src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEw_VdwLJEw_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_ElecEw_VdwLJEw_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LJEwald
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwLJEw_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     real *           vdwgridioffsetptr0;
  88     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  89     real *           vdwioffsetptr1;
  90     real *           vdwgridioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     real *           vdwgridioffsetptr2;
  94     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  95     real *           vdwioffsetptr3;
  96     real *           vdwgridioffsetptr3;
  97     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  98     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  99     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 100     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
 101     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
 102     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
 103     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 104     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 105     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 106     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 107     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 108     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 109     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 110     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 111     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 112     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 113     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 114     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 115     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 116     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 117     real             *charge;
 118     int              nvdwtype;
 119     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 120     int              *vdwtype;
 121     real             *vdwparam;
 122     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 123     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 124     __m256           c6grid_00;
 125     __m256           c6grid_11;
 126     __m256           c6grid_12;
 127     __m256           c6grid_13;
 128     __m256           c6grid_21;
 129     __m256           c6grid_22;
 130     __m256           c6grid_23;
 131     __m256           c6grid_31;
 132     __m256           c6grid_32;
 133     __m256           c6grid_33;
 134     real             *vdwgridparam;
 135     __m256           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 136     __m256           one_half  = _mm256_set1_ps(0.5);
 137     __m256           minus_one = _mm256_set1_ps(-1.0);
 138     __m256i          ewitab;
 139     __m128i          ewitab_lo,ewitab_hi;
 140     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 141     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 142     real             *ewtab;
 143     __m256           dummy_mask,cutoff_mask;
 144     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 145     __m256           one     = _mm256_set1_ps(1.0);
 146     __m256           two     = _mm256_set1_ps(2.0);
 147     x                = xx[0];
 148     f                = ff[0];
 149
 150     nri              = nlist->nri;
 151     iinr             = nlist->iinr;
 152     jindex           = nlist->jindex;
 153     jjnr             = nlist->jjnr;
 154     shiftidx         = nlist->shift;
 155     gid              = nlist->gid;
 156     shiftvec         = fr->shift_vec[0];
 157     fshift           = fr->fshift[0];
 158     facel            = _mm256_set1_ps(fr->epsfac);
 159     charge           = mdatoms->chargeA;
 160     nvdwtype         = fr->ntype;
 161     vdwparam         = fr->nbfp;
 162     vdwtype          = mdatoms->typeA;
 163     vdwgridparam     = fr->ljpme_c6grid;
 164     sh_lj_ewald      = _mm256_set1_ps(fr->ic->sh_lj_ewald);
 165     ewclj            = _mm256_set1_ps(fr->ewaldcoeff_lj);
 166     ewclj2           = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
 167
 168     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 169     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 170     beta2            = _mm256_mul_ps(beta,beta);
 171     beta3            = _mm256_mul_ps(beta,beta2);
 172
 173     ewtab            = fr->ic->tabq_coul_FDV0;
 174     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 175     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 176
 177     /* Setup water-specific parameters */
 178     inr              = nlist->iinr[0];
 179     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 180     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 181     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 182     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 183     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
 184
 185     jq1              = _mm256_set1_ps(charge[inr+1]);
 186     jq2              = _mm256_set1_ps(charge[inr+2]);
 187     jq3              = _mm256_set1_ps(charge[inr+3]);
 188     vdwjidx0A        = 2*vdwtype[inr+0];
 189     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 190     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 191     c6grid_00        = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
 192     qq11             = _mm256_mul_ps(iq1,jq1);
 193     qq12             = _mm256_mul_ps(iq1,jq2);
 194     qq13             = _mm256_mul_ps(iq1,jq3);
 195     qq21             = _mm256_mul_ps(iq2,jq1);
 196     qq22             = _mm256_mul_ps(iq2,jq2);
 197     qq23             = _mm256_mul_ps(iq2,jq3);
 198     qq31             = _mm256_mul_ps(iq3,jq1);
 199     qq32             = _mm256_mul_ps(iq3,jq2);
 200     qq33             = _mm256_mul_ps(iq3,jq3);
 201
 202     /* Avoid stupid compiler warnings */
 203     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 204     j_coord_offsetA = 0;
 205     j_coord_offsetB = 0;
 206     j_coord_offsetC = 0;
 207     j_coord_offsetD = 0;
 208     j_coord_offsetE = 0;
 209     j_coord_offsetF = 0;
 210     j_coord_offsetG = 0;
 211     j_coord_offsetH = 0;
 212
 213     outeriter        = 0;
 214     inneriter        = 0;
 215
 216     for(iidx=0;iidx<4*DIM;iidx++)
 217     {
 218         scratch[iidx] = 0.0;
 219     }
 220
 221     /* Start outer loop over neighborlists */
 222     for(iidx=0; iidx<nri; iidx++)
 223     {
 224         /* Load shift vector for this list */
 225         i_shift_offset   = DIM*shiftidx[iidx];
 226
 227         /* Load limits for loop over neighbors */
 228         j_index_start    = jindex[iidx];
 229         j_index_end      = jindex[iidx+1];
 230
 231         /* Get outer coordinate index */
 232         inr              = iinr[iidx];
 233         i_coord_offset   = DIM*inr;
 234
 235         /* Load i particle coords and add shift vector */
 236         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 237                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 238
 239         fix0             = _mm256_setzero_ps();
 240         fiy0             = _mm256_setzero_ps();
 241         fiz0             = _mm256_setzero_ps();
 242         fix1             = _mm256_setzero_ps();
 243         fiy1             = _mm256_setzero_ps();
 244         fiz1             = _mm256_setzero_ps();
 245         fix2             = _mm256_setzero_ps();
 246         fiy2             = _mm256_setzero_ps();
 247         fiz2             = _mm256_setzero_ps();
 248         fix3             = _mm256_setzero_ps();
 249         fiy3             = _mm256_setzero_ps();
 250         fiz3             = _mm256_setzero_ps();
 251
 252         /* Reset potential sums */
 253         velecsum         = _mm256_setzero_ps();
 254         vvdwsum          = _mm256_setzero_ps();
 255
 256         /* Start inner kernel loop */
 257         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 258         {
 259
 260             /* Get j neighbor index, and coordinate index */
 261             jnrA             = jjnr[jidx];
 262             jnrB             = jjnr[jidx+1];
 263             jnrC             = jjnr[jidx+2];
 264             jnrD             = jjnr[jidx+3];
 265             jnrE             = jjnr[jidx+4];
 266             jnrF             = jjnr[jidx+5];
 267             jnrG             = jjnr[jidx+6];
 268             jnrH             = jjnr[jidx+7];
 269             j_coord_offsetA  = DIM*jnrA;
 270             j_coord_offsetB  = DIM*jnrB;
 271             j_coord_offsetC  = DIM*jnrC;
 272             j_coord_offsetD  = DIM*jnrD;
 273             j_coord_offsetE  = DIM*jnrE;
 274             j_coord_offsetF  = DIM*jnrF;
 275             j_coord_offsetG  = DIM*jnrG;
 276             j_coord_offsetH  = DIM*jnrH;
 277
 278             /* load j atom coordinates */
 279             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 280                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 281                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 282                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 283                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 284                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 285
 286             /* Calculate displacement vector */
 287             dx00             = _mm256_sub_ps(ix0,jx0);
 288             dy00             = _mm256_sub_ps(iy0,jy0);
 289             dz00             = _mm256_sub_ps(iz0,jz0);
 290             dx11             = _mm256_sub_ps(ix1,jx1);
 291             dy11             = _mm256_sub_ps(iy1,jy1);
 292             dz11             = _mm256_sub_ps(iz1,jz1);
 293             dx12             = _mm256_sub_ps(ix1,jx2);
 294             dy12             = _mm256_sub_ps(iy1,jy2);
 295             dz12             = _mm256_sub_ps(iz1,jz2);
 296             dx13             = _mm256_sub_ps(ix1,jx3);
 297             dy13             = _mm256_sub_ps(iy1,jy3);
 298             dz13             = _mm256_sub_ps(iz1,jz3);
 299             dx21             = _mm256_sub_ps(ix2,jx1);
 300             dy21             = _mm256_sub_ps(iy2,jy1);
 301             dz21             = _mm256_sub_ps(iz2,jz1);
 302             dx22             = _mm256_sub_ps(ix2,jx2);
 303             dy22             = _mm256_sub_ps(iy2,jy2);
 304             dz22             = _mm256_sub_ps(iz2,jz2);
 305             dx23             = _mm256_sub_ps(ix2,jx3);
 306             dy23             = _mm256_sub_ps(iy2,jy3);
 307             dz23             = _mm256_sub_ps(iz2,jz3);
 308             dx31             = _mm256_sub_ps(ix3,jx1);
 309             dy31             = _mm256_sub_ps(iy3,jy1);
 310             dz31             = _mm256_sub_ps(iz3,jz1);
 311             dx32             = _mm256_sub_ps(ix3,jx2);
 312             dy32             = _mm256_sub_ps(iy3,jy2);
 313             dz32             = _mm256_sub_ps(iz3,jz2);
 314             dx33             = _mm256_sub_ps(ix3,jx3);
 315             dy33             = _mm256_sub_ps(iy3,jy3);
 316             dz33             = _mm256_sub_ps(iz3,jz3);
 317
 318             /* Calculate squared distance and things based on it */
 319             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 320             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 321             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 322             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 323             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 324             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 325             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 326             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 327             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 328             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 329
 330             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 331             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 332             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 333             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 334             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 335             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 336             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 337             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 338             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 339             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 340
 341             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 342             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 343             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 344             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 345             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 346             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 347             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 348             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 349             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 350             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 351
 352             fjx0             = _mm256_setzero_ps();
 353             fjy0             = _mm256_setzero_ps();
 354             fjz0             = _mm256_setzero_ps();
 355             fjx1             = _mm256_setzero_ps();
 356             fjy1             = _mm256_setzero_ps();
 357             fjz1             = _mm256_setzero_ps();
 358             fjx2             = _mm256_setzero_ps();
 359             fjy2             = _mm256_setzero_ps();
 360             fjz2             = _mm256_setzero_ps();
 361             fjx3             = _mm256_setzero_ps();
 362             fjy3             = _mm256_setzero_ps();
 363             fjz3             = _mm256_setzero_ps();
 364
 365             /**************************
 366              * CALCULATE INTERACTIONS *
 367              **************************/
 368
 369             r00              = _mm256_mul_ps(rsq00,rinv00);
 370
 371             /* Analytical LJ-PME */
 372             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 373             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
 374             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
 375             exponent         = gmx_simd_exp_r(ewcljrsq);
 376             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 377             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
 378             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 379             vvdw6            = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
 380             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 381             vvdw             = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
 382             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 383             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
 384
 385             /* Update potential sum for this i atom from the interaction with this j atom. */
 386             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 387
 388             fscal            = fvdw;
 389
 390             /* Calculate temporary vectorial force */
 391             tx               = _mm256_mul_ps(fscal,dx00);
 392             ty               = _mm256_mul_ps(fscal,dy00);
 393             tz               = _mm256_mul_ps(fscal,dz00);
 394
 395             /* Update vectorial force */
