_ps(ix0,jx0);
 712             dy00             = _mm256_sub_ps(iy0,jy0);
 713             dz00             = _mm256_sub_ps(iz0,jz0);
 714             dx11             = _mm256_sub_ps(ix1,jx1);
 715             dy11             = _mm256_sub_ps(iy1,jy1);
 716             dz11             = _mm256_sub_ps(iz1,jz1);
 717             dx12             = _mm256_sub_ps(ix1,jx2);
 718             dy12             = _mm256_sub_ps(iy1,jy2);
 719             dz12             = _mm256_sub_ps(iz1,jz2);
 720             dx13             = _mm256_sub_ps(ix1,jx3);
 721             dy13             = _mm256_sub_ps(iy1,jy3);
 722             dz13             = _mm256_sub_ps(iz1,jz3);
 723             dx21             = _mm256_sub_ps(ix2,jx1);
 724             dy21             = _mm256_sub_ps(iy2,jy1);
 725             dz21             = _mm256_sub_ps(iz2,jz1);
 726             dx22             = _mm256_sub_ps(ix2,jx2);
 727             dy22             = _mm256_sub_ps(iy2,jy2);
 728             dz22             = _mm256_sub_ps(iz2,jz2);
 729             dx23             = _mm256_sub_ps(ix2,jx3);
 730             dy23             = _mm256_sub_ps(iy2,jy3);
 731             dz23             = _mm256_sub_ps(iz2,jz3);
 732             dx31             = _mm256_sub_ps(ix3,jx1);
 733             dy31             = _mm256_sub_ps(iy3,jy1);
 734             dz31             = _mm256_sub_ps(iz3,jz1);
 735             dx32             = _mm256_sub_ps(ix3,jx2);
 736             dy32             = _mm256_sub_ps(iy3,jy2);
 737             dz32             = _mm256_sub_ps(iz3,jz2);
 738             dx33             = _mm256_sub_ps(ix3,jx3);
 739             dy33             = _mm256_sub_ps(iy3,jy3);
 740             dz33             = _mm256_sub_ps(iz3,jz3);
 741
 742             /* Calculate squared distance and things based on it */
 743             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 744             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 745             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 746             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 747             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 748             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 749             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 750             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 751             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 752             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 753
 754             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 755             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 756             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 757             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 758             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 759             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 760             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 761             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 762             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 763             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 764
 765             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 766             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 767             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 768             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 769             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 770             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 771             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 772             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 773             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 774
 775             fjx0             = _mm256_setzero_ps();
 776             fjy0             = _mm256_setzero_ps();
 777             fjz0             = _mm256_setzero_ps();
 778             fjx1             = _mm256_setzero_ps();
 779             fjy1             = _mm256_setzero_ps();
 780             fjz1             = _mm256_setzero_ps();
 781             fjx2             = _mm256_setzero_ps();
 782             fjy2             = _mm256_setzero_ps();
 783             fjz2             = _mm256_setzero_ps();
 784             fjx3             = _mm256_setzero_ps();
 785             fjy3             = _mm256_setzero_ps();
 786             fjz3             = _mm256_setzero_ps();
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             r00              = _mm256_mul_ps(rsq00,rinv00);
 793             r00              = _mm256_andnot_ps(dummy_mask,r00);
 794
 795             /* Calculate table index by multiplying r with table scale and truncate to integer */
 796             rt               = _mm256_mul_ps(r00,vftabscale);
 797             vfitab           = _mm256_cvttps_epi32(rt);
 798             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 799             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 800             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 801             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 802             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 803             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 804
 805             /* CUBIC SPLINE TABLE DISPERSION */
 806             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 807                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 808             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 809                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 810             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 811                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 812             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 813                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 814             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 815             Heps             = _mm256_mul_ps(vfeps,H);
 816             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 817             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 818             vvdw6            = _mm256_mul_ps(c6_00,VV);
 819             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 820             fvdw6            = _mm256_mul_ps(c6_00,FF);
 821
 822             /* CUBIC SPLINE TABLE REPULSION */
 823             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 824             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 825             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 826                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 827             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 828                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 829             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 830                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 831             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 832                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 833             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 834             Heps             = _mm256_mul_ps(vfeps,H);
 835             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 836             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 837             vvdw12           = _mm256_mul_ps(c12_00,VV);
 838             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 839             fvdw12           = _mm256_mul_ps(c12_00,FF);
 840             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 841             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 842
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 845             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 846
 847             fscal            = fvdw;
 848
 849             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 850
 851             /* Calculate temporary vectorial force */
 852             tx               = _mm256_mul_ps(fscal,dx00);
 853             ty               = _mm256_mul_ps(fscal,dy00);
 854             tz               = _mm256_mul_ps(fscal,dz00);
 855
 856             /* Update vectorial force */
 857             fix0             = _mm256_add_ps(fix0,tx);
 858             fiy0             = _mm256_add_ps(fiy0,ty);
