src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEw_VdwLJ_GeomW4W4_avx_256_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_single.h"
  34 #include "kernelutil_x86_avx_256_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrE,jnrF,jnrG,jnrH;
  62     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  63     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  64     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  65     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  66     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  67     real             rcutoff_scalar;
  68     real             *shiftvec,*fshift,*x,*f;
  69     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  70     real             scratch[4*DIM];
  71     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  72     real *           vdwioffsetptr0;
  73     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  74     real *           vdwioffsetptr1;
  75     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  76     real *           vdwioffsetptr2;
  77     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  78     real *           vdwioffsetptr3;
  79     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  80     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  81     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  82     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  83     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  84     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  85     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  86     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  87     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  88     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  89     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  90     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  91     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  92     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  93     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  94     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  95     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  96     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  97     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  98     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  99     real             *charge;
 100     int              nvdwtype;
 101     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 102     int              *vdwtype;
 103     real             *vdwparam;
 104     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 105     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 106     __m256i          ewitab;
 107     __m128i          ewitab_lo,ewitab_hi;
 108     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 109     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 110     real             *ewtab;
 111     __m256           dummy_mask,cutoff_mask;
 112     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 113     __m256           one     = _mm256_set1_ps(1.0);
 114     __m256           two     = _mm256_set1_ps(2.0);
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = _mm256_set1_ps(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128     nvdwtype         = fr->ntype;
 129     vdwparam         = fr->nbfp;
 130     vdwtype          = mdatoms->typeA;
 131
 132     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 133     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff);
 134     beta2            = _mm256_mul_ps(beta,beta);
 135     beta3            = _mm256_mul_ps(beta,beta2);
 136
 137     ewtab            = fr->ic->tabq_coul_FDV0;
 138     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 139     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 140
 141     /* Setup water-specific parameters */
 142     inr              = nlist->iinr[0];
 143     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 144     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 145     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 146     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 147
 148     jq1              = _mm256_set1_ps(charge[inr+1]);
 149     jq2              = _mm256_set1_ps(charge[inr+2]);
 150     jq3              = _mm256_set1_ps(charge[inr+3]);
 151     vdwjidx0A        = 2*vdwtype[inr+0];
 152     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 153     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 154     qq11             = _mm256_mul_ps(iq1,jq1);
 155     qq12             = _mm256_mul_ps(iq1,jq2);
 156     qq13             = _mm256_mul_ps(iq1,jq3);
 157     qq21             = _mm256_mul_ps(iq2,jq1);
 158     qq22             = _mm256_mul_ps(iq2,jq2);
 159     qq23             = _mm256_mul_ps(iq2,jq3);
 160     qq31             = _mm256_mul_ps(iq3,jq1);
 161     qq32             = _mm256_mul_ps(iq3,jq2);
 162     qq33             = _mm256_mul_ps(iq3,jq3);
 163
 164     /* Avoid stupid compiler warnings */
 165     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 166     j_coord_offsetA = 0;
 167     j_coord_offsetB = 0;
 168     j_coord_offsetC = 0;
 169     j_coord_offsetD = 0;
 170     j_coord_offsetE = 0;
 171     j_coord_offsetF = 0;
 172     j_coord_offsetG = 0;
 173     j_coord_offsetH = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm256_setzero_ps();
 202         fiy0             = _mm256_setzero_ps();
 203         fiz0             = _mm256_setzero_ps();
 204         fix1             = _mm256_setzero_ps();
 205         fiy1             = _mm256_setzero_ps();
 206         fiz1             = _mm256_setzero_ps();
 207         fix2             = _mm256_setzero_ps();
 208         fiy2             = _mm256_setzero_ps();
 209         fiz2             = _mm256_setzero_ps();
 210         fix3             = _mm256_setzero_ps();
 211         fiy3             = _mm256_setzero_ps();
 212         fiz3             = _mm256_setzero_ps();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm256_setzero_ps();
 216         vvdwsum          = _mm256_setzero_ps();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             jnrC             = jjnr[jidx+2];
 226             jnrD             = jjnr[jidx+3];
 227             jnrE             = jjnr[jidx+4];
 228             jnrF             = jjnr[jidx+5];
 229             jnrG             = jjnr[jidx+6];
 230             jnrH             = jjnr[jidx+7];
 231             j_coord_offsetA  = DIM*jnrA;
 232             j_coord_offsetB  = DIM*jnrB;
 233             j_coord_offsetC  = DIM*jnrC;
 234             j_coord_offsetD  = DIM*jnrD;
 235             j_coord_offsetE  = DIM*jnrE;
 236             j_coord_offsetF  = DIM*jnrF;
 237             j_coord_offsetG  = DIM*jnrG;
 238             j_coord_offsetH  = DIM*jnrH;
 239
 240             /* load j atom coordinates */
 241             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 242                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 243                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 244                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 245                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 246                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 247
 248             /* Calculate displacement vector */
 249             dx00             = _mm256_sub_ps(ix0,jx0);
 250             dy00             = _mm256_sub_ps(iy0,jy0);
 251             dz00             = _mm256_sub_ps(iz0,jz0);
 252             dx11             = _mm256_sub_ps(ix1,jx1);
 253             dy11             = _mm256_sub_ps(iy1,jy1);
 254             dz11             = _mm256_sub_ps(iz1,jz1);
 255             dx12             = _mm256_sub_ps(ix1,jx2);
 256             dy12             = _mm256_sub_ps(iy1,jy2);
 257             dz12             = _mm256_sub_ps(iz1,jz2);
 258             dx13             = _mm256_sub_ps(ix1,jx3);
 259             dy13             = _mm256_sub_ps(iy1,jy3);
 260             dz13             = _mm256_sub_ps(iz1,jz3);
 261             dx21             = _mm256_sub_ps(ix2,jx1);
 262             dy21             = _mm256_sub_ps(iy2,jy1);
 263             dz21             = _mm256_sub_ps(iz2,jz1);
 264             dx22             = _mm256_sub_ps(ix2,jx2);
 265             dy22             = _mm256_sub_ps(iy2,jy2);
 266             dz22             = _mm256_sub_ps(iz2,jz2);
 267             dx23             = _mm256_sub_ps(ix2,jx3);
 268             dy23             = _mm256_sub_ps(iy2,jy3);
 269             dz23             = _mm256_sub_ps(iz2,jz3);
 270             dx31             = _mm256_sub_ps(ix3,jx1);
 271             dy31             = _mm256_sub_ps(iy3,jy1);
 272             dz31             = _mm256_sub_ps(iz3,jz1);
 273             dx32             = _mm256_sub_ps(ix3,jx2);
 274             dy32             = _mm256_sub_ps(iy3,jy2);
 275             dz32             = _mm256_sub_ps(iz3,jz2);
 276             dx33             = _mm256_sub_ps(ix3,jx3);
 277             dy33             = _mm256_sub_ps(iy3,jy3);
 278             dz33             = _mm256_sub_ps(iz3,jz3);
 279
 280             /* Calculate squared distance and things based on it */
 281             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 282             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 283             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 284             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 285             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 286             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 287             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 288             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 289             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 290             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 291