 396             fix0             = _mm256_add_ps(fix0,tx);
 397             fiy0             = _mm256_add_ps(fiy0,ty);
 398             fiz0             = _mm256_add_ps(fiz0,tz);
 399
 400             fjx0             = _mm256_add_ps(fjx0,tx);
 401             fjy0             = _mm256_add_ps(fjy0,ty);
 402             fjz0             = _mm256_add_ps(fjz0,tz);
 403
 404             /**************************
 405              * CALCULATE INTERACTIONS *
 406              **************************/
 407
 408             r11              = _mm256_mul_ps(rsq11,rinv11);
 409
 410             /* EWALD ELECTROSTATICS */
 411
 412             /* Analytical PME correction */
 413             zeta2            = _mm256_mul_ps(beta2,rsq11);
 414             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 415             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 416             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 417             felec            = _mm256_mul_ps(qq11,felec);
 418             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 419             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 420             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 421             velec            = _mm256_mul_ps(qq11,velec);
 422
 423             /* Update potential sum for this i atom from the interaction with this j atom. */
 424             velecsum         = _mm256_add_ps(velecsum,velec);
 425
 426             fscal            = felec;
 427
 428             /* Calculate temporary vectorial force */
 429             tx               = _mm256_mul_ps(fscal,dx11);
 430             ty               = _mm256_mul_ps(fscal,dy11);
 431             tz               = _mm256_mul_ps(fscal,dz11);
 432
 433             /* Update vectorial force */
 434             fix1             = _mm256_add_ps(fix1,tx);
 435             fiy1             = _mm256_add_ps(fiy1,ty);
 436             fiz1             = _mm256_add_ps(fiz1,tz);
 437
 438             fjx1             = _mm256_add_ps(fjx1,tx);
 439             fjy1             = _mm256_add_ps(fjy1,ty);
 440             fjz1             = _mm256_add_ps(fjz1,tz);
 441
 442             /**************************
 443              * CALCULATE INTERACTIONS *
 444              **************************/
 445
 446             r12              = _mm256_mul_ps(rsq12,rinv12);
 447
 448             /* EWALD ELECTROSTATICS */
 449
 450             /* Analytical PME correction */
 451             zeta2            = _mm256_mul_ps(beta2,rsq12);
 452             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 453             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 454             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 455             felec            = _mm256_mul_ps(qq12,felec);
 456             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 457             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 458             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 459             velec            = _mm256_mul_ps(qq12,velec);
 460
 461             /* Update potential sum for this i atom from the interaction with this j atom. */
 462             velecsum         = _mm256_add_ps(velecsum,velec);
 463
 464             fscal            = felec;
 465
 466             /* Calculate temporary vectorial force */
 467             tx               = _mm256_mul_ps(fscal,dx12);
 468             ty               = _mm256_mul_ps(fscal,dy12);
 469             tz               = _mm256_mul_ps(fscal,dz12);
 470
 471             /* Update vectorial force */
 472             fix1             = _mm256_add_ps(fix1,tx);
 473             fiy1             = _mm256_add_ps(fiy1,ty);
 474             fiz1             = _mm256_add_ps(fiz1,tz);
 475
 476             fjx2             = _mm256_add_ps(fjx2,tx);
 477             fjy2             = _mm256_add_ps(fjy2,ty);
 478             fjz2             = _mm256_add_ps(fjz2,tz);
 479
 480             /**************************
 481              * CALCULATE INTERACTIONS *
 482              **************************/
 483
 484             r13              = _mm256_mul_ps(rsq13,rinv13);
 485
 486             /* EWALD ELECTROSTATICS */
 487
 488             /* Analytical PME correction */
 489             zeta2            = _mm256_mul_ps(beta2,rsq13);
 490             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
 491             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 492             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 493             felec            = _mm256_mul_ps(qq13,felec);
 494             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 495             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 496             velec            = _mm256_sub_ps(rinv13,pmecorrV);
 497             velec            = _mm256_mul_ps(qq13,velec);
 498
 499             /* Update potential sum for this i atom from the interaction with this j atom. */
 500             velecsum         = _mm256_add_ps(velecsum,velec);
 501
 502             fscal            = felec;
 503
 504             /* Calculate temporary vectorial force */
 505             tx               = _mm256_mul_ps(fscal,dx13);
 506             ty               = _mm256_mul_ps(fscal,dy13);
 507             tz               = _mm256_mul_ps(fscal,dz13);
 508
 509             /* Update vectorial force */
 510             fix1             = _mm256_add_ps(fix1,tx);
 511             fiy1             = _mm256_add_ps(fiy1,ty);
 512             fiz1             = _mm256_add_ps(fiz1,tz);
 513
 514             fjx3             = _mm256_add_ps(fjx3,tx);
 515             fjy3             = _mm256_add_ps(fjy3,ty);
 516             fjz3             = _mm256_add_ps(fjz3,tz);
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             r21              = _mm256_mul_ps(rsq21,rinv21);
 523
 524             /* EWALD ELECTROSTATICS */
 525
 526             /* Analytical PME correction */
 527             zeta2            = _mm256_mul_ps(beta2,rsq21);
 528             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
 529             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 530             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 531             felec            = _mm256_mul_ps(qq21,felec);
 532             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 533             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 534             velec            = _mm256_sub_ps(rinv21,pmecorrV);
 535             velec            = _mm256_mul_ps(qq21,velec);
 536
 537             /* Update potential sum for this i atom from the interaction with this j atom. */
 538             velecsum         = _mm256_add_ps(velecsum,velec);
 539
 540             fscal            = felec;
 541
 542             /* Calculate temporary vectorial force */
 543             tx               = _mm256_mul_ps(fscal,dx21);
 544             ty               = _mm256_mul_ps(fscal,dy21);
 545             tz               = _mm256_mul_ps(fscal,dz21);
 546
 547             /* Update vectorial force */
 548             fix2             = _mm256_add_ps(fix2,tx);
 549             fiy2             = _mm256_add_ps(fiy2,ty);
 550             fiz2             = _mm256_add_ps(fiz2,tz);
 551
 552             fjx1             = _mm256_add_ps(fjx1,tx);
 553             fjy1             = _mm256_add_ps(fjy1,ty);
 554             fjz1             = _mm256_add_ps(fjz1,tz);
 555
 556             /**************************
 557              * CALCULATE INTERACTIONS *
 558              **************************/
 559
 560             r22              = _mm256_mul_ps(rsq22,rinv22);
 561
 562             /* EWALD ELECTROSTATICS */
 563
 564             /* Analytical PME correction */
 565             zeta2            = _mm256_mul_ps(beta2,rsq22);
 566             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
 567             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 568             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 569             felec            = _mm256_mul_ps(qq22,felec);
 570             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 571             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 572             velec            = _mm256_sub_ps(rinv22,pmecorrV);
 573             velec            = _mm256_mul_ps(qq22,velec);
 574
 575             /* Update potential sum for this i atom from the interaction with this j atom. */
 576             velecsum         = _mm256_add_ps(velecsum,velec);
 577
 578             fscal            = felec;
 579
 580             /* Calculate temporary vectorial force */
 581             tx               = _mm256_mul_ps(fscal,dx22);
 582             ty               = _mm256_mul_ps(fscal,dy22);
 583             tz               = _mm256_mul_ps(fscal,dz22);
 584
 585             /* Update vectorial force */
 586             fix2             = _mm256_add_ps(fix2,tx);
 587             fiy2             = _mm256_add_ps(fiy2,ty);
 588             fiz2             = _mm256_add_ps(fiz2,tz);
 589
 590             fjx2             = _mm256_add_ps(fjx2,tx);
 591             fjy2             = _mm256_add_ps(fjy2,ty);
 592             fjz2             = _mm256_add_ps(fjz2,tz);
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             r23              = _mm256_mul_ps(rsq23,rinv23);
 599
 600             /* EWALD ELECTROSTATICS */
 601
 602             /* Analytical PME correction */
 603             zeta2            = _mm256_mul_ps(beta2,rsq23);
 604             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
 605             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 606             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 607             felec            = _mm256_mul_ps(qq23,felec);
 608             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 609             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 610             velec            = _mm256_sub_ps(rinv23,pmecorrV);
 611             velec            = _mm256_mul_ps(qq23,velec);
 612
 613             /* Update potential sum for this i atom from the interaction with this j atom. */
 614             velecsum         = _mm256_add_ps(velecsum,velec);
 615
 616             fscal            = felec;
 617
 618             /* Calculate temporary vectorial force */
 619             tx               = _mm256_mul_ps(fscal,dx23);
 620             ty               = _mm256_mul_ps(fscal,dy23);
 621             tz               = _mm256_mul_ps(fscal,dz23);
 622
 623             /* Update vectorial force */
 624             fix2             = _mm256_add_ps(fix2,tx);
 625             fiy2             = _mm256_add_ps(fiy2,ty);
 626             fiz2             = _mm256_add_ps(fiz2,tz);
 627
 628             fjx3             = _mm256_add_ps(fjx3,tx);
 629             fjy3             = _mm256_add_ps(fjy3,ty);
 630             fjz3             = _mm256_add_ps(fjz3,tz);
 631
 632             /**************************
 633              * CALCULATE INTERACTIONS *
 634              **************************/
 635
 636             r31              = _mm256_mul_ps(rsq31,rinv31);
 637
 638             /* EWALD ELECTROSTATICS */
 639
 640             /* Analytical PME correction */
 641             zeta2            = _mm256_mul_ps(beta2,rsq31);
 642             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
 643             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 644             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 645             felec            = _mm256_mul_ps(qq31,felec);
 646             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 647             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 648             velec            = _mm256_sub_ps(rinv31,pmecorrV);
 649             velec            = _mm256_mul_ps(qq31,velec);
 650
 651             /* Update potential sum for this i atom from the interaction with this j atom. */
 652             velecsum         = _mm256_add_ps(velecsum,velec);
 653
 654             fscal            = felec;
 655
 656             /* Calculate temporary vectorial force */
 657             tx               = _mm256_mul_ps(fscal,dx31);
 658             ty               = _mm256_mul_ps(fscal,dy31);
 659             tz               = _mm256_mul_ps(fscal,dz31);
 660
 661             /* Update vectorial force */
 662             fix3             = _mm256_add_ps(fix3,tx);
 663             fiy3             = _mm256_add_ps(fiy3,ty);
 664             fiz3             = _mm256_add_ps(fiz3,tz);
 665
 666             fjx1             = _mm256_add_ps(fjx1,tx);
 667             fjy1             = _mm256_add_ps(fjy1,ty);
 668             fjz1             = _mm256_add_ps(fjz1,tz);
 669
 670             /**************************
 671              * CALCULATE INTERACTIONS *
 672              **************************/
 673
 674             r32              = _mm256_mul_ps(rsq32,rinv32);
 675
 676             /* EWALD ELECTROSTATICS */
 677
 678             /* Analytical PME correction */
 679             zeta2            = _mm256_mul_ps(beta2,rsq32);
 680             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
 681             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 682             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 683             felec            = _mm256_mul_ps(qq32,felec);
 684             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 685             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 686             velec            = _mm256_sub_ps(rinv32,pmecorrV);
 687             velec            = _mm256_mul_ps(qq32,velec);
 688
 689             /* Update potential sum for this i atom from the interaction with this j atom. */
 690             velecsum         = _mm256_add_ps(velecsum,velec);
 691
 692             fscal            = felec;
 693
 694             /* Calculate temporary vectorial force */
 695             tx               = _mm256_mul_ps(fscal,dx32);
 696             ty               = _mm256_mul_ps(fscal,dy32);
 697             tz               = _mm256_mul_ps(fscal,dz32);
 698
 699             /* Update vectorial force */
 700             fix3             = _mm256_add_ps(fix3,tx);
 701             fiy3             = _mm256_add_ps(fiy3,ty);
 702             fiz3             = _mm256_add_ps(fiz3,tz);
 703
 704             fjx2             = _mm256_add_ps(fjx2,tx);
 705             fjy2             = _mm256_add_ps(fjy2,ty);
 706             fjz2             = _mm256_add_ps(fjz2,tz);
 707
 708             /**************************
 709              * CALCULATE INTERACTIONS *
 710              **************************/
 711
 712             r33              = _mm256_mul_ps(rsq33,rinv33);
 713
 714             /* EWALD ELECTROSTATICS */
 715
 716             /* Analytical PME correction */
 717             zeta2            = _mm256_mul_ps(beta2,rsq33);
 718             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
 719             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 720             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 721             felec            = _mm256_mul_ps(qq33,felec);
 722             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 723             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 724             velec            = _mm256_sub_ps(rinv33,pmecorrV);
 725             velec            = _mm256_mul_ps(qq33,velec);
 726
 727             /* Update potential sum for this i atom from the interaction with this j atom. */
 728             velecsum         = _mm256_add_ps(velecsum,velec);
 729
 730             fscal            = felec;
 731