 859             fiz0             = _mm256_add_ps(fiz0,tz);
 860
 861             fjx0             = _mm256_add_ps(fjx0,tx);
 862             fjy0             = _mm256_add_ps(fjy0,ty);
 863             fjz0             = _mm256_add_ps(fjz0,tz);
 864
 865             /**************************
 866              * CALCULATE INTERACTIONS *
 867              **************************/
 868
 869             /* COULOMB ELECTROSTATICS */
 870             velec            = _mm256_mul_ps(qq11,rinv11);
 871             felec            = _mm256_mul_ps(velec,rinvsq11);
 872
 873             /* Update potential sum for this i atom from the interaction with this j atom. */
 874             velec            = _mm256_andnot_ps(dummy_mask,velec);
 875             velecsum         = _mm256_add_ps(velecsum,velec);
 876
 877             fscal            = felec;
 878
 879             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 880
 881             /* Calculate temporary vectorial force */
 882             tx               = _mm256_mul_ps(fscal,dx11);
 883             ty               = _mm256_mul_ps(fscal,dy11);
 884             tz               = _mm256_mul_ps(fscal,dz11);
 885
 886             /* Update vectorial force */
 887             fix1             = _mm256_add_ps(fix1,tx);
 888             fiy1             = _mm256_add_ps(fiy1,ty);
 889             fiz1             = _mm256_add_ps(fiz1,tz);
 890
 891             fjx1             = _mm256_add_ps(fjx1,tx);
 892             fjy1             = _mm256_add_ps(fjy1,ty);
 893             fjz1             = _mm256_add_ps(fjz1,tz);
 894
 895             /**************************
 896              * CALCULATE INTERACTIONS *
 897              **************************/
 898
 899             /* COULOMB ELECTROSTATICS */
 900             velec            = _mm256_mul_ps(qq12,rinv12);
 901             felec            = _mm256_mul_ps(velec,rinvsq12);
 902
 903             /* Update potential sum for this i atom from the interaction with this j atom. */
 904             velec            = _mm256_andnot_ps(dummy_mask,velec);
 905             velecsum         = _mm256_add_ps(velecsum,velec);
 906
 907             fscal            = felec;
 908
 909             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 910
 911             /* Calculate temporary vectorial force */
 912             tx               = _mm256_mul_ps(fscal,dx12);
 913             ty               = _mm256_mul_ps(fscal,dy12);
 914             tz               = _mm256_mul_ps(fscal,dz12);
 915
 916             /* Update vectorial force */
 917             fix1             = _mm256_add_ps(fix1,tx);
 918             fiy1             = _mm256_add_ps(fiy1,ty);
 919             fiz1             = _mm256_add_ps(fiz1,tz);
 920
 921             fjx2             = _mm256_add_ps(fjx2,tx);
 922             fjy2             = _mm256_add_ps(fjy2,ty);
 923             fjz2             = _mm256_add_ps(fjz2,tz);
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             /* COULOMB ELECTROSTATICS */
 930             velec            = _mm256_mul_ps(qq13,rinv13);
 931             felec            = _mm256_mul_ps(velec,rinvsq13);
 932
 933             /* Update potential sum for this i atom from the interaction with this j atom. */
 934             velec            = _mm256_andnot_ps(dummy_mask,velec);
 935             velecsum         = _mm256_add_ps(velecsum,velec);
 936
 937             fscal            = felec;
 938
 939             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 940
 941             /* Calculate temporary vectorial force */
 942             tx               = _mm256_mul_ps(fscal,dx13);
 943             ty               = _mm256_mul_ps(fscal,dy13);
 944             tz               = _mm256_mul_ps(fscal,dz13);
 945
 946             /* Update vectorial force */
 947             fix1             = _mm256_add_ps(fix1,tx);
 948             fiy1             = _mm256_add_ps(fiy1,ty);
 949             fiz1             = _mm256_add_ps(fiz1,tz);
 950
 951             fjx3             = _mm256_add_ps(fjx3,tx);
 952             fjy3             = _mm256_add_ps(fjy3,ty);
 953             fjz3             = _mm256_add_ps(fjz3,tz);
 954
 955             /**************************
 956              * CALCULATE INTERACTIONS *
 957              **************************/
 958
 959             /* COULOMB ELECTROSTATICS */
 960             velec            = _mm256_mul_ps(qq21,rinv21);
 961             felec            = _mm256_mul_ps(velec,rinvsq21);
 962
 963             /* Update potential sum for this i atom from the interaction with this j atom. */
 964             velec            = _mm256_andnot_ps(dummy_mask,velec);
 965             velecsum         = _mm256_add_ps(velecsum,velec);
 966
 967             fscal            = felec;
 968
 969             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 970
 971             /* Calculate temporary vectorial force */
 972             tx               = _mm256_mul_ps(fscal,dx21);
 973             ty               = _mm256_mul_ps(fscal,dy21);
 974             tz               = _mm256_mul_ps(fscal,dz21);
 975
 976             /* Update vectorial force */
 977             fix2             = _mm256_add_ps(fix2,tx);
 978             fiy2             = _mm256_add_ps(fiy2,ty);
 979             fiz2             = _mm256_add_ps(fiz2,tz);
 980
 981             fjx1             = _mm256_add_ps(fjx1,tx);
 982             fjy1             = _mm256_add_ps(fjy1,ty);
 983             fjz1             = _mm256_add_ps(fjz1,tz);
 984
 985             /**************************
 986              * CALCULATE INTERACTIONS *
 987              **************************/
 988
 989             /* COULOMB ELECTROSTATICS */
 990             velec            = _mm256_mul_ps(qq22,rinv22);
 991             felec            = _mm256_mul_ps(velec,rinvsq22);
 992
 993             /* Update potential sum for this i atom from the interaction with this j atom. */
 994             velec            = _mm256_andnot_ps(dummy_mask,velec);
 995             velecsum         = _mm256_add_ps(velecsum,velec);
 996
 997             fscal            = felec;
 998
 999             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1000
1001             /* Calculate temporary vectorial force */
1002             tx               = _mm256_mul_ps(fscal,dx22);
1003             ty               = _mm256_mul_ps(fscal,dy22);
1004             tz               = _mm256_mul_ps(fscal,dz22);
1005
1006             /* Update vectorial force */
1007             fix2             = _mm256_add_ps(fix2,tx);
1008             fiy2             = _mm256_add_ps(fiy2,ty);
1009             fiz2             = _mm256_add_ps(fiz2,tz);
1010
1011             fjx2             = _mm256_add_ps(fjx2,tx);
1012             fjy2             = _mm256_add_ps(fjy2,ty);
1013             fjz2             = _mm256_add_ps(fjz2,tz);
1014
1015             /**************************
1016              * CALCULATE INTERACTIONS *
1017              **************************/
1018
1019             /* COULOMB ELECTROSTATICS */
1020             velec            = _mm256_mul_ps(qq23,rinv23);
1021             felec            = _mm256_mul_ps(velec,rinvsq23);
1022
1023             /* Update potential sum for this i atom from the interaction with this j atom. */
1024             velec            = _mm256_andnot_ps(dummy_mask,velec);
1025             velecsum         = _mm256_add_ps(velecsum,velec);
1026
1027             fscal            = felec;
1028
1029             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1030
1031             /* Calculate temporary vectorial force */
1032             tx               = _mm256_mul_ps(fscal,dx23);
1033             ty               = _mm256_mul_ps(fscal,dy23);
1034             tz               = _mm256_mul_ps(fscal,dz23);
1035
1036             /* Update vectorial force */
1037             fix2             = _mm256_add_ps(fix2,tx);
1038             fiy2             = _mm256_add_ps(fiy2,ty);
1039             fiz2             = _mm256_add_ps(fiz2,tz);
1040
1041             fjx3             = _mm256_add_ps(fjx3,tx);
1042             fjy3             = _mm256_add_ps(fjy3,ty);
1043             fjz3             = _mm256_add_ps(fjz3,tz);
1044
1045             /**************************
1046              * CALCULATE INTERACTIONS *
1047              **************************/
1048
1049             /* COULOMB ELECTROSTATICS */
1050             velec            = _mm256_mul_ps(qq31,rinv31);