 292             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 293             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 294             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 295             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 296             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 297             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 298             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 299             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 300             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 301
 302             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 303             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 304             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 305             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 306             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 307             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 308             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 309             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 310             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 311             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 312
 313             fjx0             = _mm256_setzero_ps();
 314             fjy0             = _mm256_setzero_ps();
 315             fjz0             = _mm256_setzero_ps();
 316             fjx1             = _mm256_setzero_ps();
 317             fjy1             = _mm256_setzero_ps();
 318             fjz1             = _mm256_setzero_ps();
 319             fjx2             = _mm256_setzero_ps();
 320             fjy2             = _mm256_setzero_ps();
 321             fjz2             = _mm256_setzero_ps();
 322             fjx3             = _mm256_setzero_ps();
 323             fjy3             = _mm256_setzero_ps();
 324             fjz3             = _mm256_setzero_ps();
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             /* LENNARD-JONES DISPERSION/REPULSION */
 331
 332             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 333             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 334             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 335             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 336             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 340
 341             fscal            = fvdw;
 342
 343             /* Calculate temporary vectorial force */
 344             tx               = _mm256_mul_ps(fscal,dx00);
 345             ty               = _mm256_mul_ps(fscal,dy00);
 346             tz               = _mm256_mul_ps(fscal,dz00);
 347
 348             /* Update vectorial force */
 349             fix0             = _mm256_add_ps(fix0,tx);
 350             fiy0             = _mm256_add_ps(fiy0,ty);
 351             fiz0             = _mm256_add_ps(fiz0,tz);
 352
 353             fjx0             = _mm256_add_ps(fjx0,tx);
 354             fjy0             = _mm256_add_ps(fjy0,ty);
 355             fjz0             = _mm256_add_ps(fjz0,tz);
 356
 357             /**************************
 358              * CALCULATE INTERACTIONS *
 359              **************************/
 360
 361             r11              = _mm256_mul_ps(rsq11,rinv11);
 362
 363             /* EWALD ELECTROSTATICS */
 364
 365             /* Analytical PME correction */
 366             zeta2            = _mm256_mul_ps(beta2,rsq11);
 367             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 368             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 369             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 370             felec            = _mm256_mul_ps(qq11,felec);
 371             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 372             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 373             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 374             velec            = _mm256_mul_ps(qq11,velec);
 375
 376             /* Update potential sum for this i atom from the interaction with this j atom. */
 377             velecsum         = _mm256_add_ps(velecsum,velec);
 378
 379             fscal            = felec;
 380
 381             /* Calculate temporary vectorial force */
 382             tx               = _mm256_mul_ps(fscal,dx11);
 383             ty               = _mm256_mul_ps(fscal,dy11);
 384             tz               = _mm256_mul_ps(fscal,dz11);
 385
 386             /* Update vectorial force */
 387             fix1             = _mm256_add_ps(fix1,tx);
 388             fiy1             = _mm256_add_ps(fiy1,ty);
 389             fiz1             = _mm256_add_ps(fiz1,tz);
 390
 391             fjx1             = _mm256_add_ps(fjx1,tx);
 392             fjy1             = _mm256_add_ps(fjy1,ty);
 393             fjz1             = _mm256_add_ps(fjz1,tz);
 394
 395             /**************************
 396              * CALCULATE INTERACTIONS *
 397              **************************/
 398
 399             r12              = _mm256_mul_ps(rsq12,rinv12);
 400
 401             /* EWALD ELECTROSTATICS */
 402
 403             /* Analytical PME correction */
 404             zeta2            = _mm256_mul_ps(beta2,rsq12);
 405             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 406             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 407             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 408             felec            = _mm256_mul_ps(qq12,felec);
 409             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 410             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 411             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 412             velec            = _mm256_mul_ps(qq12,velec);
 413
 414             /* Update potential sum for this i atom from the interaction with this j atom. */
 415             velecsum         = _mm256_add_ps(velecsum,velec);
 416
 417             fscal            = felec;
 418
 419             /* Calculate temporary vectorial force */
 420             tx               = _mm256_mul_ps(fscal,dx12);
 421             ty               = _mm256_mul_ps(fscal,dy12);
 422             tz               = _mm256_mul_ps(fscal,dz12);
 423
 424             /* Update vectorial force */
 425             fix1             = _mm256_add_ps(fix1,tx);
 426             fiy1             = _mm256_add_ps(fiy1,ty);
 427             fiz1             = _mm256_add_ps(fiz1,tz);
 428
 429             fjx2             = _mm256_add_ps(fjx2,tx);
 430             fjy2             = _mm256_add_ps(fjy2,ty);
 431             fjz2             = _mm256_add_ps(fjz2,tz);
 432
 433             /**************************
 434              * CALCULATE INTERACTIONS *
 435              **************************/
 436
 437             r13              = _mm256_mul_ps(rsq13,rinv13);
 438
 439             /* EWALD ELECTROSTATICS */
 440
 441             /* Analytical PME correction */
 442             zeta2            = _mm256_mul_ps(beta2,rsq13);
 443             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
 444             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 445             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 446             felec            = _mm256_mul_ps(qq13,felec);
 447             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 448             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 449             velec            = _mm256_sub_ps(rinv13,pmecorrV);
 450             velec            = _mm256_mul_ps(qq13,velec);
 451
 452             /* Update potential sum for this i atom from the interaction with this j atom. */
 453             velecsum         = _mm256_add_ps(velecsum,velec);
 454
 455             fscal            = felec;
 456
 457             /* Calculate temporary vectorial force */
 458             tx               = _mm256_mul_ps(fscal,dx13);
 459             ty               = _mm256_mul_ps(fscal,dy13);
 460             tz               = _mm256_mul_ps(fscal,dz13);
 461
 462             /* Update vectorial force */
 463             fix1             = _mm256_add_ps(fix1,tx);
 464             fiy1             = _mm256_add_ps(fiy1,ty);
 465             fiz1             = _mm256_add_ps(fiz1,tz);
 466
 467             fjx3             = _mm256_add_ps(fjx3,tx);
 468             fjy3             = _mm256_add_ps(fjy3,ty);
 469             fjz3             = _mm256_add_ps(fjz3,tz);
 470
 471             /**************************
 472              * CALCULATE INTERACTIONS *
 473              **************************/
 474
 475             r21              = _mm256_mul_ps(rsq21,rinv21);
 476
 477             /* EWALD ELECTROSTATICS */
 478
 479             /* Analytical PME correction */
 480             zeta2            = _mm256_mul_ps(beta2,rsq21);
 481             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
 482             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 483             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 484             felec            = _mm256_mul_ps(qq21,felec);
 485             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 486             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 487             velec            = _mm256_sub_ps(rinv21,pmecorrV);
 488             velec            = _mm256_mul_ps(qq21,velec);
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velecsum         = _mm256_add_ps(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm256_mul_ps(fscal,dx21);
 497             ty               = _mm256_mul_ps(fscal,dy21);
 498             tz               = _mm256_mul_ps(fscal,dz21);
 499
 500             /* Update vectorial force */
 501             fix2             = _mm256_add_ps(fix2,tx);
 502             fiy2             = _mm256_add_ps(fiy2,ty);
 503             fiz2             = _mm256_add_ps(fiz2,tz);
 504
 505             fjx1             = _mm256_add_ps(fjx1,tx);
 506             fjy1             = _mm256_add_ps(fjy1,ty);
 507             fjz1             = _mm256_add_ps(fjz1,tz);
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             r22              = _mm256_mul_ps(rsq22,rinv22);