 732             /* Calculate temporary vectorial force */
 733             tx               = _mm256_mul_ps(fscal,dx33);
 734             ty               = _mm256_mul_ps(fscal,dy33);
 735             tz               = _mm256_mul_ps(fscal,dz33);
 736
 737             /* Update vectorial force */
 738             fix3             = _mm256_add_ps(fix3,tx);
 739             fiy3             = _mm256_add_ps(fiy3,ty);
 740             fiz3             = _mm256_add_ps(fiz3,tz);
 741
 742             fjx3             = _mm256_add_ps(fjx3,tx);
 743             fjy3             = _mm256_add_ps(fjy3,ty);
 744             fjz3             = _mm256_add_ps(fjz3,tz);
 745
 746             fjptrA             = f+j_coord_offsetA;
 747             fjptrB             = f+j_coord_offsetB;
 748             fjptrC             = f+j_coord_offsetC;
 749             fjptrD             = f+j_coord_offsetD;
 750             fjptrE             = f+j_coord_offsetE;
 751             fjptrF             = f+j_coord_offsetF;
 752             fjptrG             = f+j_coord_offsetG;
 753             fjptrH             = f+j_coord_offsetH;
 754
 755             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 756                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 757                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 758
 759             /* Inner loop uses 810 flops */
 760         }
 761
 762         if(jidx<j_index_end)
 763         {
 764
 765             /* Get j neighbor index, and coordinate index */
 766             jnrlistA         = jjnr[jidx];
 767             jnrlistB         = jjnr[jidx+1];
 768             jnrlistC         = jjnr[jidx+2];
 769             jnrlistD         = jjnr[jidx+3];
 770             jnrlistE         = jjnr[jidx+4];
 771             jnrlistF         = jjnr[jidx+5];
 772             jnrlistG         = jjnr[jidx+6];
 773             jnrlistH         = jjnr[jidx+7];
 774             /* Sign of each element will be negative for non-real atoms.
 775              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 776              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 777              */
 778             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 779                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 780
 781             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 782             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 783             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 784             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 785             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 786             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 787             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 788             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 789             j_coord_offsetA  = DIM*jnrA;
 790             j_coord_offsetB  = DIM*jnrB;
 791             j_coord_offsetC  = DIM*jnrC;
 792             j_coord_offsetD  = DIM*jnrD;
 793             j_coord_offsetE  = DIM*jnrE;
 794             j_coord_offsetF  = DIM*jnrF;
 795             j_coord_offsetG  = DIM*jnrG;
 796             j_coord_offsetH  = DIM*jnrH;
 797
 798             /* load j atom coordinates */
 799             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 800                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 801                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 802                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 803                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 804                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 805
 806             /* Calculate displacement vector */
 807             dx00             = _mm256_sub_ps(ix0,jx0);
 808             dy00             = _mm256_sub_ps(iy0,jy0);
 809             dz00             = _mm256_sub_ps(iz0,jz0);
 810             dx11             = _mm256_sub_ps(ix1,jx1);
 811             dy11             = _mm256_sub_ps(iy1,jy1);
 812             dz11             = _mm256_sub_ps(iz1,jz1);
 813             dx12             = _mm256_sub_ps(ix1,jx2);
 814             dy12             = _mm256_sub_ps(iy1,jy2);
 815             dz12             = _mm256_sub_ps(iz1,jz2);
 816             dx13             = _mm256_sub_ps(ix1,jx3);
 817             dy13             = _mm256_sub_ps(iy1,jy3);
 818             dz13             = _mm256_sub_ps(iz1,jz3);
 819             dx21             = _mm256_sub_ps(ix2,jx1);
 820             dy21             = _mm256_sub_ps(iy2,jy1);
 821             dz21             = _mm256_sub_ps(iz2,jz1);
 822             dx22             = _mm256_sub_ps(ix2,jx2);
 823             dy22             = _mm256_sub_ps(iy2,jy2);
 824             dz22             = _mm256_sub_ps(iz2,jz2);
 825             dx23             = _mm256_sub_ps(ix2,jx3);
 826             dy23             = _mm256_sub_ps(iy2,jy3);
 827             dz23             = _mm256_sub_ps(iz2,jz3);
 828             dx31             = _mm256_sub_ps(ix3,jx1);
 829             dy31             = _mm256_sub_ps(iy3,jy1);
 830             dz31             = _mm256_sub_ps(iz3,jz1);
 831             dx32             = _mm256_sub_ps(ix3,jx2);
 832             dy32             = _mm256_sub_ps(iy3,jy2);
 833             dz32             = _mm256_sub_ps(iz3,jz2);
 834             dx33             = _mm256_sub_ps(ix3,jx3);
 835             dy33             = _mm256_sub_ps(iy3,jy3);
 836             dz33             = _mm256_sub_ps(iz3,jz3);
 837
 838             /* Calculate squared distance and things based on it */
 839             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 840             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 841             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 842             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 843             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 844             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 845             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 846             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 847             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 848             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 849
 850             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 851             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 852             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 853             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 854             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 855             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 856             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 857             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 858             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 859             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 860
 861             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 862             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 863             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 864             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 865             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 866             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 867             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 868             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 869             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 870             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 871
 872             fjx0             = _mm256_setzero_ps();
 873             fjy0             = _mm256_setzero_ps();
 874             fjz0             = _mm256_setzero_ps();
 875             fjx1             = _mm256_setzero_ps();
 876             fjy1             = _mm256_setzero_ps();
 877             fjz1             = _mm256_setzero_ps();
 878             fjx2             = _mm256_setzero_ps();
 879             fjy2             = _mm256_setzero_ps();
 880             fjz2             = _mm256_setzero_ps();
 881             fjx3             = _mm256_setzero_ps();
 882             fjy3             = _mm256_setzero_ps();
 883             fjz3             = _mm256_setzero_ps();
 884
 885             /**************************
 886              * CALCULATE INTERACTIONS *
 887              **************************/
 888
 889             r00              = _mm256_mul_ps(rsq00,rinv00);
 890             r00              = _mm256_andnot_ps(dummy_mask,r00);
 891
 892             /* Analytical LJ-PME */
 893             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 894             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
 895             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
 896             exponent         = gmx_simd_exp_r(ewcljrsq);
 897             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 898             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
 899             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 900             vvdw6            = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
 901             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 902             vvdw             = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
 903             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 904             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
 905
 906             /* Update potential sum for this i atom from the interaction with this j atom. */
 907             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 908             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 909
 910             fscal            = fvdw;
 911
 912             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 913
 914             /* Calculate temporary vectorial force */
 915             tx               = _mm256_mul_ps(fscal,dx00);
 916             ty               = _mm256_mul_ps(fscal,dy00);
 917             tz               = _mm256_mul_ps(fscal,dz00);
 918
 919             /* Update vectorial force */
 920             fix0             = _mm256_add_ps(fix0,tx);
 921             fiy0             = _mm256_add_ps(fiy0,ty);
 922             fiz0             = _mm256_add_ps(fiz0,tz);
 923
 924             fjx0             = _mm256_add_ps(fjx0,tx);
 925             fjy0             = _mm256_add_ps(fjy0,ty);
 926             fjz0             = _mm256_add_ps(fjz0,tz);
 927
 928             /**************************
 929              * CALCULATE INTERACTIONS *
 930              **************************/
 931
 932             r11              = _mm256_mul_ps(rsq11,rinv11);
 933             r11              = _mm256_andnot_ps(dummy_mask,r11);
 934
 935             /* EWALD ELECTROSTATICS */
 936
 937             /* Analytical PME correction */
 938             zeta2            = _mm256_mul_ps(beta2,rsq11);
 939             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 940             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 941             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 942             felec            = _mm256_mul_ps(qq11,felec);
 943             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 944             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 945             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 946             velec            = _mm256_mul_ps(qq11,velec);
 947
 948             /* Update potential sum for this i atom from the interaction with this j atom. */
 949             velec            = _mm256_andnot_ps(dummy_mask,velec);
 950             velecsum         = _mm256_add_ps(velecsum,velec);
 951
 952             fscal            = felec;
 953
 954             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 955
 956             /* Calculate temporary vectorial force */
 957             tx               = _mm256_mul_ps(fscal,dx11);
 958             ty               = _mm256_mul_ps(fscal,dy11);
 959             tz               = _mm256_mul_ps(fscal,dz11);
 960
 961             /* Update vectorial force */
 962             fix1             = _mm256_add_ps(fix1,tx);
 963             fiy1             = _mm256_add_ps(fiy1,ty);
 964             fiz1             = _mm256_add_ps(fiz1,tz);
 965
 966             fjx1             = _mm256_add_ps(fjx1,tx);
 967             fjy1             = _mm256_add_ps(fjy1,ty);
 968             fjz1             = _mm256_add_ps(fjz1,tz);
 969
 970             /**************************
 971              * CALCULATE INTERACTIONS *
 972              **************************/
 973
 974             r12              = _mm256_mul_ps(rsq12,rinv12);
 975             r12              = _mm256_andnot_ps(dummy_mask,r12);
 976
 977             /* EWALD ELECTROSTATICS */
 978
 979             /* Analytical PME correction */
 980             zeta2            = _mm256_mul_ps(beta2,rsq12);
 981             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 982             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 983             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 984             felec            = _mm256_mul_ps(qq12,felec);
 985             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 986             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 987             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 988             velec            = _mm256_mul_ps(qq12,velec);
 989
 990             /* Update potential sum for this i atom from the interaction with this j atom. */
 991             velec            = _mm256_andnot_ps(dummy_mask,velec);
 992             velecsum         = _mm256_add_ps(velecsum,velec);
 993
 994             fscal            = felec;
 995
 996             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 997
 998             /* Calculate temporary vectorial force */
 999             tx               = _mm256_mul_ps(fscal,dx12);
1000             ty               = _mm256_mul_ps(fscal,dy12);
1001             tz               = _mm256_mul_ps(fscal,dz12);
1002
1003             /* Update vectorial force */
1004             fix1             = _mm256_add_ps(fix1,tx);
1005             fiy1             = _mm256_add_ps(fiy1,ty);
1006             fiz1             = _mm256_add_ps(fiz1,tz);
1007
1008             fjx2             = _mm256_add_ps(fjx2,tx);
1009             fjy2             = _mm256_add_ps(fjy2,ty);
1010             fjz2             = _mm256_add_ps(fjz2,tz);
1011
1012             /**************************
1013              * CALCULATE INTERACTIONS *
1014              **************************/
1015
1016             r13              = _mm256_mul_ps(rsq13,rinv13);
1017             r13              = _mm256_andnot_ps(dummy_mask,r13);
1018
1019             /* EWALD ELECTROSTATICS */
1020
1021             /* Analytical PME correction */
1022             zeta2            = _mm256_mul_ps(beta2,rsq13);
1023             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
1024             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1025             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1026             felec            = _mm256_mul_ps(qq13,felec);
1027             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1028             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1029             velec            = _mm256_sub_ps(rinv13,pmecorrV);
1030             velec            = _mm256_mul_ps(qq13,velec);
1031
1032             /* Update potential sum for this i atom from the interaction with this j atom. */