1051             felec            = _mm256_mul_ps(velec,rinvsq31);
1052
1053             /* Update potential sum for this i atom from the interaction with this j atom. */
1054             velec            = _mm256_andnot_ps(dummy_mask,velec);
1055             velecsum         = _mm256_add_ps(velecsum,velec);
1056
1057             fscal            = felec;
1058
1059             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1060
1061             /* Calculate temporary vectorial force */
1062             tx               = _mm256_mul_ps(fscal,dx31);
1063             ty               = _mm256_mul_ps(fscal,dy31);
1064             tz               = _mm256_mul_ps(fscal,dz31);
1065
1066             /* Update vectorial force */
1067             fix3             = _mm256_add_ps(fix3,tx);
1068             fiy3             = _mm256_add_ps(fiy3,ty);
1069             fiz3             = _mm256_add_ps(fiz3,tz);
1070
1071             fjx1             = _mm256_add_ps(fjx1,tx);
1072             fjy1             = _mm256_add_ps(fjy1,ty);
1073             fjz1             = _mm256_add_ps(fjz1,tz);
1074
1075             /**************************
1076              * CALCULATE INTERACTIONS *
1077              **************************/
1078
1079             /* COULOMB ELECTROSTATICS */
1080             velec            = _mm256_mul_ps(qq32,rinv32);
1081             felec            = _mm256_mul_ps(velec,rinvsq32);
1082
1083             /* Update potential sum for this i atom from the interaction with this j atom. */
1084             velec            = _mm256_andnot_ps(dummy_mask,velec);
1085             velecsum         = _mm256_add_ps(velecsum,velec);
1086
1087             fscal            = felec;
1088
1089             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1090
1091             /* Calculate temporary vectorial force */
1092             tx               = _mm256_mul_ps(fscal,dx32);
1093             ty               = _mm256_mul_ps(fscal,dy32);
1094             tz               = _mm256_mul_ps(fscal,dz32);
1095
1096             /* Update vectorial force */
1097             fix3             = _mm256_add_ps(fix3,tx);
1098             fiy3             = _mm256_add_ps(fiy3,ty);
1099             fiz3             = _mm256_add_ps(fiz3,tz);
1100
1101             fjx2             = _mm256_add_ps(fjx2,tx);
1102             fjy2             = _mm256_add_ps(fjy2,ty);
1103             fjz2             = _mm256_add_ps(fjz2,tz);
1104
1105             /**************************
1106              * CALCULATE INTERACTIONS *
1107              **************************/
1108
1109             /* COULOMB ELECTROSTATICS */
1110             velec            = _mm256_mul_ps(qq33,rinv33);
1111             felec            = _mm256_mul_ps(velec,rinvsq33);
1112
1113             /* Update potential sum for this i atom from the interaction with this j atom. */
1114             velec            = _mm256_andnot_ps(dummy_mask,velec);
1115             velecsum         = _mm256_add_ps(velecsum,velec);
1116
1117             fscal            = felec;
1118
1119             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1120
1121             /* Calculate temporary vectorial force */
1122             tx               = _mm256_mul_ps(fscal,dx33);
1123             ty               = _mm256_mul_ps(fscal,dy33);
1124             tz               = _mm256_mul_ps(fscal,dz33);
1125
1126             /* Update vectorial force */
1127             fix3             = _mm256_add_ps(fix3,tx);
1128             fiy3             = _mm256_add_ps(fiy3,ty);
1129             fiz3             = _mm256_add_ps(fiz3,tz);
1130
1131             fjx3             = _mm256_add_ps(fjx3,tx);
1132             fjy3             = _mm256_add_ps(fjy3,ty);
1133             fjz3             = _mm256_add_ps(fjz3,tz);
1134
1135             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1136             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1137             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1138             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1139             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1140             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1141             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1142             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1143
1144             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1145                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1146                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1147
1148             /* Inner loop uses 303 flops */
1149         }
1150
1151         /* End of innermost loop */
1152
1153         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1154                                                  f+i_coord_offset,fshift+i_shift_offset);
1155
1156         ggid                        = gid[iidx];
1157         /* Update potential energies */
1158         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1159         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1160
1161         /* Increment number of inner iterations */
1162         inneriter                  += j_index_end - j_index_start;
1163
1164         /* Outer loop uses 26 flops */
1165     }
1166
1167     /* Increment number of outer iterations */
1168     outeriter        += nri;
1169
1170     /* Update outer/inner flops */
1171
1172     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*303);
1173 }
1174 /*
1175  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_single
1176  * Electrostatics interaction: Coulomb
1177  * VdW interaction:            CubicSplineTable
1178  * Geometry:                   Water4-Water4
1179  * Calculate force/pot:        Force
1180  */
1181 void
1182 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_single
1183                     (t_nblist                    * gmx_restrict       nlist,
1184                      rvec                        * gmx_restrict          xx,
1185                      rvec                        * gmx_restrict          ff,
1186                      t_forcerec                  * gmx_restrict          fr,
1187                      t_mdatoms                   * gmx_restrict     mdatoms,
1188                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1189                      t_nrnb                      * gmx_restrict        nrnb)
1190 {
1191     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1192      * just 0 for non-waters.
1193      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1194      * jnr indices corresponding to data put in the four positions in the SIMD register.
1195      */
1196     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1197     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1198     int              jnrA,jnrB,jnrC,jnrD;
1199     int              jnrE,jnrF,jnrG,jnrH;
1200     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1201     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1202     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1203     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1204     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1205     real             rcutoff_scalar;
1206     real             *shiftvec,*fshift,*x,*f;
1207     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1208     real             scratch[4*DIM];
1209     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1210     real *           vdwioffsetptr0;
1211     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1212     real *           vdwioffsetptr1;
1213     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1214     real *           vdwioffsetptr2;
1215     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1216     real *           vdwioffsetptr3;
1217     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1218     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1219     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1220     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1221     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1222     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1223     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1224     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1225     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1226     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1227     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1228     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1229     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1230     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1231     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1232     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1233     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1234     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1235     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1236     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1237     real             *charge;