 514
 515             /* EWALD ELECTROSTATICS */
 516
 517             /* Analytical PME correction */
 518             zeta2            = _mm256_mul_ps(beta2,rsq22);
 519             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
 520             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 521             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 522             felec            = _mm256_mul_ps(qq22,felec);
 523             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 524             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 525             velec            = _mm256_sub_ps(rinv22,pmecorrV);
 526             velec            = _mm256_mul_ps(qq22,velec);
 527
 528             /* Update potential sum for this i atom from the interaction with this j atom. */
 529             velecsum         = _mm256_add_ps(velecsum,velec);
 530
 531             fscal            = felec;
 532
 533             /* Calculate temporary vectorial force */
 534             tx               = _mm256_mul_ps(fscal,dx22);
 535             ty               = _mm256_mul_ps(fscal,dy22);
 536             tz               = _mm256_mul_ps(fscal,dz22);
 537
 538             /* Update vectorial force */
 539             fix2             = _mm256_add_ps(fix2,tx);
 540             fiy2             = _mm256_add_ps(fiy2,ty);
 541             fiz2             = _mm256_add_ps(fiz2,tz);
 542
 543             fjx2             = _mm256_add_ps(fjx2,tx);
 544             fjy2             = _mm256_add_ps(fjy2,ty);
 545             fjz2             = _mm256_add_ps(fjz2,tz);
 546
 547             /**************************
 548              * CALCULATE INTERACTIONS *
 549              **************************/
 550
 551             r23              = _mm256_mul_ps(rsq23,rinv23);
 552
 553             /* EWALD ELECTROSTATICS */
 554
 555             /* Analytical PME correction */
 556             zeta2            = _mm256_mul_ps(beta2,rsq23);
 557             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
 558             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 559             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 560             felec            = _mm256_mul_ps(qq23,felec);
 561             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 562             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 563             velec            = _mm256_sub_ps(rinv23,pmecorrV);
 564             velec            = _mm256_mul_ps(qq23,velec);
 565
 566             /* Update potential sum for this i atom from the interaction with this j atom. */
 567             velecsum         = _mm256_add_ps(velecsum,velec);
 568
 569             fscal            = felec;
 570
 571             /* Calculate temporary vectorial force */
 572             tx               = _mm256_mul_ps(fscal,dx23);
 573             ty               = _mm256_mul_ps(fscal,dy23);
 574             tz               = _mm256_mul_ps(fscal,dz23);
 575
 576             /* Update vectorial force */
 577             fix2             = _mm256_add_ps(fix2,tx);
 578             fiy2             = _mm256_add_ps(fiy2,ty);
 579             fiz2             = _mm256_add_ps(fiz2,tz);
 580
 581             fjx3             = _mm256_add_ps(fjx3,tx);
 582             fjy3             = _mm256_add_ps(fjy3,ty);
 583             fjz3             = _mm256_add_ps(fjz3,tz);
 584
 585             /**************************
 586              * CALCULATE INTERACTIONS *
 587              **************************/
 588
 589             r31              = _mm256_mul_ps(rsq31,rinv31);
 590
 591             /* EWALD ELECTROSTATICS */
 592
 593             /* Analytical PME correction */
 594             zeta2            = _mm256_mul_ps(beta2,rsq31);
 595             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
 596             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 597             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 598             felec            = _mm256_mul_ps(qq31,felec);
 599             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 600             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 601             velec            = _mm256_sub_ps(rinv31,pmecorrV);
 602             velec            = _mm256_mul_ps(qq31,velec);
 603
 604             /* Update potential sum for this i atom from the interaction with this j atom. */
 605             velecsum         = _mm256_add_ps(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             /* Calculate temporary vectorial force */
 610             tx               = _mm256_mul_ps(fscal,dx31);
 611             ty               = _mm256_mul_ps(fscal,dy31);
 612             tz               = _mm256_mul_ps(fscal,dz31);
 613
 614             /* Update vectorial force */
 615             fix3             = _mm256_add_ps(fix3,tx);
 616             fiy3             = _mm256_add_ps(fiy3,ty);
 617             fiz3             = _mm256_add_ps(fiz3,tz);
 618
 619             fjx1             = _mm256_add_ps(fjx1,tx);
 620             fjy1             = _mm256_add_ps(fjy1,ty);
 621             fjz1             = _mm256_add_ps(fjz1,tz);
 622
 623             /**************************
 624              * CALCULATE INTERACTIONS *
 625              **************************/
 626
 627             r32              = _mm256_mul_ps(rsq32,rinv32);
 628
 629             /* EWALD ELECTROSTATICS */
 630
 631             /* Analytical PME correction */
 632             zeta2            = _mm256_mul_ps(beta2,rsq32);
 633             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
 634             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 635             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 636             felec            = _mm256_mul_ps(qq32,felec);
 637             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 638             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 639             velec            = _mm256_sub_ps(rinv32,pmecorrV);
 640             velec            = _mm256_mul_ps(qq32,velec);
 641
 642             /* Update potential sum for this i atom from the interaction with this j atom. */
 643             velecsum         = _mm256_add_ps(velecsum,velec);
 644
 645             fscal            = felec;
 646
 647             /* Calculate temporary vectorial force */
 648             tx               = _mm256_mul_ps(fscal,dx32);
 649             ty               = _mm256_mul_ps(fscal,dy32);
 650             tz               = _mm256_mul_ps(fscal,dz32);
 651
 652             /* Update vectorial force */
 653             fix3             = _mm256_add_ps(fix3,tx);
 654             fiy3             = _mm256_add_ps(fiy3,ty);
 655             fiz3             = _mm256_add_ps(fiz3,tz);
 656
 657             fjx2             = _mm256_add_ps(fjx2,tx);
 658             fjy2             = _mm256_add_ps(fjy2,ty);
 659             fjz2             = _mm256_add_ps(fjz2,tz);
 660
 661             /**************************
 662              * CALCULATE INTERACTIONS *
 663              **************************/
 664
 665             r33              = _mm256_mul_ps(rsq33,rinv33);
 666
 667             /* EWALD ELECTROSTATICS */
 668
 669             /* Analytical PME correction */
 670             zeta2            = _mm256_mul_ps(beta2,rsq33);
 671             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
 672             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 673             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 674             felec            = _mm256_mul_ps(qq33,felec);
 675             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 676             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 677             velec            = _mm256_sub_ps(rinv33,pmecorrV);
 678             velec            = _mm256_mul_ps(qq33,velec);
 679
 680             /* Update potential sum for this i atom from the interaction with this j atom. */
 681             velecsum         = _mm256_add_ps(velecsum,velec);
 682
 683             fscal            = felec;
 684
 685             /* Calculate temporary vectorial force */
 686             tx               = _mm256_mul_ps(fscal,dx33);
 687             ty               = _mm256_mul_ps(fscal,dy33);
 688             tz               = _mm256_mul_ps(fscal,dz33);
 689
 690             /* Update vectorial force */
 691             fix3             = _mm256_add_ps(fix3,tx);
 692             fiy3             = _mm256_add_ps(fiy3,ty);
 693             fiz3             = _mm256_add_ps(fiz3,tz);
 694
 695             fjx3             = _mm256_add_ps(fjx3,tx);
 696             fjy3             = _mm256_add_ps(fjy3,ty);
 697             fjz3             = _mm256_add_ps(fjz3,tz);
 698
 699             fjptrA             = f+j_coord_offsetA;
 700             fjptrB             = f+j_coord_offsetB;
 701             fjptrC             = f+j_coord_offsetC;
 702             fjptrD             = f+j_coord_offsetD;
 703             fjptrE             = f+j_coord_offsetE;
 704             fjptrF             = f+j_coord_offsetF;
 705             fjptrG             = f+j_coord_offsetG;
 706             fjptrH             = f+j_coord_offsetH;
 707
 708             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 709                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 710                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 711
 712             /* Inner loop uses 791 flops */
 713         }
 714
 715         if(jidx<j_index_end)
 716         {
 717
 718             /* Get j neighbor index, and coordinate index */
 719             jnrlistA         = jjnr[jidx];
 720             jnrlistB         = jjnr[jidx+1];
 721             jnrlistC         = jjnr[jidx+2];
 722             jnrlistD         = jjnr[jidx+3];
 723             jnrlistE         = jjnr[jidx+4];
 724             jnrlistF         = jjnr[jidx+5];
 725             jnrlistG         = jjnr[jidx+6];
 726             jnrlistH         = jjnr[jidx+7];
 727             /* Sign of each element will be negative for non-real atoms.