1033             velec            = _mm256_andnot_ps(dummy_mask,velec);
1034             velecsum         = _mm256_add_ps(velecsum,velec);
1035
1036             fscal            = felec;
1037
1038             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1039
1040             /* Calculate temporary vectorial force */
1041             tx               = _mm256_mul_ps(fscal,dx13);
1042             ty               = _mm256_mul_ps(fscal,dy13);
1043             tz               = _mm256_mul_ps(fscal,dz13);
1044
1045             /* Update vectorial force */
1046             fix1             = _mm256_add_ps(fix1,tx);
1047             fiy1             = _mm256_add_ps(fiy1,ty);
1048             fiz1             = _mm256_add_ps(fiz1,tz);
1049
1050             fjx3             = _mm256_add_ps(fjx3,tx);
1051             fjy3             = _mm256_add_ps(fjy3,ty);
1052             fjz3             = _mm256_add_ps(fjz3,tz);
1053
1054             /**************************
1055              * CALCULATE INTERACTIONS *
1056              **************************/
1057
1058             r21              = _mm256_mul_ps(rsq21,rinv21);
1059             r21              = _mm256_andnot_ps(dummy_mask,r21);
1060
1061             /* EWALD ELECTROSTATICS */
1062
1063             /* Analytical PME correction */
1064             zeta2            = _mm256_mul_ps(beta2,rsq21);
1065             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1066             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1067             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1068             felec            = _mm256_mul_ps(qq21,felec);
1069             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1070             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1071             velec            = _mm256_sub_ps(rinv21,pmecorrV);
1072             velec            = _mm256_mul_ps(qq21,velec);
1073
1074             /* Update potential sum for this i atom from the interaction with this j atom. */
1075             velec            = _mm256_andnot_ps(dummy_mask,velec);
1076             velecsum         = _mm256_add_ps(velecsum,velec);
1077
1078             fscal            = felec;
1079
1080             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1081
1082             /* Calculate temporary vectorial force */
1083             tx               = _mm256_mul_ps(fscal,dx21);
1084             ty               = _mm256_mul_ps(fscal,dy21);
1085             tz               = _mm256_mul_ps(fscal,dz21);
1086
1087             /* Update vectorial force */
1088             fix2             = _mm256_add_ps(fix2,tx);
1089             fiy2             = _mm256_add_ps(fiy2,ty);
1090             fiz2             = _mm256_add_ps(fiz2,tz);
1091
1092             fjx1             = _mm256_add_ps(fjx1,tx);
1093             fjy1             = _mm256_add_ps(fjy1,ty);
1094             fjz1             = _mm256_add_ps(fjz1,tz);
1095
1096             /**************************
1097              * CALCULATE INTERACTIONS *
1098              **************************/
1099
1100             r22              = _mm256_mul_ps(rsq22,rinv22);
1101             r22              = _mm256_andnot_ps(dummy_mask,r22);
1102
1103             /* EWALD ELECTROSTATICS */
1104
1105             /* Analytical PME correction */
1106             zeta2            = _mm256_mul_ps(beta2,rsq22);
1107             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1108             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1109             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1110             felec            = _mm256_mul_ps(qq22,felec);
1111             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1112             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1113             velec            = _mm256_sub_ps(rinv22,pmecorrV);
1114             velec            = _mm256_mul_ps(qq22,velec);
1115
1116             /* Update potential sum for this i atom from the interaction with this j atom. */
1117             velec            = _mm256_andnot_ps(dummy_mask,velec);
1118             velecsum         = _mm256_add_ps(velecsum,velec);
1119
1120             fscal            = felec;
1121
1122             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1123
1124             /* Calculate temporary vectorial force */
1125             tx               = _mm256_mul_ps(fscal,dx22);
1126             ty               = _mm256_mul_ps(fscal,dy22);
1127             tz               = _mm256_mul_ps(fscal,dz22);
1128
1129             /* Update vectorial force */
1130             fix2             = _mm256_add_ps(fix2,tx);
1131             fiy2             = _mm256_add_ps(fiy2,ty);
1132             fiz2             = _mm256_add_ps(fiz2,tz);
1133
1134             fjx2             = _mm256_add_ps(fjx2,tx);
1135             fjy2             = _mm256_add_ps(fjy2,ty);
1136             fjz2             = _mm256_add_ps(fjz2,tz);
1137
1138             /**************************
1139              * CALCULATE INTERACTIONS *
1140              **************************/
1141
1142             r23              = _mm256_mul_ps(rsq23,rinv23);
1143             r23              = _mm256_andnot_ps(dummy_mask,r23);
1144
1145             /* EWALD ELECTROSTATICS */
1146
1147             /* Analytical PME correction */
1148             zeta2            = _mm256_mul_ps(beta2,rsq23);
1149             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1150             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1151             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1152             felec            = _mm256_mul_ps(qq23,felec);
1153             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1154             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1155             velec            = _mm256_sub_ps(rinv23,pmecorrV);
1156             velec            = _mm256_mul_ps(qq23,velec);
1157
1158             /* Update potential sum for this i atom from the interaction with this j atom. */
1159             velec            = _mm256_andnot_ps(dummy_mask,velec);
1160             velecsum         = _mm256_add_ps(velecsum,velec);
1161
1162             fscal            = felec;
1163
1164             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1165
1166             /* Calculate temporary vectorial force */
1167             tx               = _mm256_mul_ps(fscal,dx23);
1168             ty               = _mm256_mul_ps(fscal,dy23);
1169             tz               = _mm256_mul_ps(fscal,dz23);
1170
1171             /* Update vectorial force */
1172             fix2             = _mm256_add_ps(fix2,tx);
1173             fiy2             = _mm256_add_ps(fiy2,ty);
1174             fiz2             = _mm256_add_ps(fiz2,tz);
1175
1176             fjx3             = _mm256_add_ps(fjx3,tx);
1177             fjy3             = _mm256_add_ps(fjy3,ty);
1178             fjz3             = _mm256_add_ps(fjz3,tz);
1179
1180             /**************************
1181              * CALCULATE INTERACTIONS *
1182              **************************/
1183
1184             r31              = _mm256_mul_ps(rsq31,rinv31);
1185             r31              = _mm256_andnot_ps(dummy_mask,r31);
1186
1187             /* EWALD ELECTROSTATICS */
1188
1189             /* Analytical PME correction */
1190             zeta2            = _mm256_mul_ps(beta2,rsq31);
1191             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1192             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1193             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1194             felec            = _mm256_mul_ps(qq31,felec);
1195             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1196             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1197             velec            = _mm256_sub_ps(rinv31,pmecorrV);
1198             velec            = _mm256_mul_ps(qq31,velec);
1199
1200             /* Update potential sum for this i atom from the interaction with this j atom. */
1201             velec            = _mm256_andnot_ps(dummy_mask,velec);
1202             velecsum         = _mm256_add_ps(velecsum,velec);
1203
1204             fscal            = felec;
1205
1206             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1207
1208             /* Calculate temporary vectorial force */
1209             tx               = _mm256_mul_ps(fscal,dx31);
1210             ty               = _mm256_mul_ps(fscal,dy31);
1211             tz               = _mm256_mul_ps(fscal,dz31);
1212
1213             /* Update vectorial force */
1214             fix3             = _mm256_add_ps(fix3,tx);
1215             fiy3             = _mm256_add_ps(fiy3,ty);
1216             fiz3             = _mm256_add_ps(fiz3,tz);
1217
1218             fjx1             = _mm256_add_ps(fjx1,tx);
1219             fjy1             = _mm256_add_ps(fjy1,ty);
1220             fjz1             = _mm256_add_ps(fjz1,tz);
1221
1222             /**************************
1223              * CALCULATE INTERACTIONS *
1224              **************************/
1225
1226             r32              = _mm256_mul_ps(rsq32,rinv32);
1227             r32              = _mm256_andnot_ps(dummy_mask,r32);
1228
1229             /* EWALD ELECTROSTATICS */
1230
1231             /* Analytical PME correction */
1232             zeta2            = _mm256_mul_ps(beta2,rsq32);
1233             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1234             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1235             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1236             felec            = _mm256_mul_ps(qq32,felec);
1237             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1238             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1239             velec            = _mm256_sub_ps(rinv32,pmecorrV);
1240             velec            = _mm256_mul_ps(qq32,velec);
1241
1242             /* Update potential sum for this i atom from the interaction with this j atom. */
1243             velec            = _mm256_andnot_ps(dummy_mask,velec);
1244             velecsum         = _mm256_add_ps(velecsum,velec);
1245
1246             fscal            = felec;
1247
1248             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1249
1250             /* Calculate temporary vectorial force */
1251             tx               = _mm256_mul_ps(fscal,dx32);
1252             ty               = _mm256_mul_ps(fscal,dy32);
1253             tz               = _mm256_mul_ps(fscal,dz32);
1254
1255             /* Update vectorial force */
1256             fix3             = _mm256_add_ps(fix3,tx);
1257             fiy3             = _mm256_add_ps(fiy3,ty);
1258             fiz3             = _mm256_add_ps(fiz3,tz);
1259
1260             fjx2             = _mm256_add_ps(fjx2,tx);
1261             fjy2             = _mm256_add_ps(fjy2,ty);
1262             fjz2             = _mm256_add_ps(fjz2,tz);
1263
1264             /**************************
1265              * CALCULATE INTERACTIONS *
1266              **************************/
1267
1268             r33              = _mm256_mul_ps(rsq33,rinv33);
1269             r33              = _mm256_andnot_ps(dummy_mask,r33);
1270
1271             /* EWALD ELECTROSTATICS */
1272
1273             /* Analytical PME correction */
1274             zeta2            = _mm256_mul_ps(beta2,rsq33);
1275             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1276             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1277             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1278             felec            = _mm256_mul_ps(qq33,felec);
1279             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1280             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1281             velec            = _mm256_sub_ps(rinv33,pmecorrV);
1282             velec            = _mm256_mul_ps(qq33,velec);
1283
1284             /* Update potential sum for this i atom from the interaction with this j atom. */
1285             velec            = _mm256_andnot_ps(dummy_mask,velec);
1286             velecsum         = _mm256_add_ps(velecsum,velec);
1287
1288             fscal            = felec;
1289
1290             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1291
1292             /* Calculate temporary vectorial force */
1293             tx               = _mm256_mul_ps(fscal,dx33);
1294             ty               = _mm256_mul_ps(fscal,dy33);
1295             tz               = _mm256_mul_ps(fscal,dz33);
1296
1297             /* Update vectorial force */
1298             fix3             = _mm256_add_ps(fix3,tx);
1299             fiy3             = _mm256_add_ps(fiy3,ty);
1300             fiz3             = _mm256_add_ps(fiz3,tz);
1301
1302             fjx3             = _mm256_add_ps(fjx3,tx);
1303             fjy3             = _mm256_add_ps(fjy3,ty);
1304             fjz3             = _mm256_add_ps(fjz3,tz);
1305
1306             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1307             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1308             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1309             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1310             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1311             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1312             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1313             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1314
1315             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1316                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1317                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1318
1319             /* Inner loop uses 820 flops */
1320         }
1321
1322         /* End of innermost loop */
1323
1324         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1325                                                  f+i_coord_offset,fshift+i_shift_offset);
1326
1327         ggid                        = gid[iidx];
1328         /* Update potential energies */
1329         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1330         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1331
1332         /* Increment number of inner iterations */
1333         inneriter                  += j_index_end - j_index_start;
1334
1335         /* Outer loop uses 26 flops */
1336     }
1337
1338     /* Increment number of outer iterations */
1339     outeriter        += nri;
1340
1341     /* Update outer/inner flops */
1342
1343     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*820);
1344 }
1345 /*
1346  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_avx_256_single
1347  * Electrostatics interaction: Ewald
1348  * VdW interaction:            LJEwald
1349  * Geometry:                   Water4-Water4
1350  * Calculate force/pot:        Force
1351  */
1352 void
1353 nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_avx_256_single
1354                     (t_nblist                    * gmx_restrict       nlist,
1355                      rvec                        * gmx_restrict          xx,
1356                      rvec                        * gmx_restrict          ff,
1357                      t_forcerec                  * gmx_restrict          fr,
1358                      t_mdatoms                   * gmx_restrict     mdatoms,
1359                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1360                      t_nrnb                      * gmx_restrict        nrnb)
1361 {
1362     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1363      * just 0 for non-waters.