1238     int              nvdwtype;
1239     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1240     int              *vdwtype;
1241     real             *vdwparam;
1242     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1243     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1244     __m256i          vfitab;
1245     __m128i          vfitab_lo,vfitab_hi;
1246     __m128i          ifour       = _mm_set1_epi32(4);
1247     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1248     real             *vftab;
1249     __m256           dummy_mask,cutoff_mask;
1250     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1251     __m256           one     = _mm256_set1_ps(1.0);
1252     __m256           two     = _mm256_set1_ps(2.0);
1253     x                = xx[0];
1254     f                = ff[0];
1255
1256     nri              = nlist->nri;
1257     iinr             = nlist->iinr;
1258     jindex           = nlist->jindex;
1259     jjnr             = nlist->jjnr;
1260     shiftidx         = nlist->shift;
1261     gid              = nlist->gid;
1262     shiftvec         = fr->shift_vec[0];
1263     fshift           = fr->fshift[0];
1264     facel            = _mm256_set1_ps(fr->epsfac);
1265     charge           = mdatoms->chargeA;
1266     nvdwtype         = fr->ntype;
1267     vdwparam         = fr->nbfp;
1268     vdwtype          = mdatoms->typeA;
1269
1270     vftab            = kernel_data->table_vdw->data;
1271     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
1272
1273     /* Setup water-specific parameters */
1274     inr              = nlist->iinr[0];
1275     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1276     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1277     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1278     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1279
1280     jq1              = _mm256_set1_ps(charge[inr+1]);
1281     jq2              = _mm256_set1_ps(charge[inr+2]);
1282     jq3              = _mm256_set1_ps(charge[inr+3]);
1283     vdwjidx0A        = 2*vdwtype[inr+0];
1284     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1285     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1286     qq11             = _mm256_mul_ps(iq1,jq1);
1287     qq12             = _mm256_mul_ps(iq1,jq2);
1288     qq13             = _mm256_mul_ps(iq1,jq3);
1289     qq21             = _mm256_mul_ps(iq2,jq1);
1290     qq22             = _mm256_mul_ps(iq2,jq2);
1291     qq23             = _mm256_mul_ps(iq2,jq3);
1292     qq31             = _mm256_mul_ps(iq3,jq1);
1293     qq32             = _mm256_mul_ps(iq3,jq2);
1294     qq33             = _mm256_mul_ps(iq3,jq3);
1295
1296     /* Avoid stupid compiler warnings */
1297     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1298     j_coord_offsetA = 0;
1299     j_coord_offsetB = 0;
1300     j_coord_offsetC = 0;
1301     j_coord_offsetD = 0;
1302     j_coord_offsetE = 0;
1303     j_coord_offsetF = 0;
1304     j_coord_offsetG = 0;
1305     j_coord_offsetH = 0;
1306
1307     outeriter        = 0;
1308     inneriter        = 0;
1309
1310     for(iidx=0;iidx<4*DIM;iidx++)
1311     {
1312         scratch[iidx] = 0.0;
1313     }
1314
1315     /* Start outer loop over neighborlists */
1316     for(iidx=0; iidx<nri; iidx++)
1317     {
1318         /* Load shift vector for this list */
1319         i_shift_offset   = DIM*shiftidx[iidx];
1320
1321         /* Load limits for loop over neighbors */
1322         j_index_start    = jindex[iidx];
1323         j_index_end      = jindex[iidx+1];
1324
1325         /* Get outer coordinate index */
1326         inr              = iinr[iidx];
1327         i_coord_offset   = DIM*inr;
1328
1329         /* Load i particle coords and add shift vector */
1330         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1331                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1332
1333         fix0             = _mm256_setzero_ps();
1334         fiy0             = _mm256_setzero_ps();
1335         fiz0             = _mm256_setzero_ps();
1336         fix1             = _mm256_setzero_ps();
1337         fiy1             = _mm256_setzero_ps();
1338         fiz1             = _mm256_setzero_ps();
1339         fix2             = _mm256_setzero_ps();
1340         fiy2             = _mm256_setzero_ps();
1341         fiz2             = _mm256_setzero_ps();
1342         fix3             = _mm256_setzero_ps();
1343         fiy3             = _mm256_setzero_ps();
1344         fiz3             = _mm256_setzero_ps();
1345
1346         /* Start inner kernel loop */
1347         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1348         {
1349
1350             /* Get j neighbor index, and coordinate index */
1351             jnrA             = jjnr[jidx];
1352             jnrB             = jjnr[jidx+1];
1353             jnrC             = jjnr[jidx+2];
1354             jnrD             = jjnr[jidx+3];
1355             jnrE             = jjnr[jidx+4];
1356             jnrF             = jjnr[jidx+5];
1357             jnrG             = jjnr[jidx+6];
1358             jnrH             = jjnr[jidx+7];
1359             j_coord_offsetA  = DIM*jnrA;
1360             j_coord_offsetB  = DIM*jnrB;
1361             j_coord_offsetC  = DIM*jnrC;
1362             j_coord_offsetD  = DIM*jnrD;
1363             j_coord_offsetE  = DIM*jnrE;
1364             j_coord_offsetF  = DIM*jnrF;
1365             j_coord_offsetG  = DIM*jnrG;
1366             j_coord_offsetH  = DIM*jnrH;
1367
1368             /* load j atom coordinates */
1369             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1370                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1371                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1372                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1373                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1374                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1375
1376             /* Calculate displacement vector */
1377             dx00             = _mm256_sub_ps(ix0,jx0);
1378             dy00             = _mm256_sub_ps(iy0,jy0);
1379             dz00             = _mm256_sub_ps(iz0,jz0);
1380             dx11             = _mm256_sub_ps(ix1,jx1);
1381             dy11             = _mm256_sub_ps(iy1,jy1);
1382             dz11             = _mm256_sub_ps(iz1,jz1);
1383             dx12             = _mm256_sub_ps(ix1,jx2);
1384             dy12             = _mm256_sub_ps(iy1,jy2);
1385             dz12             = _mm256_sub_ps(iz1,jz2);
1386             dx13             = _mm256_sub_ps(ix1,jx3);
1387             dy13             = _mm256_sub_ps(iy1,jy3);
1388             dz13             = _mm256_sub_ps(iz1,jz3);
1389             dx21             = _mm256_sub_ps(ix2,jx1);
1390             dy21             = _mm256_sub_ps(iy2,jy1);
1391             dz21             = _mm256_sub_ps(iz2,jz1);
1392             dx22             = _mm256_sub_ps(ix2,jx2);
1393             dy22             = _mm256_sub_ps(iy2,jy2);
1394             dz22             = _mm256_sub_ps(iz2,jz2);
1395             dx23             = _mm256_sub_ps(ix2,jx3);
1396             dy23             = _mm256_sub_ps(iy2,jy3);
1397             dz23             = _mm256_sub_ps(iz2,jz3);
1398             dx31             = _mm256_sub_ps(ix3,jx1);
1399             dy31             = _mm256_sub_ps(iy3,jy1);
1400             dz31             = _mm256_sub_ps(iz3,jz1);
1401             dx32             = _mm256_sub_ps(ix3,jx2);
1402             dy32             = _mm256_sub_ps(iy3,jy2);
1403             dz32             = _mm256_sub_ps(iz3,jz2);
1404             dx33             = _mm256_sub_ps(ix3,jx3);
1405             dy33             = _mm256_sub_ps(iy3,jy3);
1406             dz33             = _mm256_sub_ps(iz3,jz3);
1407
1408             /* Calculate squared distance and things based on it */