 728              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 729              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 730              */
 731             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 732                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 733
 734             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 735             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 736             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 737             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 738             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 739             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 740             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 741             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 742             j_coord_offsetA  = DIM*jnrA;
 743             j_coord_offsetB  = DIM*jnrB;
 744             j_coord_offsetC  = DIM*jnrC;
 745             j_coord_offsetD  = DIM*jnrD;
 746             j_coord_offsetE  = DIM*jnrE;
 747             j_coord_offsetF  = DIM*jnrF;
 748             j_coord_offsetG  = DIM*jnrG;
 749             j_coord_offsetH  = DIM*jnrH;
 750
 751             /* load j atom coordinates */
 752             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 753                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 754                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 755                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 756                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 757                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 758
 759             /* Calculate displacement vector */
 760             dx00             = _mm256_sub_ps(ix0,jx0);
 761             dy00             = _mm256_sub_ps(iy0,jy0);
 762             dz00             = _mm256_sub_ps(iz0,jz0);
 763             dx11             = _mm256_sub_ps(ix1,jx1);
 764             dy11             = _mm256_sub_ps(iy1,jy1);
 765             dz11             = _mm256_sub_ps(iz1,jz1);
 766             dx12             = _mm256_sub_ps(ix1,jx2);
 767             dy12             = _mm256_sub_ps(iy1,jy2);
 768             dz12             = _mm256_sub_ps(iz1,jz2);
 769             dx13             = _mm256_sub_ps(ix1,jx3);
 770             dy13             = _mm256_sub_ps(iy1,jy3);
 771             dz13             = _mm256_sub_ps(iz1,jz3);
 772             dx21             = _mm256_sub_ps(ix2,jx1);
 773             dy21             = _mm256_sub_ps(iy2,jy1);
 774             dz21             = _mm256_sub_ps(iz2,jz1);
 775             dx22             = _mm256_sub_ps(ix2,jx2);
 776             dy22             = _mm256_sub_ps(iy2,jy2);
 777             dz22             = _mm256_sub_ps(iz2,jz2);
 778             dx23             = _mm256_sub_ps(ix2,jx3);
 779             dy23             = _mm256_sub_ps(iy2,jy3);
 780             dz23             = _mm256_sub_ps(iz2,jz3);
 781             dx31             = _mm256_sub_ps(ix3,jx1);
 782             dy31             = _mm256_sub_ps(iy3,jy1);
 783             dz31             = _mm256_sub_ps(iz3,jz1);
 784             dx32             = _mm256_sub_ps(ix3,jx2);
 785             dy32             = _mm256_sub_ps(iy3,jy2);
 786             dz32             = _mm256_sub_ps(iz3,jz2);
 787             dx33             = _mm256_sub_ps(ix3,jx3);
 788             dy33             = _mm256_sub_ps(iy3,jy3);
 789             dz33             = _mm256_sub_ps(iz3,jz3);
 790
 791             /* Calculate squared distance and things based on it */
 792             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 793             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 794             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 795             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 796             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 797             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 798             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 799             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 800             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 801             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 802
 803             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 804             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 805             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 806             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 807             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 808             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 809             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 810             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 811             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 812
 813             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 814             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 815             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 816             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 817             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 818             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 819             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 820             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 821             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 822             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 823
 824             fjx0             = _mm256_setzero_ps();
 825             fjy0             = _mm256_setzero_ps();
 826             fjz0             = _mm256_setzero_ps();
 827             fjx1             = _mm256_setzero_ps();
 828             fjy1             = _mm256_setzero_ps();
 829             fjz1             = _mm256_setzero_ps();
 830             fjx2             = _mm256_setzero_ps();
 831             fjy2             = _mm256_setzero_ps();
 832             fjz2             = _mm256_setzero_ps();
 833             fjx3             = _mm256_setzero_ps();
 834             fjy3             = _mm256_setzero_ps();
 835             fjz3             = _mm256_setzero_ps();
 836
 837             /**************************
 838              * CALCULATE INTERACTIONS *
 839              **************************/
 840
 841             /* LENNARD-JONES DISPERSION/REPULSION */
 842
 843             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 844             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 845             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 846             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 847             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 848
 849             /* Update potential sum for this i atom from the interaction with this j atom. */
 850             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 851             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 852
 853             fscal            = fvdw;
 854
 855             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 856
 857             /* Calculate temporary vectorial force */
 858             tx               = _mm256_mul_ps(fscal,dx00);
 859             ty               = _mm256_mul_ps(fscal,dy00);
 860             tz               = _mm256_mul_ps(fscal,dz00);
 861
 862             /* Update vectorial force */
 863             fix0             = _mm256_add_ps(fix0,tx);
 864             fiy0             = _mm256_add_ps(fiy0,ty);
 865             fiz0             = _mm256_add_ps(fiz0,tz);
 866
 867             fjx0             = _mm256_add_ps(fjx0,tx);
 868             fjy0             = _mm256_add_ps(fjy0,ty);
 869             fjz0             = _mm256_add_ps(fjz0,tz);
 870
 871             /**************************
 872              * CALCULATE INTERACTIONS *
 873              **************************/
 874
 875             r11              = _mm256_mul_ps(rsq11,rinv11);
 876             r11              = _mm256_andnot_ps(dummy_mask,r11);
 877
 878             /* EWALD ELECTROSTATICS */
 879
 880             /* Analytical PME correction */
 881             zeta2            = _mm256_mul_ps(beta2,rsq11);
 882             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 883             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 884             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 885             felec            = _mm256_mul_ps(qq11,felec);
 886             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 887             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 888             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 889             velec            = _mm256_mul_ps(qq11,velec);
 890
 891             /* Update potential sum for this i atom from the interaction with this j atom. */
 892             velec            = _mm256_andnot_ps(dummy_mask,velec);
 893             velecsum         = _mm256_add_ps(velecsum,velec);
 894
 895             fscal            = felec;
 896
 897             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 898
 899             /* Calculate temporary vectorial force */
 900             tx               = _mm256_mul_ps(fscal,dx11);
 901             ty               = _mm256_mul_ps(fscal,dy11);
 902             tz               = _mm256_mul_ps(fscal,dz11);
 903
 904             /* Update vectorial force */
 905             fix1             = _mm256_add_ps(fix1,tx);