1364      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1365      * jnr indices corresponding to data put in the four positions in the SIMD register.
1366      */
1367     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1368     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1369     int              jnrA,jnrB,jnrC,jnrD;
1370     int              jnrE,jnrF,jnrG,jnrH;
1371     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1372     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1373     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1374     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1375     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1376     real             rcutoff_scalar;
1377     real             *shiftvec,*fshift,*x,*f;
1378     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1379     real             scratch[4*DIM];
1380     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1381     real *           vdwioffsetptr0;
1382     real *           vdwgridioffsetptr0;
1383     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1384     real *           vdwioffsetptr1;
1385     real *           vdwgridioffsetptr1;
1386     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1387     real *           vdwioffsetptr2;
1388     real *           vdwgridioffsetptr2;
1389     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1390     real *           vdwioffsetptr3;
1391     real *           vdwgridioffsetptr3;
1392     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1393     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1394     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1395     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1396     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1397     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1398     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1399     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1400     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1401     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1402     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1403     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1404     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1405     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1406     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1407     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1408     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1409     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1410     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1411     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1412     real             *charge;
1413     int              nvdwtype;
1414     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1415     int              *vdwtype;
1416     real             *vdwparam;
1417     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1418     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1419     __m256           c6grid_00;
1420     __m256           c6grid_11;
1421     __m256           c6grid_12;
1422     __m256           c6grid_13;
1423     __m256           c6grid_21;
1424     __m256           c6grid_22;
1425     __m256           c6grid_23;
1426     __m256           c6grid_31;
1427     __m256           c6grid_32;
1428     __m256           c6grid_33;
1429     real             *vdwgridparam;
1430     __m256           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1431     __m256           one_half  = _mm256_set1_ps(0.5);
1432     __m256           minus_one = _mm256_set1_ps(-1.0);
1433     __m256i          ewitab;
1434     __m128i          ewitab_lo,ewitab_hi;
1435     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1436     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1437     real             *ewtab;
1438     __m256           dummy_mask,cutoff_mask;
1439     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1440     __m256           one     = _mm256_set1_ps(1.0);
1441     __m256           two     = _mm256_set1_ps(2.0);
1442     x                = xx[0];
1443     f                = ff[0];
1444
1445     nri              = nlist->nri;
1446     iinr             = nlist->iinr;
1447     jindex           = nlist->jindex;
1448     jjnr             = nlist->jjnr;
1449     shiftidx         = nlist->shift;
1450     gid              = nlist->gid;
1451     shiftvec         = fr->shift_vec[0];
1452     fshift           = fr->fshift[0];
1453     facel            = _mm256_set1_ps(fr->epsfac);
1454     charge           = mdatoms->chargeA;
1455     nvdwtype         = fr->ntype;
1456     vdwparam         = fr->nbfp;
1457     vdwtype          = mdatoms->typeA;
1458     vdwgridparam     = fr->ljpme_c6grid;
1459     sh_lj_ewald      = _mm256_set1_ps(fr->ic->sh_lj_ewald);
1460     ewclj            = _mm256_set1_ps(fr->ewaldcoeff_lj);
1461     ewclj2           = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
1462
1463     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
1464     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1465     beta2            = _mm256_mul_ps(beta,beta);
1466     beta3            = _mm256_mul_ps(beta,beta2);
1467
1468     ewtab            = fr->ic->tabq_coul_F;
1469     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
1470     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1471
1472     /* Setup water-specific parameters */
1473     inr              = nlist->iinr[0];
1474     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1475     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1476     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1477     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1478     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1479
1480     jq1              = _mm256_set1_ps(charge[inr+1]);
1481     jq2              = _mm256_set1_ps(charge[inr+2]);
1482     jq3              = _mm256_set1_ps(charge[inr+3]);
1483     vdwjidx0A        = 2*vdwtype[inr+0];
1484     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1485     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1486     c6grid_00        = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
1487     qq11             = _mm256_mul_ps(iq1,jq1);
1488     qq12             = _mm256_mul_ps(iq1,jq2);
1489     qq13             = _mm256_mul_ps(iq1,jq3);
1490     qq21             = _mm256_mul_ps(iq2,jq1);
1491     qq22             = _mm256_mul_ps(iq2,jq2);
1492     qq23             = _mm256_mul_ps(iq2,jq3);
1493     qq31             = _mm256_mul_ps(iq3,jq1);
1494     qq32             = _mm256_mul_ps(iq3,jq2);
1495     qq33             = _mm256_mul_ps(iq3,jq3);
1496
1497     /* Avoid stupid compiler warnings */
1498     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1499     j_coord_offsetA = 0;
1500     j_coord_offsetB = 0;
1501     j_coord_offsetC = 0;
1502     j_coord_offsetD = 0;
1503     j_coord_offsetE = 0;
1504     j_coord_offsetF = 0;
1505     j_coord_offsetG = 0;
1506     j_coord_offsetH = 0;
1507
1508     outeriter        = 0;
1509     inneriter        = 0;
1510
1511     for(iidx=0;iidx<4*DIM;iidx++)
1512     {
1513         scratch[iidx] = 0.0;
1514     }
1515
1516     /* Start outer loop over neighborlists */
1517     for(iidx=0; iidx<nri; iidx++)
1518     {
1519         /* Load shift vector for this list */
1520         i_shift_offset   = DIM*shiftidx[iidx];
1521
1522         /* Load limits for loop over neighbors */
1523         j_index_start    = jindex[iidx];
1524         j_index_end      = jindex[iidx+1];
1525
1526         /* Get outer coordinate index */
1527         inr              = iinr[iidx];
1528         i_coord_offset   = DIM*inr;
1529
1530         /* Load i particle coords and add shift vector */
1531         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1532                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1533
1534         fix0             = _mm256_setzero_ps();
1535         fiy0             = _mm256_setzero_ps();
1536         fiz0             = _mm256_setzero_ps();
1537         fix1             = _mm256_setzero_ps();
1538         fiy1             = _mm256_setzero_ps();
1539         fiz1             = _mm256_setzero_ps();
1540         fix2             = _mm256_setzero_ps();
1541         fiy2             = _mm256_setzero_ps();
1542         fiz2             = _mm256_setzero_ps();
1543         fix3             = _mm256_setzero_ps();
1544         fiy3             = _mm256_setzero_ps();
1545         fiz3             = _mm256_setzero_ps();
1546
1547         /* Start inner kernel loop */
1548         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1549         {
1550
1551             /* Get j neighbor index, and coordinate index */
1552             jnrA             = jjnr[jidx];
1553             jnrB             = jjnr[jidx+1];
1554             jnrC             = jjnr[jidx+2];
1555             jnrD             = jjnr[jidx+3];
1556             jnrE             = jjnr[jidx+4];
1557             jnrF             = jjnr[jidx+5];
1558             jnrG             = jjnr[jidx+6];
1559             jnrH             = jjnr[jidx+7];
1560             j_coord_offsetA  = DIM*jnrA;
1561             j_coord_offsetB  = DIM*jnrB;
1562             j_coord_offsetC  = DIM*jnrC;
1563             j_coord_offsetD  = DIM*jnrD;
1564             j_coord_offsetE  = DIM*jnrE;
1565             j_coord_offsetF  = DIM*jnrF;
1566             j_coord_offsetG  = DIM*jnrG;
1567             j_coord_offsetH  = DIM*jnrH;
1568
1569             /* load j atom coordinates */
1570             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1571                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1572                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1573                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1574                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1575                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1576
1577             /* Calculate displacement vector */
1578             dx00             = _mm256_sub_ps(ix0,jx0);
1579             dy00             = _mm256_sub_ps(iy0,jy0);
1580             dz00             = _mm256_sub_ps(iz0,jz0);
1581             dx11             = _mm256_sub_ps(ix1,jx1);
1582             dy11             = _mm256_sub_ps(iy1,jy1);
1583             dz11             = _mm256_sub_ps(iz1,jz1);
1584             dx12             = _mm256_sub_ps(ix1,jx2);
1585             dy12             = _mm256_sub_ps(iy1,jy2);
1586             dz12             = _mm256_sub_ps(iz1,jz2);
1587             dx13             = _mm256_sub_ps(ix1,jx3);
1588             dy13             = _mm256_sub_ps(iy1,jy3);
1589             dz13             = _mm256_sub_ps(iz1,jz3);
1590             dx21             = _mm256_sub_ps(ix2,jx1);
1591             dy21             = _mm256_sub_ps(iy2,jy1);
1592             dz21             = _mm256_sub_ps(iz2,jz1);
1593             dx22             = _mm256_sub_ps(ix2,jx2);
1594             dy22             = _mm256_sub_ps(iy2,jy2);
1595             dz22             = _mm256_sub_ps(iz2,jz2);
1596             dx23             = _mm256_sub_ps(ix2,jx3);
1597             dy23             = _mm256_sub_ps(iy2,jy3);
1598             dz23             = _mm256_sub_ps(iz2,jz3);
1599             dx31             = _mm256_sub_ps(ix3,jx1);
1600             dy31             = _mm256_sub_ps(iy3,jy1);
1601             dz31             = _mm256_sub_ps(iz3,jz1);
1602             dx32             = _mm256_sub_ps(ix3,jx2);
1603             dy32             = _mm256_sub_ps(iy3,jy2);
1604             dz32             = _mm256_sub_ps(iz3,jz2);
1605             dx33             = _mm256_sub_ps(ix3,jx3);
1606             dy33             = _mm256_sub_ps(iy3,jy3);
1607             dz33             = _mm256_sub_ps(iz3,jz3);
1608
1609             /* Calculate squared distance and things based on it */
1610             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1611             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1612             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1613             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1614             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1615             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1616             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1617             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1618             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1619             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1620
1621             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1622             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1623             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1624             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1625             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1626             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1627             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1628             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1629             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1630             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1631