1409             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1410             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1411             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1412             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1413             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1414             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1415             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1416             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1417             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1418             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1419
1420             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1421             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1422             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1423             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1424             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1425             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1426             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1427             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1428             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1429             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1430
1431             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1432             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1433             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1434             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1435             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1436             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1437             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1438             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1439             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1440
1441             fjx0             = _mm256_setzero_ps();
1442             fjy0             = _mm256_setzero_ps();
1443             fjz0             = _mm256_setzero_ps();
1444             fjx1             = _mm256_setzero_ps();
1445             fjy1             = _mm256_setzero_ps();
1446             fjz1             = _mm256_setzero_ps();
1447             fjx2             = _mm256_setzero_ps();
1448             fjy2             = _mm256_setzero_ps();
1449             fjz2             = _mm256_setzero_ps();
1450             fjx3             = _mm256_setzero_ps();
1451             fjy3             = _mm256_setzero_ps();
1452             fjz3             = _mm256_setzero_ps();
1453
1454             /**************************
1455              * CALCULATE INTERACTIONS *
1456              **************************/
1457
1458             r00              = _mm256_mul_ps(rsq00,rinv00);
1459
1460             /* Calculate table index by multiplying r with table scale and truncate to integer */
1461             rt               = _mm256_mul_ps(r00,vftabscale);
1462             vfitab           = _mm256_cvttps_epi32(rt);
1463             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1464             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1465             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1466             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1467             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1468             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1469
1470             /* CUBIC SPLINE TABLE DISPERSION */
1471             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1472                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1473             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1474                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1475             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1476                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1477             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1478                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1479             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1480             Heps             = _mm256_mul_ps(vfeps,H);
1481             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1482             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1483             fvdw6            = _mm256_mul_ps(c6_00,FF);
1484
1485             /* CUBIC SPLINE TABLE REPULSION */
1486             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1487             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1488             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1489                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1490             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1491                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1492             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1493                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1494             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1495                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1496             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1497             Heps             = _mm256_mul_ps(vfeps,H);
1498             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1499             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1500             fvdw12           = _mm256_mul_ps(c12_00,FF);
1501             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1502
1503             fscal            = fvdw;
1504
1505             /* Calculate temporary vectorial force */
1506             tx               = _mm256_mul_ps(fscal,dx00);
1507             ty               = _mm256_mul_ps(fscal,dy00);
1508             tz               = _mm256_mul_ps(fscal,dz00);
1509
1510             /* Update vectorial force */
1511             fix0             = _mm256_add_ps(fix0,tx);
1512             fiy0             = _mm256_add_ps(fiy0,ty);
1513             fiz0             = _mm256_add_ps(fiz0,tz);
1514
1515             fjx0             = _mm256_add_ps(fjx0,tx);
1516             fjy0             = _mm256_add_ps(fjy0,ty);
1517             fjz0             = _mm256_add_ps(fjz0,tz);
1518
1519             /**************************
1520              * CALCULATE INTERACTIONS *
1521              **************************/
1522
1523             /* COULOMB ELECTROSTATICS */
1524             velec            = _mm256_mul_ps(qq11,rinv11);
1525             felec            = _mm256_mul_ps(velec,rinvsq11);
1526
1527             fscal            = felec;
1528
1529             /* Calculate temporary vectorial force */
1530             tx               = _mm256_mul_ps(fscal,dx11);
1531             ty               = _mm256_mul_ps(fscal,dy11);
1532             tz               = _mm256_mul_ps(fscal,dz11);
1533
1534             /* Update vectorial force */
1535             fix1             = _mm256_add_ps(fix1,tx);
1536             fiy1             = _mm256_add_ps(fiy1,ty);
1537             fiz1             = _mm256_add_ps(fiz1,tz);
1538
1539             fjx1             = _mm256_add_ps(fjx1,tx);
1540             fjy1             = _mm256_add_ps(fjy1,ty);
1541             fjz1             = _mm256_add_ps(fjz1,tz);
1542
1543             /**************************
1544              * CALCULATE INTERACTIONS *
1545              **************************/
1546
1547             /* COULOMB ELECTROSTATICS */
1548             velec            = _mm256_mul_ps(qq12,rinv12);
1549             felec            = _mm256_mul_ps(velec,rinvsq12);
1550
1551             fscal            = felec;
1552
1553             /* Calculate temporary vectorial force */
1554             tx               = _mm256_mul_ps(fscal,dx12);
1555             ty               = _mm256_mul_ps(fscal,dy12);
1556             tz               = _mm256_mul_ps(fscal,dz12);
1557
1558             /* Update vectorial force */
1559             fix1             = _mm256_add_ps(fix1,tx);
1560             fiy1             = _mm256_add_ps(fiy1,ty);
1561             fiz1             = _mm256_add_ps(fiz1,tz);
1562
1563             fjx2             = _mm256_add_ps(fjx2,tx);
1564             fjy2             = _mm256_add_ps(fjy2,ty);
1565             fjz2             = _mm256_add_ps(fjz2,tz);
1566
1567             /**************************
1568              * CALCULATE INTERACTIONS *