 906             fiy1             = _mm256_add_ps(fiy1,ty);
 907             fiz1             = _mm256_add_ps(fiz1,tz);
 908
 909             fjx1             = _mm256_add_ps(fjx1,tx);
 910             fjy1             = _mm256_add_ps(fjy1,ty);
 911             fjz1             = _mm256_add_ps(fjz1,tz);
 912
 913             /**************************
 914              * CALCULATE INTERACTIONS *
 915              **************************/
 916
 917             r12              = _mm256_mul_ps(rsq12,rinv12);
 918             r12              = _mm256_andnot_ps(dummy_mask,r12);
 919
 920             /* EWALD ELECTROSTATICS */
 921
 922             /* Analytical PME correction */
 923             zeta2            = _mm256_mul_ps(beta2,rsq12);
 924             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 925             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 926             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 927             felec            = _mm256_mul_ps(qq12,felec);
 928             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 929             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 930             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 931             velec            = _mm256_mul_ps(qq12,velec);
 932
 933             /* Update potential sum for this i atom from the interaction with this j atom. */
 934             velec            = _mm256_andnot_ps(dummy_mask,velec);
 935             velecsum         = _mm256_add_ps(velecsum,velec);
 936
 937             fscal            = felec;
 938
 939             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 940
 941             /* Calculate temporary vectorial force */
 942             tx               = _mm256_mul_ps(fscal,dx12);
 943             ty               = _mm256_mul_ps(fscal,dy12);
 944             tz               = _mm256_mul_ps(fscal,dz12);
 945
 946             /* Update vectorial force */
 947             fix1             = _mm256_add_ps(fix1,tx);
 948             fiy1             = _mm256_add_ps(fiy1,ty);
 949             fiz1             = _mm256_add_ps(fiz1,tz);
 950
 951             fjx2             = _mm256_add_ps(fjx2,tx);
 952             fjy2             = _mm256_add_ps(fjy2,ty);
 953             fjz2             = _mm256_add_ps(fjz2,tz);
 954
 955             /**************************
 956              * CALCULATE INTERACTIONS *
 957              **************************/
 958
 959             r13              = _mm256_mul_ps(rsq13,rinv13);
 960             r13              = _mm256_andnot_ps(dummy_mask,r13);
 961
 962             /* EWALD ELECTROSTATICS */
 963
 964             /* Analytical PME correction */
 965             zeta2            = _mm256_mul_ps(beta2,rsq13);
 966             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
 967             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 968             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 969             felec            = _mm256_mul_ps(qq13,felec);
 970             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 971             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 972             velec            = _mm256_sub_ps(rinv13,pmecorrV);
 973             velec            = _mm256_mul_ps(qq13,velec);
 974
 975             /* Update potential sum for this i atom from the interaction with this j atom. */
 976             velec            = _mm256_andnot_ps(dummy_mask,velec);
 977             velecsum         = _mm256_add_ps(velecsum,velec);
 978
 979             fscal            = felec;
 980
 981             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 982
 983             /* Calculate temporary vectorial force */
 984             tx               = _mm256_mul_ps(fscal,dx13);
 985             ty               = _mm256_mul_ps(fscal,dy13);
 986             tz               = _mm256_mul_ps(fscal,dz13);
 987
 988             /* Update vectorial force */
 989             fix1             = _mm256_add_ps(fix1,tx);
 990             fiy1             = _mm256_add_ps(fiy1,ty);
 991             fiz1             = _mm256_add_ps(fiz1,tz);
 992
 993             fjx3             = _mm256_add_ps(fjx3,tx);
 994             fjy3             = _mm256_add_ps(fjy3,ty);
 995             fjz3             = _mm256_add_ps(fjz3,tz);
 996
 997             /**************************
 998              * CALCULATE INTERACTIONS *
 999              **************************/
1000
1001             r21              = _mm256_mul_ps(rsq21,rinv21);
1002             r21              = _mm256_andnot_ps(dummy_mask,r21);
1003
1004             /* EWALD ELECTROSTATICS */
1005
1006             /* Analytical PME correction */
1007             zeta2            = _mm256_mul_ps(beta2,rsq21);
1008             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1009             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1010             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1011             felec            = _mm256_mul_ps(qq21,felec);
1012             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1013             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1014             velec            = _mm256_sub_ps(rinv21,pmecorrV);
1015             velec            = _mm256_mul_ps(qq21,velec);
1016
1017             /* Update potential sum for this i atom from the interaction with this j atom. */
1018             velec            = _mm256_andnot_ps(dummy_mask,velec);
1019             velecsum         = _mm256_add_ps(velecsum,velec);
1020
1021             fscal            = felec;
1022
1023             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1024
1025             /* Calculate temporary vectorial force */
1026             tx               = _mm256_mul_ps(fscal,dx21);
1027             ty               = _mm256_mul_ps(fscal,dy21);
1028             tz               = _mm256_mul_ps(fscal,dz21);
1029
1030             /* Update vectorial force */
1031             fix2             = _mm256_add_ps(fix2,tx);
1032             fiy2             = _mm256_add_ps(fiy2,ty);
1033             fiz2             = _mm256_add_ps(fiz2,tz);
1034
1035             fjx1             = _mm256_add_ps(fjx1,tx);
1036             fjy1             = _mm256_add_ps(fjy1,ty);
1037             fjz1             = _mm256_add_ps(fjz1,tz);
1038
1039             /**************************
1040              * CALCULATE INTERACTIONS *
1041              **************************/
1042
1043             r22              = _mm256_mul_ps(rsq22,rinv22);
1044             r22              = _mm256_andnot_ps(dummy_mask,r22);
1045
1046             /* EWALD ELECTROSTATICS */
1047
1048             /* Analytical PME correction */
1049             zeta2            = _mm256_mul_ps(beta2,rsq22);
1050             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1051             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1052             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1053             felec            = _mm256_mul_ps(qq22,felec);
1054             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1055             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1056             velec            = _mm256_sub_ps(rinv22,pmecorrV);
1057             velec            = _mm256_mul_ps(qq22,velec);
1058
1059             /* Update potential sum for this i atom from the interaction with this j atom. */
1060             velec            = _mm256_andnot_ps(dummy_mask,velec);
1061             velecsum         = _mm256_add_ps(velecsum,velec);
1062
1063             fscal            = felec;
1064
1065             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1066
1067             /* Calculate temporary vectorial force */
1068             tx               = _mm256_mul_ps(fscal,dx22);
1069             ty               = _mm256_mul_ps(fscal,dy22);
1070             tz               = _mm256_mul_ps(fscal,dz22);
1071
1072             /* Update vectorial force */
1073             fix2             = _mm256_add_ps(fix2,tx);
1074             fiy2             = _mm256_add_ps(fiy2,ty);
1075             fiz2             = _mm256_add_ps(fiz2,tz);
1076
1077             fjx2             = _mm256_add_ps(fjx2,tx);
1078             fjy2             = _mm256_add_ps(fjy2,ty);
1079             fjz2             = _mm256_add_ps(fjz2,tz);
1080
1081             /**************************
1082              * CALCULATE INTERACTIONS *
1083              **************************/
1084
1085             r23              = _mm256_mul_ps(rsq23,rinv23);
1086             r23              = _mm256_andnot_ps(dummy_mask,r23);
1087
1088             /* EWALD ELECTROSTATICS */
1089
1090             /* Analytical PME correction */
1091             zeta2            = _mm256_mul_ps(beta2,rsq23);
1092             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1093             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1094             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1095             felec            = _mm256_mul_ps(qq23,felec);
1096             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1097             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1098             velec            = _mm256_sub_ps(rinv23,pmecorrV);
1099             velec            = _mm256_mul_ps(qq23,velec);
1100
1101             /* Update potential sum for this i atom from the interaction with this j atom. */
1102             velec            = _mm256_andnot_ps(dummy_mask,velec);
1103             velecsum         = _mm256_add_ps(velecsum,velec);
1104