1632             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1633             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1634             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1635             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1636             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1637             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1638             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1639             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1640             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1641             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1642
1643             fjx0             = _mm256_setzero_ps();
1644             fjy0             = _mm256_setzero_ps();
1645             fjz0             = _mm256_setzero_ps();
1646             fjx1             = _mm256_setzero_ps();
1647             fjy1             = _mm256_setzero_ps();
1648             fjz1             = _mm256_setzero_ps();
1649             fjx2             = _mm256_setzero_ps();
1650             fjy2             = _mm256_setzero_ps();
1651             fjz2             = _mm256_setzero_ps();
1652             fjx3             = _mm256_setzero_ps();
1653             fjy3             = _mm256_setzero_ps();
1654             fjz3             = _mm256_setzero_ps();
1655
1656             /**************************
1657              * CALCULATE INTERACTIONS *
1658              **************************/
1659
1660             r00              = _mm256_mul_ps(rsq00,rinv00);
1661
1662             /* Analytical LJ-PME */
1663             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1664             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
1665             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1666             exponent         = gmx_simd_exp_r(ewcljrsq);
1667             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1668             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1669             /* f6A = 6 * C6grid * (1 - poly) */
1670             f6A              = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1671             /* f6B = C6grid * exponent * beta^6 */
1672             f6B              = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
1673             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1674             fvdw              = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1675
1676             fscal            = fvdw;
1677
1678             /* Calculate temporary vectorial force */
1679             tx               = _mm256_mul_ps(fscal,dx00);
1680             ty               = _mm256_mul_ps(fscal,dy00);
1681             tz               = _mm256_mul_ps(fscal,dz00);
1682
1683             /* Update vectorial force */
1684             fix0             = _mm256_add_ps(fix0,tx);
1685             fiy0             = _mm256_add_ps(fiy0,ty);
1686             fiz0             = _mm256_add_ps(fiz0,tz);
1687
1688             fjx0             = _mm256_add_ps(fjx0,tx);
1689             fjy0             = _mm256_add_ps(fjy0,ty);
1690             fjz0             = _mm256_add_ps(fjz0,tz);
1691
1692             /**************************
1693              * CALCULATE INTERACTIONS *
1694              **************************/
1695
1696             r11              = _mm256_mul_ps(rsq11,rinv11);
1697
1698             /* EWALD ELECTROSTATICS */
1699
1700             /* Analytical PME correction */
1701             zeta2            = _mm256_mul_ps(beta2,rsq11);
1702             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
1703             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1704             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1705             felec            = _mm256_mul_ps(qq11,felec);
1706
1707             fscal            = felec;
1708
1709             /* Calculate temporary vectorial force */
1710             tx               = _mm256_mul_ps(fscal,dx11);
1711             ty               = _mm256_mul_ps(fscal,dy11);
1712             tz               = _mm256_mul_ps(fscal,dz11);
1713
1714             /* Update vectorial force */
1715             fix1             = _mm256_add_ps(fix1,tx);
1716             fiy1             = _mm256_add_ps(fiy1,ty);
1717             fiz1             = _mm256_add_ps(fiz1,tz);
1718
1719             fjx1             = _mm256_add_ps(fjx1,tx);
1720             fjy1             = _mm256_add_ps(fjy1,ty);
1721             fjz1             = _mm256_add_ps(fjz1,tz);
1722
1723             /**************************
1724              * CALCULATE INTERACTIONS *
1725              **************************/
1726
1727             r12              = _mm256_mul_ps(rsq12,rinv12);
1728
1729             /* EWALD ELECTROSTATICS */
1730
1731             /* Analytical PME correction */
1732             zeta2            = _mm256_mul_ps(beta2,rsq12);
1733             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
1734             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1735             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1736             felec            = _mm256_mul_ps(qq12,felec);
1737
1738             fscal            = felec;
1739
1740             /* Calculate temporary vectorial force */
1741             tx               = _mm256_mul_ps(fscal,dx12);
1742             ty               = _mm256_mul_ps(fscal,dy12);
1743             tz               = _mm256_mul_ps(fscal,dz12);
1744
1745             /* Update vectorial force */
1746             fix1             = _mm256_add_ps(fix1,tx);
1747             fiy1             = _mm256_add_ps(fiy1,ty);
1748             fiz1             = _mm256_add_ps(fiz1,tz);
1749
1750             fjx2             = _mm256_add_ps(fjx2,tx);
1751             fjy2             = _mm256_add_ps(fjy2,ty);
1752             fjz2             = _mm256_add_ps(fjz2,tz);
1753
1754             /**************************
1755              * CALCULATE INTERACTIONS *
1756              **************************/
1757
1758             r13              = _mm256_mul_ps(rsq13,rinv13);
1759
1760             /* EWALD ELECTROSTATICS */
1761
1762             /* Analytical PME correction */
1763             zeta2            = _mm256_mul_ps(beta2,rsq13);
1764             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
1765             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1766             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1767             felec            = _mm256_mul_ps(qq13,felec);
1768
1769             fscal            = felec;
1770
1771             /* Calculate temporary vectorial force */
1772             tx               = _mm256_mul_ps(fscal,dx13);
1773             ty               = _mm256_mul_ps(fscal,dy13);
1774             tz               = _mm256_mul_ps(fscal,dz13);
1775
1776             /* Update vectorial force */
1777             fix1             = _mm256_add_ps(fix1,tx);
1778             fiy1             = _mm256_add_ps(fiy1,ty);
1779             fiz1             = _mm256_add_ps(fiz1,tz);
1780
1781             fjx3             = _mm256_add_ps(fjx3,tx);
1782             fjy3             = _mm256_add_ps(fjy3,ty);
1783             fjz3             = _mm256_add_ps(fjz3,tz);
1784
1785             /**************************
1786              * CALCULATE INTERACTIONS *
1787              **************************/
1788
1789             r21              = _mm256_mul_ps(rsq21,rinv21);
1790
1791             /* EWALD ELECTROSTATICS */
1792
1793             /* Analytical PME correction */
1794             zeta2            = _mm256_mul_ps(beta2,rsq21);
1795             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1796             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1797             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1798             felec            = _mm256_mul_ps(qq21,felec);
1799
1800             fscal            = felec;
1801
1802             /* Calculate temporary vectorial force */
1803             tx               = _mm256_mul_ps(fscal,dx21);
1804             ty               = _mm256_mul_ps(fscal,dy21);
1805             tz               = _mm256_mul_ps(fscal,dz21);
1806
1807             /* Update vectorial force */
1808             fix2             = _mm256_add_ps(fix2,tx);
1809             fiy2             = _mm256_add_ps(fiy2,ty);
1810             fiz2             = _mm256_add_ps(fiz2,tz);
1811
1812             fjx1             = _mm256_add_ps(fjx1,tx);
1813             fjy1             = _mm256_add_ps(fjy1,ty);
1814             fjz1             = _mm256_add_ps(fjz1,tz);
1815
1816             /**************************
1817              * CALCULATE INTERACTIONS *
1818              **************************/
1819
1820             r22              = _mm256_mul_ps(rsq22,rinv22);
1821
1822             /* EWALD ELECTROSTATICS */
1823
1824             /* Analytical PME correction */
1825             zeta2            = _mm256_mul_ps(beta2,rsq22);
1826             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1827             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1828             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1829             felec            = _mm256_mul_ps(qq22,felec);
1830
1831             fscal            = felec;
1832
1833             /* Calculate temporary vectorial force */
1834             tx               = _mm256_mul_ps(fscal,dx22);
1835             ty               = _mm256_mul_ps(fscal,dy22);
1836             tz               = _mm256_mul_ps(fscal,dz22);
1837
1838             /* Update vectorial force */
1839             fix2             = _mm256_add_ps(fix2,tx);
1840             fiy2             = _mm256_add_ps(fiy2,ty);
1841             fiz2             = _mm256_add_ps(fiz2,tz);
1842
1843             fjx2             = _mm256_add_ps(fjx2,tx);
1844             fjy2             = _mm256_add_ps(fjy2,ty);
1845             fjz2             = _mm256_add_ps(fjz2,tz);
1846
1847             /**************************
1848              * CALCULATE INTERACTIONS *
1849              **************************/
1850
1851             r23              = _mm256_mul_ps(rsq23,rinv23);
1852
1853             /* EWALD ELECTROSTATICS */
1854
1855             /* Analytical PME correction */
1856             zeta2            = _mm256_mul_ps(beta2,rsq23);
1857             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1858             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1859             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1860             felec            = _mm256_mul_ps(qq23,felec);
1861
1862             fscal            = felec;
1863
1864             /* Calculate temporary vectorial force */
1865             tx               = _mm256_mul_ps(fscal,dx23);
1866             ty               = _mm256_mul_ps(fscal,dy23);
1867             tz               = _mm256_mul_ps(fscal,dz23);
1868
1869             /* Update vectorial force */
1870             fix2             = _mm256_add_ps(fix2,tx);
1871             fiy2             = _mm256_add_ps(fiy2,ty);
1872             fiz2             = _mm256_add_ps(fiz2,tz);
1873
1874             fjx3             = _mm256_add_ps(fjx3,tx);
1875             fjy3             = _mm256_add_ps(fjy3,ty);
1876             fjz3             = _mm256_add_ps(fjz3,tz);
1877
1878             /**************************
1879              * CALCULATE INTERACTIONS *
1880              **************************/
1881
1882             r31              = _mm256_mul_ps(rsq31,rinv31);
1883
1884             /* EWALD ELECTROSTATICS */
1885
1886             /* Analytical PME correction */
1887             zeta2            = _mm256_mul_ps(beta2,rsq31);
1888             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1889             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1890             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1891             felec            = _mm256_mul_ps(qq31,felec);
1892
1893             fscal            = felec;
1894
1895             /* Calculate temporary vectorial force */
1896             tx               = _mm256_mul_ps(fscal,dx31);
1897             ty               = _mm256_mul_ps(fscal,dy31);
1898             tz               = _mm256_mul_ps(fscal,dz31);
1899
1900             /* Update vectorial force */
1901             fix3             = _mm256_add_ps(fix3,tx);
1902             fiy3             = _mm256_add_ps(fiy3,ty);
1903             fiz3             = _mm256_add_ps(fiz3,tz);
1904
1905             fjx1             = _mm256_add_ps(fjx1,tx);
1906             fjy1             = _mm256_add_ps(fjy1,ty);
1907             fjz1             = _mm256_add_ps(fjz1,tz);
1908
1909             /**************************
1910              * CALCULATE INTERACTIONS *
1911              **************************/
1912
1913             r32              = _mm256_mul_ps(rsq32,rinv32);
1914
1915             /* EWALD ELECTROSTATICS */
1916
1917             /* Analytical PME correction */
1918             zeta2            = _mm256_mul_ps(beta2,rsq32);
1919             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1920             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1921             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1922             felec            = _mm256_mul_ps(qq32,felec);
1923
1924             fscal            = felec;
1925
1926             /* Calculate temporary vectorial force */