1569              **************************/
1570
1571             /* COULOMB ELECTROSTATICS */
1572             velec            = _mm256_mul_ps(qq13,rinv13);
1573             felec            = _mm256_mul_ps(velec,rinvsq13);
1574
1575             fscal            = felec;
1576
1577             /* Calculate temporary vectorial force */
1578             tx               = _mm256_mul_ps(fscal,dx13);
1579             ty               = _mm256_mul_ps(fscal,dy13);
1580             tz               = _mm256_mul_ps(fscal,dz13);
1581
1582             /* Update vectorial force */
1583             fix1             = _mm256_add_ps(fix1,tx);
1584             fiy1             = _mm256_add_ps(fiy1,ty);
1585             fiz1             = _mm256_add_ps(fiz1,tz);
1586
1587             fjx3             = _mm256_add_ps(fjx3,tx);
1588             fjy3             = _mm256_add_ps(fjy3,ty);
1589             fjz3             = _mm256_add_ps(fjz3,tz);
1590
1591             /**************************
1592              * CALCULATE INTERACTIONS *
1593              **************************/
1594
1595             /* COULOMB ELECTROSTATICS */
1596             velec            = _mm256_mul_ps(qq21,rinv21);
1597             felec            = _mm256_mul_ps(velec,rinvsq21);
1598
1599             fscal            = felec;
1600
1601             /* Calculate temporary vectorial force */
1602             tx               = _mm256_mul_ps(fscal,dx21);
1603             ty               = _mm256_mul_ps(fscal,dy21);
1604             tz               = _mm256_mul_ps(fscal,dz21);
1605
1606             /* Update vectorial force */
1607             fix2             = _mm256_add_ps(fix2,tx);
1608             fiy2             = _mm256_add_ps(fiy2,ty);
1609             fiz2             = _mm256_add_ps(fiz2,tz);
1610
1611             fjx1             = _mm256_add_ps(fjx1,tx);
1612             fjy1             = _mm256_add_ps(fjy1,ty);
1613             fjz1             = _mm256_add_ps(fjz1,tz);
1614
1615             /**************************
1616              * CALCULATE INTERACTIONS *
1617              **************************/
1618
1619             /* COULOMB ELECTROSTATICS */
1620             velec            = _mm256_mul_ps(qq22,rinv22);
1621             felec            = _mm256_mul_ps(velec,rinvsq22);
1622
1623             fscal            = felec;
1624
1625             /* Calculate temporary vectorial force */
1626             tx               = _mm256_mul_ps(fscal,dx22);
1627             ty               = _mm256_mul_ps(fscal,dy22);
1628             tz               = _mm256_mul_ps(fscal,dz22);
1629
1630             /* Update vectorial force */
1631             fix2             = _mm256_add_ps(fix2,tx);
1632             fiy2             = _mm256_add_ps(fiy2,ty);
1633             fiz2             = _mm256_add_ps(fiz2,tz);
1634
1635             fjx2             = _mm256_add_ps(fjx2,tx);
1636             fjy2             = _mm256_add_ps(fjy2,ty);
1637             fjz2             = _mm256_add_ps(fjz2,tz);
1638
1639             /**************************
1640              * CALCULATE INTERACTIONS *
1641              **************************/
1642
1643             /* COULOMB ELECTROSTATICS */
1644             velec            = _mm256_mul_ps(qq23,rinv23);
1645             felec            = _mm256_mul_ps(velec,rinvsq23);
1646
1647             fscal            = felec;
1648
1649             /* Calculate temporary vectorial force */
1650             tx               = _mm256_mul_ps(fscal,dx23);
1651             ty               = _mm256_mul_ps(fscal,dy23);
1652             tz               = _mm256_mul_ps(fscal,dz23);
1653
1654             /* Update vectorial force */
1655             fix2             = _mm256_add_ps(fix2,tx);
1656             fiy2             = _mm256_add_ps(fiy2,ty);
1657             fiz2             = _mm256_add_ps(fiz2,tz);
1658
1659             fjx3             = _mm256_add_ps(fjx3,tx);
1660             fjy3             = _mm256_add_ps(fjy3,ty);
1661             fjz3             = _mm256_add_ps(fjz3,tz);
1662
1663             /**************************
1664              * CALCULATE INTERACTIONS *
1665              **************************/
1666
1667             /* COULOMB ELECTROSTATICS */
1668             velec            = _mm256_mul_ps(qq31,rinv31);
1669             felec            = _mm256_mul_ps(velec,rinvsq31);
1670
1671             fscal            = felec;
1672
1673             /* Calculate temporary vectorial force */
1674             tx               = _mm256_mul_ps(fscal,dx31);
1675             ty               = _mm256_mul_ps(fscal,dy31);
1676             tz               = _mm256_mul_ps(fscal,dz31);
1677
1678             /* Update vectorial force */
1679             fix3             = _mm256_add_ps(fix3,tx);
1680             fiy3             = _mm256_add_ps(fiy3,ty);
1681             fiz3             = _mm256_add_ps(fiz3,tz);
1682
1683             fjx1             = _mm256_add_ps(fjx1,tx);
1684             fjy1             = _mm256_add_ps(fjy1,ty);
1685             fjz1             = _mm256_add_ps(fjz1,tz);
1686
1687             /**************************
1688              * CALCULATE INTERACTIONS *
1689              **************************/
1690
1691             /* COULOMB ELECTROSTATICS */
1692             velec            = _mm256_mul_ps(qq32,rinv32);
1693             felec            = _mm256_mul_ps(velec,rinvsq32);
1694
1695             fscal            = felec;
1696
1697             /* Calculate temporary vectorial force */
1698             tx               = _mm256_mul_ps(fscal,dx32);
1699             ty               = _mm256_mul_ps(fscal,dy32);
1700             tz               = _mm256_mul_ps(fscal,dz32);
1701
1702             /* Update vectorial force */
1703             fix3             = _mm256_add_ps(fix3,tx);
1704             fiy3             = _mm256_add_ps(fiy3,ty);
1705             fiz3             = _mm256_add_ps(fiz3,tz);
1706
1707             fjx2             = _mm256_add_ps(fjx2,tx);
1708             fjy2             = _mm256_add_ps(fjy2,ty);
1709             fjz2             = _mm256_add_ps(fjz2,tz);
1710
1711             /**************************
1712              * CALCULATE INTERACTIONS *
1713              **************************/
1714
1715             /* COULOMB ELECTROSTATICS */
1716             velec            = _mm256_mul_ps(qq33,rinv33);
1717             felec            = _mm256_mul_ps(velec,rinvsq33);
1718
1719             fscal            = felec;
1720
1721             /* Calculate temporary vectorial force */
1722             tx               = _mm256_mul_ps(fscal,dx33);
1723             ty               = _mm256_mul_ps(fscal,dy33);
1724             tz               = _mm256_mul_ps(fscal,dz33);
1725
1726             /* Update vectorial force */
1727             fix3             = _mm256_add_ps(fix3,tx);
1728             fiy3             = _mm256_add_ps(fiy3,ty);
1729             fiz3             = _mm256_add_ps(fiz3,tz);
1730
1731             fjx3             = _mm256_add_ps(fjx3,tx);
1732             fjy3             = _mm256_add_ps(fjy3,ty);
1733             fjz3             = _mm256_add_ps(fjz3,tz);
1734
1735             fjptrA             = f+j_coord_offsetA;
1736             fjptrB             = f+j_coord_offsetB;
1737             fjptrC             = f+j_coord_offsetC;
1738             fjptrD             = f+j_coord_offsetD;
1739             fjptrE             = f+j_coord_offsetE;
1740             fjptrF             = f+j_coord_offsetF;
1741             fjptrG             = f+j_coord_offsetG;
1742             fjptrH             = f+j_coord_offsetH;
1743
1744             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1745                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1746                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1747
1748             /* Inner loop uses 285 flops */
1749         }
1750
1751         if(jidx<j_index_end)
1752         {
1753
1754             /* Get j neighbor index, and coordinate index */
1755             jnrlistA         = jjnr[jidx];
1756             jnrlistB         = jjnr[jidx+1];
1757             jnrlistC         = jjnr[jidx+2];
1758             jnrlistD         = jjnr[jidx+3];
1759             jnrlistE         = jjnr[jidx+4];
1760             jnrlistF         = jjnr[jidx+5];
1761             jnrlistG         = jjnr[jidx+6];
1762             jnrlistH         = jjnr[jidx+7];
1763             /* Sign of each element will be negative for non-real atoms.