1105             fscal            = felec;
1106
1107             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1108
1109             /* Calculate temporary vectorial force */
1110             tx               = _mm256_mul_ps(fscal,dx23);
1111             ty               = _mm256_mul_ps(fscal,dy23);
1112             tz               = _mm256_mul_ps(fscal,dz23);
1113
1114             /* Update vectorial force */
1115             fix2             = _mm256_add_ps(fix2,tx);
1116             fiy2             = _mm256_add_ps(fiy2,ty);
1117             fiz2             = _mm256_add_ps(fiz2,tz);
1118
1119             fjx3             = _mm256_add_ps(fjx3,tx);
1120             fjy3             = _mm256_add_ps(fjy3,ty);
1121             fjz3             = _mm256_add_ps(fjz3,tz);
1122
1123             /**************************
1124              * CALCULATE INTERACTIONS *
1125              **************************/
1126
1127             r31              = _mm256_mul_ps(rsq31,rinv31);
1128             r31              = _mm256_andnot_ps(dummy_mask,r31);
1129
1130             /* EWALD ELECTROSTATICS */
1131
1132             /* Analytical PME correction */
1133             zeta2            = _mm256_mul_ps(beta2,rsq31);
1134             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1135             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1136             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1137             felec            = _mm256_mul_ps(qq31,felec);
1138             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1139             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1140             velec            = _mm256_sub_ps(rinv31,pmecorrV);
1141             velec            = _mm256_mul_ps(qq31,velec);
1142
1143             /* Update potential sum for this i atom from the interaction with this j atom. */
1144             velec            = _mm256_andnot_ps(dummy_mask,velec);
1145             velecsum         = _mm256_add_ps(velecsum,velec);
1146
1147             fscal            = felec;
1148
1149             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1150
1151             /* Calculate temporary vectorial force */
1152             tx               = _mm256_mul_ps(fscal,dx31);
1153             ty               = _mm256_mul_ps(fscal,dy31);
1154             tz               = _mm256_mul_ps(fscal,dz31);
1155
1156             /* Update vectorial force */
1157             fix3             = _mm256_add_ps(fix3,tx);
1158             fiy3             = _mm256_add_ps(fiy3,ty);
1159             fiz3             = _mm256_add_ps(fiz3,tz);
1160
1161             fjx1             = _mm256_add_ps(fjx1,tx);
1162             fjy1             = _mm256_add_ps(fjy1,ty);
1163             fjz1             = _mm256_add_ps(fjz1,tz);
1164
1165             /**************************
1166              * CALCULATE INTERACTIONS *
1167              **************************/
1168
1169             r32              = _mm256_mul_ps(rsq32,rinv32);
1170             r32              = _mm256_andnot_ps(dummy_mask,r32);
1171
1172             /* EWALD ELECTROSTATICS */
1173
1174             /* Analytical PME correction */
1175             zeta2            = _mm256_mul_ps(beta2,rsq32);
1176             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1177             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1178             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1179             felec            = _mm256_mul_ps(qq32,felec);
1180             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1181             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1182             velec            = _mm256_sub_ps(rinv32,pmecorrV);
1183             velec            = _mm256_mul_ps(qq32,velec);
1184
1185             /* Update potential sum for this i atom from the interaction with this j atom. */
1186             velec            = _mm256_andnot_ps(dummy_mask,velec);
1187             velecsum         = _mm256_add_ps(velecsum,velec);
1188
1189             fscal            = felec;
1190
1191             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1192
1193             /* Calculate temporary vectorial force */
1194             tx               = _mm256_mul_ps(fscal,dx32);
1195             ty               = _mm256_mul_ps(fscal,dy32);
1196             tz               = _mm256_mul_ps(fscal,dz32);
1197
1198             /* Update vectorial force */
1199             fix3             = _mm256_add_ps(fix3,tx);
1200             fiy3             = _mm256_add_ps(fiy3,ty);
1201             fiz3             = _mm256_add_ps(fiz3,tz);
1202
1203             fjx2             = _mm256_add_ps(fjx2,tx);
1204             fjy2             = _mm256_add_ps(fjy2,ty);
1205             fjz2             = _mm256_add_ps(fjz2,tz);
1206
1207             /**************************
1208              * CALCULATE INTERACTIONS *
1209              **************************/
1210
1211             r33              = _mm256_mul_ps(rsq33,rinv33);
1212             r33              = _mm256_andnot_ps(dummy_mask,r33);
1213
1214             /* EWALD ELECTROSTATICS */
1215
1216             /* Analytical PME correction */
1217             zeta2            = _mm256_mul_ps(beta2,rsq33);
1218             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1219             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1220             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1221             felec            = _mm256_mul_ps(qq33,felec);
1222             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1223             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1224             velec            = _mm256_sub_ps(rinv33,pmecorrV);
1225             velec            = _mm256_mul_ps(qq33,velec);
1226
1227             /* Update potential sum for this i atom from the interaction with this j atom. */
1228             velec            = _mm256_andnot_ps(dummy_mask,velec);
1229             velecsum         = _mm256_add_ps(velecsum,velec);
1230
1231             fscal            = felec;
1232
1233             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1234
1235             /* Calculate temporary vectorial force */
1236             tx               = _mm256_mul_ps(fscal,dx33);
1237             ty               = _mm256_mul_ps(fscal,dy33);
1238             tz               = _mm256_mul_ps(fscal,dz33);
1239
1240             /* Update vectorial force */
1241             fix3             = _mm256_add_ps(fix3,tx);
1242             fiy3             = _mm256_add_ps(fiy3,ty);
1243             fiz3             = _mm256_add_ps(fiz3,tz);
1244
1245             fjx3             = _mm256_add_ps(fjx3,tx);
1246             fjy3             = _mm256_add_ps(fjy3,ty);
1247             fjz3             = _mm256_add_ps(fjz3,tz);
1248
1249             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1250             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1251             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1252             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1253             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1254             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1255             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1256             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1257
1258             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1259                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1260                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1261
1262             /* Inner loop uses 800 flops */
1263         }
1264
1265         /* End of innermost loop */
1266
1267         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1268                                                  f+i_coord_offset,fshift+i_shift_offset);
1269
1270         ggid                        = gid[iidx];
1271         /* Update potential energies */
1272         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1273         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1274
1275         /* Increment number of inner iterations */
1276         inneriter                  += j_index_end - j_index_start;
1277
1278         /* Outer loop uses 26 flops */
1279     }
1280
1281     /* Increment number of outer iterations */
1282     outeriter        += nri;
1283
1284     /* Update outer/inner flops */
1285
1286     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*800);
1287 }
1288 /*
1289  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_single
1290  * Electrostatics interaction: Ewald
1291  * VdW interaction:            LennardJones
1292  * Geometry:                   Water4-Water4
1293  * Calculate force/pot:        Force
1294  */
1295 void
1296 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_single
1297                     (t_nblist * gmx_restrict                nlist,
1298                      rvec * gmx_restrict                    xx,
1299                      rvec * gmx_restrict                    ff,
1300                      t_forcerec * gmx_restrict              fr,
1301                      t_mdatoms * gmx_restrict               mdatoms,
1302                      nb_kernel_data_t * gmx_restrict        kernel_data,
1303                      t_nrnb * gmx_restrict                  nrnb)
1304 {
1305     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1306      * just 0 for non-waters.
1307      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1308      * jnr indices corresponding to data put in the four positions in the SIMD register.