1927             tx               = _mm256_mul_ps(fscal,dx32);
1928             ty               = _mm256_mul_ps(fscal,dy32);
1929             tz               = _mm256_mul_ps(fscal,dz32);
1930
1931             /* Update vectorial force */
1932             fix3             = _mm256_add_ps(fix3,tx);
1933             fiy3             = _mm256_add_ps(fiy3,ty);
1934             fiz3             = _mm256_add_ps(fiz3,tz);
1935
1936             fjx2             = _mm256_add_ps(fjx2,tx);
1937             fjy2             = _mm256_add_ps(fjy2,ty);
1938             fjz2             = _mm256_add_ps(fjz2,tz);
1939
1940             /**************************
1941              * CALCULATE INTERACTIONS *
1942              **************************/
1943
1944             r33              = _mm256_mul_ps(rsq33,rinv33);
1945
1946             /* EWALD ELECTROSTATICS */
1947
1948             /* Analytical PME correction */
1949             zeta2            = _mm256_mul_ps(beta2,rsq33);
1950             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1951             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1952             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1953             felec            = _mm256_mul_ps(qq33,felec);
1954
1955             fscal            = felec;
1956
1957             /* Calculate temporary vectorial force */
1958             tx               = _mm256_mul_ps(fscal,dx33);
1959             ty               = _mm256_mul_ps(fscal,dy33);
1960             tz               = _mm256_mul_ps(fscal,dz33);
1961
1962             /* Update vectorial force */
1963             fix3             = _mm256_add_ps(fix3,tx);
1964             fiy3             = _mm256_add_ps(fiy3,ty);
1965             fiz3             = _mm256_add_ps(fiz3,tz);
1966
1967             fjx3             = _mm256_add_ps(fjx3,tx);
1968             fjy3             = _mm256_add_ps(fjy3,ty);
1969             fjz3             = _mm256_add_ps(fjz3,tz);
1970
1971             fjptrA             = f+j_coord_offsetA;
1972             fjptrB             = f+j_coord_offsetB;
1973             fjptrC             = f+j_coord_offsetC;
1974             fjptrD             = f+j_coord_offsetD;
1975             fjptrE             = f+j_coord_offsetE;
1976             fjptrF             = f+j_coord_offsetF;
1977             fjptrG             = f+j_coord_offsetG;
1978             fjptrH             = f+j_coord_offsetH;
1979
1980             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1981                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1982                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1983
1984             /* Inner loop uses 553 flops */
1985         }
1986
1987         if(jidx<j_index_end)
1988         {
1989
1990             /* Get j neighbor index, and coordinate index */
1991             jnrlistA         = jjnr[jidx];
1992             jnrlistB         = jjnr[jidx+1];
1993             jnrlistC         = jjnr[jidx+2];
1994             jnrlistD         = jjnr[jidx+3];
1995             jnrlistE         = jjnr[jidx+4];
1996             jnrlistF         = jjnr[jidx+5];
1997             jnrlistG         = jjnr[jidx+6];
1998             jnrlistH         = jjnr[jidx+7];
1999             /* Sign of each element will be negative for non-real atoms.
2000              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2001              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2002              */
2003             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
2004                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
2005
2006             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2007             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2008             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2009             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2010             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
2011             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
2012             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
2013             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
2014             j_coord_offsetA  = DIM*jnrA;
2015             j_coord_offsetB  = DIM*jnrB;
2016             j_coord_offsetC  = DIM*jnrC;
2017             j_coord_offsetD  = DIM*jnrD;
2018             j_coord_offsetE  = DIM*jnrE;
2019             j_coord_offsetF  = DIM*jnrF;
2020             j_coord_offsetG  = DIM*jnrG;
2021             j_coord_offsetH  = DIM*jnrH;
2022
2023             /* load j atom coordinates */
2024             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2025                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2026                                                  x+j_coord_offsetE,x+j_coord_offsetF,
2027                                                  x+j_coord_offsetG,x+j_coord_offsetH,
2028                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2029                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
2030
2031             /* Calculate displacement vector */
2032             dx00             = _mm256_sub_ps(ix0,jx0);
2033             dy00             = _mm256_sub_ps(iy0,jy0);
2034             dz00             = _mm256_sub_ps(iz0,jz0);
2035             dx11             = _mm256_sub_ps(ix1,jx1);
2036             dy11             = _mm256_sub_ps(iy1,jy1);
2037             dz11             = _mm256_sub_ps(iz1,jz1);
2038             dx12             = _mm256_sub_ps(ix1,jx2);
2039             dy12             = _mm256_sub_ps(iy1,jy2);
2040             dz12             = _mm256_sub_ps(iz1,jz2);
2041             dx13             = _mm256_sub_ps(ix1,jx3);
2042             dy13             = _mm256_sub_ps(iy1,jy3);
2043             dz13             = _mm256_sub_ps(iz1,jz3);
2044             dx21             = _mm256_sub_ps(ix2,jx1);
2045             dy21             = _mm256_sub_ps(iy2,jy1);
2046             dz21             = _mm256_sub_ps(iz2,jz1);
2047             dx22             = _mm256_sub_ps(ix2,jx2);
2048             dy22             = _mm256_sub_ps(iy2,jy2);
2049             dz22             = _mm256_sub_ps(iz2,jz2);
2050             dx23             = _mm256_sub_ps(ix2,jx3);
2051             dy23             = _mm256_sub_ps(iy2,jy3);
2052             dz23             = _mm256_sub_ps(iz2,jz3);
2053             dx31             = _mm256_sub_ps(ix3,jx1);
2054             dy31             = _mm256_sub_ps(iy3,jy1);
2055             dz31             = _mm256_sub_ps(iz3,jz1);
2056             dx32             = _mm256_sub_ps(ix3,jx2);
2057             dy32             = _mm256_sub_ps(iy3,jy2);
2058             dz32             = _mm256_sub_ps(iz3,jz2);
2059             dx33             = _mm256_sub_ps(ix3,jx3);
2060             dy33             = _mm256_sub_ps(iy3,jy3);
2061             dz33             = _mm256_sub_ps(iz3,jz3);
2062
2063             /* Calculate squared distance and things based on it */
2064             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
2065             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
2066             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
2067             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
2068             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
2069             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
2070             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
2071             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
2072             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
2073             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
2074
2075             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
2076             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
2077             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
2078             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
2079             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
2080             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
2081             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
2082             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
2083             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
2084             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
2085
2086             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
2087             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
2088             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
2089             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
2090             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
2091             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
2092             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
2093             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
2094             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
2095             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
2096
2097             fjx0             = _mm256_setzero_ps();
2098             fjy0             = _mm256_setzero_ps();
2099             fjz0             = _mm256_setzero_ps();
2100             fjx1             = _mm256_setzero_ps();
2101             fjy1             = _mm256_setzero_ps();
2102             fjz1             = _mm256_setzero_ps();
2103             fjx2             = _mm256_setzero_ps();
2104             fjy2             = _mm256_setzero_ps();
2105             fjz2             = _mm256_setzero_ps();
2106             fjx3             = _mm256_setzero_ps();
2107             fjy3             = _mm256_setzero_ps();
2108             fjz3             = _mm256_setzero_ps();
2109
2110             /**************************
2111              * CALCULATE INTERACTIONS *
2112              **************************/
2113
2114             r00              = _mm256_mul_ps(rsq00,rinv00);
2115             r00              = _mm256_andnot_ps(dummy_mask,r00);
2116
2117             /* Analytical LJ-PME */
2118             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2119             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
2120             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
2121             exponent         = gmx_simd_exp_r(ewcljrsq);
2122             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2123             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
2124             /* f6A = 6 * C6grid * (1 - poly) */
2125             f6A              = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
2126             /* f6B = C6grid * exponent * beta^6 */
2127             f6B              = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
2128             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2129             fvdw              = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2130
2131             fscal            = fvdw;
2132
2133             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2134
2135             /* Calculate temporary vectorial force */
2136             tx               = _mm256_mul_ps(fscal,dx00);
2137             ty               = _mm256_mul_ps(fscal,dy00);
2138             tz               = _mm256_mul_ps(fscal,dz00);
2139
2140             /* Update vectorial force */
2141             fix0             = _mm256_add_ps(fix0,tx);
2142             fiy0             = _mm256_add_ps(fiy0,ty);
2143             fiz0             = _mm256_add_ps(fiz0,tz);
2144
2145             fjx0             = _mm256_add_ps(fjx0,tx);
2146             fjy0             = _mm256_add_ps(fjy0,ty);
2147             fjz0             = _mm256_add_ps(fjz0,tz);
2148
2149             /**************************
2150              * CALCULATE INTERACTIONS *
2151              **************************/
2152
2153             r11              = _mm256_mul_ps(rsq11,rinv11);
2154             r11              = _mm256_andnot_ps(dummy_mask,r11);
2155
2156             /* EWALD ELECTROSTATICS */
2157
2158             /* Analytical PME correction */
2159             zeta2            = _mm256_mul_ps(beta2,rsq11);
2160             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
2161             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2162             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2163             felec            = _mm256_mul_ps(qq11,felec);
2164
2165             fscal            = felec;
2166
2167             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2168
2169             /* Calculate temporary vectorial force */
2170             tx               = _mm256_mul_ps(fscal,dx11);
2171             ty               = _mm256_mul_ps(fscal,dy11);
2172             tz               = _mm256_mul_ps(fscal,dz11);
2173
2174             /* Update vectorial force */
2175             fix1             = _mm256_add_ps(fix1,tx);
2176             fiy1             = _mm256_add_ps(fiy1,ty);
2177             fiz1             = _mm256_add_ps(fiz1,tz);
2178
2179             fjx1             = _mm256_add_ps(fjx1,tx);
2180             fjy1             = _mm256_add_ps(fjy1,ty);
2181             fjz1             = _mm256_add_ps(fjz1,tz);
2182
2183             /**************************
2184              * CALCULATE INTERACTIONS *