1764              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1765              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1766              */
1767             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1768                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1769
1770             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1771             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1772             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1773             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1774             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1775             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1776             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1777             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1778             j_coord_offsetA  = DIM*jnrA;
1779             j_coord_offsetB  = DIM*jnrB;
1780             j_coord_offsetC  = DIM*jnrC;
1781             j_coord_offsetD  = DIM*jnrD;
1782             j_coord_offsetE  = DIM*jnrE;
1783             j_coord_offsetF  = DIM*jnrF;
1784             j_coord_offsetG  = DIM*jnrG;
1785             j_coord_offsetH  = DIM*jnrH;
1786
1787             /* load j atom coordinates */
1788             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1789                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1790                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1791                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1792                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1793                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1794
1795             /* Calculate displacement vector */
1796             dx00             = _mm256_sub_ps(ix0,jx0);
1797             dy00             = _mm256_sub_ps(iy0,jy0);
1798             dz00             = _mm256_sub_ps(iz0,jz0);
1799             dx11             = _mm256_sub_ps(ix1,jx1);
1800             dy11             = _mm256_sub_ps(iy1,jy1);
1801             dz11             = _mm256_sub_ps(iz1,jz1);
1802             dx12             = _mm256_sub_ps(ix1,jx2);
1803             dy12             = _mm256_sub_ps(iy1,jy2);
1804             dz12             = _mm256_sub_ps(iz1,jz2);
1805             dx13             = _mm256_sub_ps(ix1,jx3);
1806             dy13             = _mm256_sub_ps(iy1,jy3);
1807             dz13             = _mm256_sub_ps(iz1,jz3);
1808             dx21             = _mm256_sub_ps(ix2,jx1);
1809             dy21             = _mm256_sub_ps(iy2,jy1);
1810             dz21             = _mm256_sub_ps(iz2,jz1);
1811             dx22             = _mm256_sub_ps(ix2,jx2);
1812             dy22             = _mm256_sub_ps(iy2,jy2);
1813             dz22             = _mm256_sub_ps(iz2,jz2);
1814             dx23             = _mm256_sub_ps(ix2,jx3);
1815             dy23             = _mm256_sub_ps(iy2,jy3);
1816             dz23             = _mm256_sub_ps(iz2,jz3);
1817             dx31             = _mm256_sub_ps(ix3,jx1);
1818             dy31             = _mm256_sub_ps(iy3,jy1);
1819             dz31             = _mm256_sub_ps(iz3,jz1);
1820             dx32             = _mm256_sub_ps(ix3,jx2);
1821             dy32             = _mm256_sub_ps(iy3,jy2);
1822             dz32             = _mm256_sub_ps(iz3,jz2);
1823             dx33             = _mm256_sub_ps(ix3,jx3);
1824             dy33             = _mm256_sub_ps(iy3,jy3);
1825             dz33             = _mm256_sub_ps(iz3,jz3);
1826
1827             /* Calculate squared distance and things based on it */
1828             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1829             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1830             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1831             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1832             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1833             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1834             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1835             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1836             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1837             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1838
1839             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1840             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1841             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1842             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1843             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1844             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1845             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1846             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1847             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1848             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1849
1850             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1851             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1852             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1853             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1854             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1855             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1856             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1857             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1858             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1859
1860             fjx0             = _mm256_setzero_ps();
1861             fjy0             = _mm256_setzero_ps();
1862             fjz0             = _mm256_setzero_ps();
1863             fjx1             = _mm256_setzero_ps();
1864             fjy1             = _mm256_setzero_ps();
1865             fjz1             = _mm256_setzero_ps();
1866             fjx2             = _mm256_setzero_ps();
1867             fjy2             = _mm256_setzero_ps();
1868             fjz2             = _mm256_setzero_ps();
1869             fjx3             = _mm256_setzero_ps();
1870             fjy3             = _mm256_setzero_ps();
1871             fjz3             = _mm256_setzero_ps();
1872
1873             /**************************
1874              * CALCULATE INTERACTIONS *
1875              **************************/
1876
1877             r00              = _mm256_mul_ps(rsq00,rinv00);
1878             r00              = _mm256_andnot_ps(dummy_mask,r00);
1879
1880             /* Calculate table index by multiplying r with table scale and truncate to integer */
1881             rt               = _mm256_mul_ps(r00,vftabscale);
1882             vfitab           = _mm256_cvttps_epi32(rt);
1883             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1884             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1885             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1886             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1887             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1888             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1889
1890             /* CUBIC SPLINE TABLE DISPERSION */
1891             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1892                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1893             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1894                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1895             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1896                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1897             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1898                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1899             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1900             Heps             = _mm256_mul_ps(vfeps,H);
1901             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1902             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1903             fvdw6            = _mm256_mul_ps(c6_00,FF);
1904
1905             /* CUBIC SPLINE TABLE REPULSION */
1906             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1907             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1908             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1909                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1910             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1911                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1912             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1913                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1914             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1915                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1916             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1917             Heps             = _mm256_mul_ps(vfeps,H);
1918             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1919             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1920             fvdw12           = _mm256_mul_ps(c12_00,FF);
1921             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1922
1923             fscal            = fvdw;
1924
1925             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1926
1927             /* Calculate temporary vectorial force */
1928             tx               = _mm256_mul_ps(fscal,dx00);
1929             ty               = _mm256_mul_ps(fscal,dy00);
1930             tz               = _mm256_mul_ps(fscal,dz00);
1931
1932             /* Update vectorial force */
1933             fix0             = _mm256_add_ps(fix0,tx);
1934             fiy0             = _mm256_add_ps(fiy0,ty);
1935             fiz0             = _mm256_add_ps(fiz0,tz);
1936
1937             fjx0             = _mm256_add_ps(fjx0,tx);
1938             fjy0             = _mm256_add_ps(fjy0,ty);
1939             fjz0             = _mm256_add_ps(fjz0,tz);
1940
1941             /**************************
1942              * CALCULATE INTERACTIONS *
1943              **************************/
1944
1945             /* COULOMB ELECTROSTATICS */
1946             velec            = _mm256_mul_ps(qq11,rinv11);
1947             felec            = _mm256_mul_ps(velec,rinvsq11);
1948
1949             fscal            = felec;
1950
1951             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1952
1953             /* Calculate temporary vectorial force */
1954             tx               = _mm256_mul_ps(fscal,dx11);
1955             ty               = _mm256_mul_ps(fscal,dy11);