1309      */
1310     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1311     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1312     int              jnrA,jnrB,jnrC,jnrD;
1313     int              jnrE,jnrF,jnrG,jnrH;
1314     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1315     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1316     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1317     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1318     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1319     real             rcutoff_scalar;
1320     real             *shiftvec,*fshift,*x,*f;
1321     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1322     real             scratch[4*DIM];
1323     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1324     real *           vdwioffsetptr0;
1325     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1326     real *           vdwioffsetptr1;
1327     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1328     real *           vdwioffsetptr2;
1329     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1330     real *           vdwioffsetptr3;
1331     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1332     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1333     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1334     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1335     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1336     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1337     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1338     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1339     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1340     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1341     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1342     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1343     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1344     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1345     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1346     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1347     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1348     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1349     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1350     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1351     real             *charge;
1352     int              nvdwtype;
1353     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1354     int              *vdwtype;
1355     real             *vdwparam;
1356     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1357     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1358     __m256i          ewitab;
1359     __m128i          ewitab_lo,ewitab_hi;
1360     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1361     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1362     real             *ewtab;
1363     __m256           dummy_mask,cutoff_mask;
1364     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1365     __m256           one     = _mm256_set1_ps(1.0);
1366     __m256           two     = _mm256_set1_ps(2.0);
1367     x                = xx[0];
1368     f                = ff[0];
1369
1370     nri              = nlist->nri;
1371     iinr             = nlist->iinr;
1372     jindex           = nlist->jindex;
1373     jjnr             = nlist->jjnr;
1374     shiftidx         = nlist->shift;
1375     gid              = nlist->gid;
1376     shiftvec         = fr->shift_vec[0];
1377     fshift           = fr->fshift[0];
1378     facel            = _mm256_set1_ps(fr->epsfac);
1379     charge           = mdatoms->chargeA;
1380     nvdwtype         = fr->ntype;
1381     vdwparam         = fr->nbfp;
1382     vdwtype          = mdatoms->typeA;
1383
1384     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
1385     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff);
1386     beta2            = _mm256_mul_ps(beta,beta);
1387     beta3            = _mm256_mul_ps(beta,beta2);
1388
1389     ewtab            = fr->ic->tabq_coul_F;
1390     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
1391     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1392
1393     /* Setup water-specific parameters */
1394     inr              = nlist->iinr[0];
1395     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1396     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1397     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1398     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1399
1400     jq1              = _mm256_set1_ps(charge[inr+1]);
1401     jq2              = _mm256_set1_ps(charge[inr+2]);
1402     jq3              = _mm256_set1_ps(charge[inr+3]);
1403     vdwjidx0A        = 2*vdwtype[inr+0];
1404     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1405     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1406     qq11             = _mm256_mul_ps(iq1,jq1);
1407     qq12             = _mm256_mul_ps(iq1,jq2);
1408     qq13             = _mm256_mul_ps(iq1,jq3);
1409     qq21             = _mm256_mul_ps(iq2,jq1);
1410     qq22             = _mm256_mul_ps(iq2,jq2);
1411     qq23             = _mm256_mul_ps(iq2,jq3);
1412     qq31             = _mm256_mul_ps(iq3,jq1);
1413     qq32             = _mm256_mul_ps(iq3,jq2);
1414     qq33             = _mm256_mul_ps(iq3,jq3);
1415
1416     /* Avoid stupid compiler warnings */
1417     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1418     j_coord_offsetA = 0;
1419     j_coord_offsetB = 0;
1420     j_coord_offsetC = 0;
1421     j_coord_offsetD = 0;
1422     j_coord_offsetE = 0;
1423     j_coord_offsetF = 0;
1424     j_coord_offsetG = 0;
1425     j_coord_offsetH = 0;
1426
1427     outeriter        = 0;
1428     inneriter        = 0;
1429
1430     for(iidx=0;iidx<4*DIM;iidx++)
1431     {
1432         scratch[iidx] = 0.0;
1433     }
1434
1435     /* Start outer loop over neighborlists */
1436     for(iidx=0; iidx<nri; iidx++)
1437     {
1438         /* Load shift vector for this list */
1439         i_shift_offset   = DIM*shiftidx[iidx];
1440
1441         /* Load limits for loop over neighbors */
1442         j_index_start    = jindex[iidx];
1443         j_index_end      = jindex[iidx+1];
1444
1445         /* Get outer coordinate index */
1446         inr              = iinr[iidx];
1447         i_coord_offset   = DIM*inr;
1448
1449         /* Load i particle coords and add shift vector */
1450         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1451                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1452
1453         fix0             = _mm256_setzero_ps();
1454         fiy0             = _mm256_setzero_ps();
1455         fiz0             = _mm256_setzero_ps();
1456         fix1             = _mm256_setzero_ps();
1457         fiy1             = _mm256_setzero_ps();
1458         fiz1             = _mm256_setzero_ps();
1459         fix2             = _mm256_setzero_ps();
1460         fiy2             = _mm256_setzero_ps();
1461         fiz2             = _mm256_setzero_ps();
1462         fix3             = _mm256_setzero_ps();
1463         fiy3             = _mm256_setzero_ps();
1464         fiz3             = _mm256_setzero_ps();
1465
1466         /* Start inner kernel loop */
1467         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1468         {
1469
1470             /* Get j neighbor index, and coordinate index */
1471             jnrA             = jjnr[jidx];
1472             jnrB             = jjnr[jidx+1];
1473             jnrC             = jjnr[jidx+2];
1474             jnrD             = jjnr[jidx+3];
1475             jnrE             = jjnr[jidx+4];
1476             jnrF             = jjnr[jidx+5];
1477             jnrG             = jjnr[jidx+6];
1478             jnrH             = jjnr[jidx+7];
1479             j_coord_offsetA  = DIM*jnrA;
1480             j_coord_offsetB  = DIM*jnrB;
1481             j_coord_offsetC  = DIM*jnrC;
1482             j_coord_offsetD  = DIM*jnrD;
1483             j_coord_offsetE  = DIM*jnrE;
1484             j_coord_offsetF  = DIM*jnrF;
1485             j_coord_offsetG  = DIM*jnrG;
1486             j_coord_offsetH  = DIM*jnrH;
1487
1488             /* load j atom coordinates */
1489             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1490                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1491                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1492                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1493                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1494                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1495
1496             /* Calculate displacement vector */
1497             dx00             = _mm256_sub_ps(ix0,jx0);
1498             dy00             = _mm256_sub_ps(iy0,jy0);
1499             dz00             = _mm256_sub_ps(iz0,jz0);
1500             dx11             = _mm256_sub_ps(ix1,jx1);
1501             dy11             = _mm256_sub_ps(iy1,jy1);
1502             dz11             = _mm256_sub_ps(iz1,jz1);
1503             dx12             = _mm256_sub_ps(ix1,jx2);
1504             dy12             = _mm256_sub_ps(iy1,jy2);
1505             dz12             = _mm256_sub_ps(iz1,jz2);
1506             dx13             = _mm256_sub_ps(ix1,jx3);
1507             dy13             = _mm256_sub_ps(iy1,jy3);
1508             dz13             = _mm256_sub_ps(iz1,jz3);
1509             dx21             = _mm256_sub_ps(ix2,jx1);
1510             dy21             = _mm256_sub_ps(iy2,jy1);
1511             dz21             = _mm256_sub_ps(iz2,jz1);
1512             dx22             = _mm256_sub_ps(ix2,jx2);
1513             dy22             = _mm256_sub_ps(iy2,jy2);