2185              **************************/
2186
2187             r12              = _mm256_mul_ps(rsq12,rinv12);
2188             r12              = _mm256_andnot_ps(dummy_mask,r12);
2189
2190             /* EWALD ELECTROSTATICS */
2191
2192             /* Analytical PME correction */
2193             zeta2            = _mm256_mul_ps(beta2,rsq12);
2194             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
2195             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2196             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2197             felec            = _mm256_mul_ps(qq12,felec);
2198
2199             fscal            = felec;
2200
2201             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2202
2203             /* Calculate temporary vectorial force */
2204             tx               = _mm256_mul_ps(fscal,dx12);
2205             ty               = _mm256_mul_ps(fscal,dy12);
2206             tz               = _mm256_mul_ps(fscal,dz12);
2207
2208             /* Update vectorial force */
2209             fix1             = _mm256_add_ps(fix1,tx);
2210             fiy1             = _mm256_add_ps(fiy1,ty);
2211             fiz1             = _mm256_add_ps(fiz1,tz);
2212
2213             fjx2             = _mm256_add_ps(fjx2,tx);
2214             fjy2             = _mm256_add_ps(fjy2,ty);
2215             fjz2             = _mm256_add_ps(fjz2,tz);
2216
2217             /**************************
2218              * CALCULATE INTERACTIONS *
2219              **************************/
2220
2221             r13              = _mm256_mul_ps(rsq13,rinv13);
2222             r13              = _mm256_andnot_ps(dummy_mask,r13);
2223
2224             /* EWALD ELECTROSTATICS */
2225
2226             /* Analytical PME correction */
2227             zeta2            = _mm256_mul_ps(beta2,rsq13);
2228             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
2229             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2230             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2231             felec            = _mm256_mul_ps(qq13,felec);
2232
2233             fscal            = felec;
2234
2235             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2236
2237             /* Calculate temporary vectorial force */
2238             tx               = _mm256_mul_ps(fscal,dx13);
2239             ty               = _mm256_mul_ps(fscal,dy13);
2240             tz               = _mm256_mul_ps(fscal,dz13);
2241
2242             /* Update vectorial force */
2243             fix1             = _mm256_add_ps(fix1,tx);
2244             fiy1             = _mm256_add_ps(fiy1,ty);
2245             fiz1             = _mm256_add_ps(fiz1,tz);
2246
2247             fjx3             = _mm256_add_ps(fjx3,tx);
2248             fjy3             = _mm256_add_ps(fjy3,ty);
2249             fjz3             = _mm256_add_ps(fjz3,tz);
2250
2251             /**************************
2252              * CALCULATE INTERACTIONS *
2253              **************************/
2254
2255             r21              = _mm256_mul_ps(rsq21,rinv21);
2256             r21              = _mm256_andnot_ps(dummy_mask,r21);
2257
2258             /* EWALD ELECTROSTATICS */
2259
2260             /* Analytical PME correction */
2261             zeta2            = _mm256_mul_ps(beta2,rsq21);
2262             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
2263             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2264             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2265             felec            = _mm256_mul_ps(qq21,felec);
2266
2267             fscal            = felec;
2268
2269             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2270
2271             /* Calculate temporary vectorial force */
2272             tx               = _mm256_mul_ps(fscal,dx21);
2273             ty               = _mm256_mul_ps(fscal,dy21);
2274             tz               = _mm256_mul_ps(fscal,dz21);
2275
2276             /* Update vectorial force */
2277             fix2             = _mm256_add_ps(fix2,tx);
2278             fiy2             = _mm256_add_ps(fiy2,ty);
2279             fiz2             = _mm256_add_ps(fiz2,tz);
2280
2281             fjx1             = _mm256_add_ps(fjx1,tx);
2282             fjy1             = _mm256_add_ps(fjy1,ty);
2283             fjz1             = _mm256_add_ps(fjz1,tz);
2284
2285             /**************************
2286              * CALCULATE INTERACTIONS *
2287              **************************/
2288
2289             r22              = _mm256_mul_ps(rsq22,rinv22);
2290             r22              = _mm256_andnot_ps(dummy_mask,r22);
2291
2292             /* EWALD ELECTROSTATICS */
2293
2294             /* Analytical PME correction */
2295             zeta2            = _mm256_mul_ps(beta2,rsq22);
2296             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
2297             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2298             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2299             felec            = _mm256_mul_ps(qq22,felec);
2300
2301             fscal            = felec;
2302
2303             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2304
2305             /* Calculate temporary vectorial force */
2306             tx               = _mm256_mul_ps(fscal,dx22);
2307             ty               = _mm256_mul_ps(fscal,dy22);
2308             tz               = _mm256_mul_ps(fscal,dz22);
2309
2310             /* Update vectorial force */
2311             fix2             = _mm256_add_ps(fix2,tx);
2312             fiy2             = _mm256_add_ps(fiy2,ty);
2313             fiz2             = _mm256_add_ps(fiz2,tz);
2314
2315             fjx2             = _mm256_add_ps(fjx2,tx);
2316             fjy2             = _mm256_add_ps(fjy2,ty);
2317             fjz2             = _mm256_add_ps(fjz2,tz);
2318
2319             /**************************
2320              * CALCULATE INTERACTIONS *
2321              **************************/
2322
2323             r23              = _mm256_mul_ps(rsq23,rinv23);
2324             r23              = _mm256_andnot_ps(dummy_mask,r23);
2325
2326             /* EWALD ELECTROSTATICS */
2327
2328             /* Analytical PME correction */
2329             zeta2            = _mm256_mul_ps(beta2,rsq23);
2330             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
2331             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2332             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2333             felec            = _mm256_mul_ps(qq23,felec);
2334
2335             fscal            = felec;
2336
2337             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2338
2339             /* Calculate temporary vectorial force */
2340             tx               = _mm256_mul_ps(fscal,dx23);
2341             ty               = _mm256_mul_ps(fscal,dy23);
2342             tz               = _mm256_mul_ps(fscal,dz23);
2343
2344             /* Update vectorial force */
2345             fix2             = _mm256_add_ps(fix2,tx);
2346             fiy2             = _mm256_add_ps(fiy2,ty);
2347             fiz2             = _mm256_add_ps(fiz2,tz);
2348
2349             fjx3             = _mm256_add_ps(fjx3,tx);
2350             fjy3             = _mm256_add_ps(fjy3,ty);
2351             fjz3             = _mm256_add_ps(fjz3,tz);
2352
2353             /**************************
2354              * CALCULATE INTERACTIONS *
2355              **************************/
2356
2357             r31              = _mm256_mul_ps(rsq31,rinv31);
2358             r31              = _mm256_andnot_ps(dummy_mask,r31);
2359
2360             /* EWALD ELECTROSTATICS */
2361
2362             /* Analytical PME correction */
2363             zeta2            = _mm256_mul_ps(beta2,rsq31);
2364             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
2365             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2366             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2367             felec            = _mm256_mul_ps(qq31,felec);
2368
2369             fscal            = felec;
2370
2371             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2372
2373             /* Calculate temporary vectorial force */
2374             tx               = _mm256_mul_ps(fscal,dx31);
2375             ty               = _mm256_mul_ps(fscal,dy31);
2376             tz               = _mm256_mul_ps(fscal,dz31);
2377
2378             /* Update vectorial force */
2379             fix3             = _mm256_add_ps(fix3,tx);
2380             fiy3             = _mm256_add_ps(fiy3,ty);
2381             fiz3             = _mm256_add_ps(fiz3,tz);
2382
2383             fjx1             = _mm256_add_ps(fjx1,tx);
2384             fjy1             = _mm256_add_ps(fjy1,ty);
2385             fjz1             = _mm256_add_ps(fjz1,tz);
2386
2387             /**************************
2388              * CALCULATE INTERACTIONS *
2389              **************************/
2390
2391             r32              = _mm256_mul_ps(rsq32,rinv32);
2392             r32              = _mm256_andnot_ps(dummy_mask,r32);
2393
2394             /* EWALD ELECTROSTATICS */
2395
2396             /* Analytical PME correction */
2397             zeta2            = _mm256_mul_ps(beta2,rsq32);
2398             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
2399             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2400             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2401             felec            = _mm256_mul_ps(qq32,felec);
2402
2403             fscal            = felec;
2404
2405             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2406
2407             /* Calculate temporary vectorial force */
2408             tx               = _mm256_mul_ps(fscal,dx32);
2409             ty               = _mm256_mul_ps(fscal,dy32);
2410             tz               = _mm256_mul_ps(fscal,dz32);
2411
2412             /* Update vectorial force */
2413             fix3             = _mm256_add_ps(fix3,tx);
2414             fiy3             = _mm256_add_ps(fiy3,ty);
2415             fiz3             = _mm256_add_ps(fiz3,tz);
2416
2417             fjx2             = _mm256_add_ps(fjx2,tx);
2418             fjy2             = _mm256_add_ps(fjy2,ty);
2419             fjz2             = _mm256_add_ps(fjz2,tz);
2420
2421             /**************************
2422              * CALCULATE INTERACTIONS *
2423              **************************/
2424
2425             r33              = _mm256_mul_ps(rsq33,rinv33);
2426             r33              = _mm256_andnot_ps(dummy_mask,r33);
2427
2428             /* EWALD ELECTROSTATICS */
2429
2430             /* Analytical PME correction */
2431             zeta2            = _mm256_mul_ps(beta2,rsq33);
2432             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
2433             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2434             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2435             felec            = _mm256_mul_ps(qq33,felec);
2436
2437             fscal            = felec;
2438
2439             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2440
2441             /* Calculate temporary vectorial force */
2442             tx               = _mm256_mul_ps(fscal,dx33);
2443             ty               = _mm256_mul_ps(fscal,dy33);
2444             tz               = _mm256_mul_ps(fscal,dz33);
2445
2446             /* Update vectorial force */
2447             fix3             = _mm256_add_ps(fix3,tx);
2448             fiy3             = _mm256_add_ps(fiy3,ty);
2449             fiz3             = _mm256_add_ps(fiz3,tz);
2450
2451             fjx3             = _mm256_add_ps(fjx3,tx);
2452             fjy3             = _mm256_add_ps(fjy3,ty);
2453             fjz3             = _mm256_add_ps(fjz3,tz);
2454
2455             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2456             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2457             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2458             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2459             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2460             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2461             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2462             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2463
2464             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2465                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2466                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2467
2468             /* Inner loop uses 563 flops */
2469         }
2470
2471         /* End of innermost loop */
2472
2473         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2474                                                  f+i_coord_offset,fshift+i_shift_offset);
2475
2476         /* Increment number of inner iterations */
2477         inneriter                  += j_index_end - j_index_start;
2478
2479         /* Outer loop uses 24 flops */
2480     }
2481
2482     /* Increment number of outer iterations */
2483     outeriter        += nri;
2484
2485     /* Update outer/inner flops */
2486
2487     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*563);
2488 }