1956             tz               = _mm256_mul_ps(fscal,dz11);
1957
1958             /* Update vectorial force */
1959             fix1             = _mm256_add_ps(fix1,tx);
1960             fiy1             = _mm256_add_ps(fiy1,ty);
1961             fiz1             = _mm256_add_ps(fiz1,tz);
1962
1963             fjx1             = _mm256_add_ps(fjx1,tx);
1964             fjy1             = _mm256_add_ps(fjy1,ty);
1965             fjz1             = _mm256_add_ps(fjz1,tz);
1966
1967             /**************************
1968              * CALCULATE INTERACTIONS *
1969              **************************/
1970
1971             /* COULOMB ELECTROSTATICS */
1972             velec            = _mm256_mul_ps(qq12,rinv12);
1973             felec            = _mm256_mul_ps(velec,rinvsq12);
1974
1975             fscal            = felec;
1976
1977             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1978
1979             /* Calculate temporary vectorial force */
1980             tx               = _mm256_mul_ps(fscal,dx12);
1981             ty               = _mm256_mul_ps(fscal,dy12);
1982             tz               = _mm256_mul_ps(fscal,dz12);
1983
1984             /* Update vectorial force */
1985             fix1             = _mm256_add_ps(fix1,tx);
1986             fiy1             = _mm256_add_ps(fiy1,ty);
1987             fiz1             = _mm256_add_ps(fiz1,tz);
1988
1989             fjx2             = _mm256_add_ps(fjx2,tx);
1990             fjy2             = _mm256_add_ps(fjy2,ty);
1991             fjz2             = _mm256_add_ps(fjz2,tz);
1992
1993             /**************************
1994              * CALCULATE INTERACTIONS *
1995              **************************/
1996
1997             /* COULOMB ELECTROSTATICS */
1998             velec            = _mm256_mul_ps(qq13,rinv13);
1999             felec            = _mm256_mul_ps(velec,rinvsq13);
2000
2001             fscal            = felec;
2002
2003             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2004
2005             /* Calculate temporary vectorial force */
2006             tx               = _mm256_mul_ps(fscal,dx13);
2007             ty               = _mm256_mul_ps(fscal,dy13);
2008             tz               = _mm256_mul_ps(fscal,dz13);
2009
2010             /* Update vectorial force */
2011             fix1             = _mm256_add_ps(fix1,tx);
2012             fiy1             = _mm256_add_ps(fiy1,ty);
2013             fiz1             = _mm256_add_ps(fiz1,tz);
2014
2015             fjx3             = _mm256_add_ps(fjx3,tx);
2016             fjy3             = _mm256_add_ps(fjy3,ty);
2017             fjz3             = _mm256_add_ps(fjz3,tz);
2018
2019             /**************************
2020              * CALCULATE INTERACTIONS *
2021              **************************/
2022
2023             /* COULOMB ELECTROSTATICS */
2024             velec            = _mm256_mul_ps(qq21,rinv21);
2025             felec            = _mm256_mul_ps(velec,rinvsq21);
2026
2027             fscal            = felec;
2028
2029             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2030
2031             /* Calculate temporary vectorial force */
2032             tx               = _mm256_mul_ps(fscal,dx21);
2033             ty               = _mm256_mul_ps(fscal,dy21);
2034             tz               = _mm256_mul_ps(fscal,dz21);
2035
2036             /* Update vectorial force */
2037             fix2             = _mm256_add_ps(fix2,tx);
2038             fiy2             = _mm256_add_ps(fiy2,ty);
2039             fiz2             = _mm256_add_ps(fiz2,tz);
2040
2041             fjx1             = _mm256_add_ps(fjx1,tx);
2042             fjy1             = _mm256_add_ps(fjy1,ty);
2043             fjz1             = _mm256_add_ps(fjz1,tz);
2044
2045             /**************************
2046              * CALCULATE INTERACTIONS *
2047              **************************/
2048
2049             /* COULOMB ELECTROSTATICS */
2050             velec            = _mm256_mul_ps(qq22,rinv22);
2051             felec            = _mm256_mul_ps(velec,rinvsq22);
2052
2053             fscal            = felec;
2054
2055             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2056
2057             /* Calculate temporary vectorial force */
2058             tx               = _mm256_mul_ps(fscal,dx22);
2059             ty               = _mm256_mul_ps(fscal,dy22);
2060             tz               = _mm256_mul_ps(fscal,dz22);
2061
2062             /* Update vectorial force */
2063             fix2             = _mm256_add_ps(fix2,tx);
2064             fiy2             = _mm256_add_ps(fiy2,ty);
2065             fiz2             = _mm256_add_ps(fiz2,tz);
2066
2067             fjx2             = _mm256_add_ps(fjx2,tx);
2068             fjy2             = _mm256_add_ps(fjy2,ty);
2069             fjz2             = _mm256_add_ps(fjz2,tz);
2070
2071             /**************************
2072              * CALCULATE INTERACTIONS *
2073              **************************/
2074
2075             /* COULOMB ELECTROSTATICS */
2076             velec            = _mm256_mul_ps(qq23,rinv23);
2077             felec            = _mm256_mul_ps(velec,rinvsq23);
2078
2079             fscal            = felec;
2080
2081             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2082
2083             /* Calculate temporary vectorial force */
2084             tx               = _mm256_mul_ps(fscal,dx23);
2085             ty               = _mm256_mul_ps(fscal,dy23);
2086             tz               = _mm256_mul_ps(fscal,dz23);
2087
2088             /* Update vectorial force */
2089             fix2             = _mm256_add_ps(fix2,tx);
2090             fiy2             = _mm256_add_ps(fiy2,ty);
2091             fiz2             = _mm256_add_ps(fiz2,tz);
2092
2093             fjx3             = _mm256_add_ps(fjx3,tx);
2094             fjy3             = _mm256_add_ps(fjy3,ty);
2095             fjz3             = _mm256_add_ps(fjz3,tz);
2096
2097             /**************************
2098              * CALCULATE INTERACTIONS *
2099              **************************/
2100
2101             /* COULOMB ELECTROSTATICS */
2102             velec            = _mm256_mul_ps(qq31,rinv31);
2103             felec            = _mm256_mul_ps(velec,rinvsq31);
2104
2105             fscal            = felec;
2106
2107             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2108
2109             /* Calculate temporary vectorial force */
2110             tx               = _mm256_mul_ps(fscal,dx31);
2111             ty               = _mm256_mul_ps(fscal,dy31);
2112             tz               = _mm256_mul_ps(fscal,dz31);
2113
2114             /* Update vectorial force */
2115             fix3             = _mm256_add_ps(fix3,tx);
2116             fiy3             = _mm256_add_ps(fiy3,ty);
2117             fiz3             = _mm256_add_ps(fiz3,tz);
2118
2119             fjx1             = _mm256_add_ps(fjx1,tx);
2120             fjy1             = _mm256_add_ps(fjy1,ty);
2121             fjz1             = _mm256_add_ps(fjz1,tz);
2122
2123             /**************************
2124              * CALCULATE INTERACTIONS *
2125              **************************/
2126
2127             /* COULOMB ELECTROSTATICS */
2128             velec            = _mm256_mul_ps(qq32,rinv32);
2129             felec            = _mm256_mul_ps(velec,rinvsq32);
2130
2131             fscal            = felec;
2132
2133             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2134
2135             /* Calculate temporary vectorial force */
2136             tx               = _mm256_mul_ps(fscal,dx32);
2137             ty               = _mm256_mul_ps(fscal,dy32);
2138             tz               = _mm256_mul_ps(fscal,dz32);
2139
2140             /* Update vectorial force */
2141             fix3             = _mm256_add_ps(fix3,tx);
2142             fiy3             = _mm256_add_ps(fiy3,ty);
2143             fiz3             = _mm256_add_ps(fiz3,tz);
2144
2145             fjx2             = _mm256_add_ps(fjx2,tx);
2146             fjy2             = _mm256_add_ps(fjy2,ty);
2147             fjz2             = _mm256_add_ps(fjz2,tz);
2148
2149             /**************************
2150              * CALCULATE INTERACTIONS *
2151              **************************/
2152
2153             /* COULOMB ELECTROSTATICS */
2154             velec            = _mm256_mul_ps(qq33,rinv33);
2155             felec            = _mm256_mul_ps(velec,rinvsq33);
2156
2157             fscal            = felec;
2158
2159             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2160
2161             /* Calculate temporary vectorial force */
2162             tx               = _mm256_mul_ps(fscal,dx33);
2163             ty               = _mm256_mul_ps(fscal,dy33);
2164             tz               = _mm256_mul_ps(fscal,dz33);
2165
2166             /* Update vectorial force */
2167             fix3             = _mm256_add_ps(fix3,tx);
2168             fiy3             = _mm256_add_ps(fiy3,ty);
2169             fiz3             = _mm256_add_ps(fiz3,tz);
2170
2171             fjx3             = _mm256_add_ps(fjx3,tx);
2172             fjy3             = _mm256_add_ps(fjy3,ty);
2173             fjz3             = _mm256_add_ps(fjz3,tz);
2174
2175             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2176             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2177             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2178             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2179             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2180             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2181             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2182             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2183
2184             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2185                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2186                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2187
2188             /* Inner loop uses 286 flops */
2189         }
2190
2191         /* End of innermost loop */
2192
2193         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2194                                                  f+i_coord_offset,fshift+i_shift_offset);
2195
2196         /* Increment number of inner iterations */
2197         inneriter                  += j_index_end - j_index_start;
2198
2199         /* Outer loop uses 24 flops */
2200     }
2201
2202     /* Increment number of outer iterations */
2203     outeriter        += nri;
2204
2205     /* Update outer/inner flops */
2206
2207     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*286);
2208 }