1514             dz22             = _mm256_sub_ps(iz2,jz2);
1515             dx23             = _mm256_sub_ps(ix2,jx3);
1516             dy23             = _mm256_sub_ps(iy2,jy3);
1517             dz23             = _mm256_sub_ps(iz2,jz3);
1518             dx31             = _mm256_sub_ps(ix3,jx1);
1519             dy31             = _mm256_sub_ps(iy3,jy1);
1520             dz31             = _mm256_sub_ps(iz3,jz1);
1521             dx32             = _mm256_sub_ps(ix3,jx2);
1522             dy32             = _mm256_sub_ps(iy3,jy2);
1523             dz32             = _mm256_sub_ps(iz3,jz2);
1524             dx33             = _mm256_sub_ps(ix3,jx3);
1525             dy33             = _mm256_sub_ps(iy3,jy3);
1526             dz33             = _mm256_sub_ps(iz3,jz3);
1527
1528             /* Calculate squared distance and things based on it */
1529             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1530             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1531             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1532             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1533             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1534             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1535             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1536             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1537             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1538             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1539
1540             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1541             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1542             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1543             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1544             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1545             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1546             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1547             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1548             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1549
1550             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1551             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1552             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1553             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1554             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1555             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1556             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1557             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1558             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1559             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1560
1561             fjx0             = _mm256_setzero_ps();
1562             fjy0             = _mm256_setzero_ps();
1563             fjz0             = _mm256_setzero_ps();
1564             fjx1             = _mm256_setzero_ps();
1565             fjy1             = _mm256_setzero_ps();
1566             fjz1             = _mm256_setzero_ps();
1567             fjx2             = _mm256_setzero_ps();
1568             fjy2             = _mm256_setzero_ps();
1569             fjz2             = _mm256_setzero_ps();
1570             fjx3             = _mm256_setzero_ps();
1571             fjy3             = _mm256_setzero_ps();
1572             fjz3             = _mm256_setzero_ps();
1573
1574             /**************************
1575              * CALCULATE INTERACTIONS *
1576              **************************/
1577
1578             /* LENNARD-JONES DISPERSION/REPULSION */
1579
1580             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1581             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1582
1583             fscal            = fvdw;
1584
1585             /* Calculate temporary vectorial force */
1586             tx               = _mm256_mul_ps(fscal,dx00);
1587             ty               = _mm256_mul_ps(fscal,dy00);
1588             tz               = _mm256_mul_ps(fscal,dz00);
1589
1590             /* Update vectorial force */
1591             fix0             = _mm256_add_ps(fix0,tx);
1592             fiy0             = _mm256_add_ps(fiy0,ty);
1593             fiz0             = _mm256_add_ps(fiz0,tz);
1594
1595             fjx0             = _mm256_add_ps(fjx0,tx);
1596             fjy0             = _mm256_add_ps(fjy0,ty);
1597             fjz0             = _mm256_add_ps(fjz0,tz);
1598
1599             /**************************
1600              * CALCULATE INTERACTIONS *
1601              **************************/
1602
1603             r11              = _mm256_mul_ps(rsq11,rinv11);
1604
1605             /* EWALD ELECTROSTATICS */
1606
1607             /* Analytical PME correction */
1608             zeta2            = _mm256_mul_ps(beta2,rsq11);
1609             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
1610             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1611             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1612             felec            = _mm256_mul_ps(qq11,felec);
1613
1614             fscal            = felec;
1615
1616             /* Calculate temporary vectorial force */
1617             tx               = _mm256_mul_ps(fscal,dx11);
1618             ty               = _mm256_mul_ps(fscal,dy11);
1619             tz               = _mm256_mul_ps(fscal,dz11);
1620
1621             /* Update vectorial force */
1622             fix1             = _mm256_add_ps(fix1,tx);
1623             fiy1             = _mm256_add_ps(fiy1,ty);
1624             fiz1             = _mm256_add_ps(fiz1,tz);
1625
1626             fjx1             = _mm256_add_ps(fjx1,tx);
1627             fjy1             = _mm256_add_ps(fjy1,ty);
1628             fjz1             = _mm256_add_ps(fjz1,tz);
1629
1630             /**************************
1631              * CALCULATE INTERACTIONS *
1632              **************************/
1633
1634             r12              = _mm256_mul_ps(rsq12,rinv12);
1635
1636             /* EWALD ELECTROSTATICS */
1637
1638             /* Analytical PME correction */
1639             zeta2            = _mm256_mul_ps(beta2,rsq12);
1640             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
1641             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1642             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1643             felec            = _mm256_mul_ps(qq12,felec);
1644
1645             fscal            = felec;
1646
1647             /* Calculate temporary vectorial force */
1648             tx               = _mm256_mul_ps(fscal,dx12);
1649             ty               = _mm256_mul_ps(fscal,dy12);
1650             tz               = _mm256_mul_ps(fscal,dz12);
1651
1652             /* Update vectorial force */
1653             fix1             = _mm256_add_ps(fix1,tx);
1654             fiy1             = _mm256_add_ps(fiy1,ty);
1655             fiz1             = _mm256_add_ps(fiz1,tz);
1656
1657             fjx2             = _mm256_add_ps(fjx2,tx);
1658             fjy2             = _mm256_add_ps(fjy2,ty);
1659             fjz2             = _mm256_add_ps(fjz2,tz);
1660
1661             /**************************
1662              * CALCULATE INTERACTIONS *
1663              **************************/
1664
1665             r13              = _mm256_mul_ps(rsq13,rinv13);
1666
1667             /* EWALD ELECTROSTATICS */
1668
1669             /* Analytical PME correction */
1670             zeta2            = _mm256_mul_ps(beta2,rsq13);
1671             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
1672             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1673             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1674             felec            = _mm256_mul_ps(qq13,felec);
1675
1676             fscal            = felec;
1677
1678             /* Calculate temporary vectorial force */
1679             tx               = _mm256_mul_ps(fscal,dx13);
1680             ty               = _mm256_mul_ps(fscal,dy13);
1681             tz               = _mm256_mul_ps(fscal,dz13);
1682
1683             /* Update vectorial force */
1684             fix1             = _mm256_add_ps(fix1,tx);
1685             fiy1             = _mm256_add_ps(fiy1,ty);
1686             fiz1             = _mm256_add_ps(fiz1,tz);
1687
1688             fjx3             = _mm256_add_ps(fjx3,tx);
1689             fjy3             = _mm256_add_ps(fjy3,ty);
1690             fjz3             = _mm256_add_ps(fjz3,tz);
1691
1692             /**************************
1693              * CALCULATE INTERACTIONS *
1694              **************************/
1695
1696             r21              = _mm256_mul_ps(rsq21,rinv21);
1697
1698             /* EWALD ELECTROSTATICS */
1699
1700             /* Analytical PME correction */
1701             zeta2            = _mm256_mul_ps(beta2,rsq21);
1702             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1703             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1704             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1705             felec            = _mm256_mul_ps(qq21,felec);
1706
1707             fscal            = felec;
1708
1709             /* Calculate temporary vectorial force */
1710             tx               = _mm256_mul_ps(fscal,dx21);
1711             ty               = _mm256_mul_ps(fscal,dy21);
1712             tz               = _mm256_mul_ps(fscal,dz21);
1713
1714             /* Update vectorial force */
1715             fix2             = _mm256_add_ps(fix2,tx);
1716             fiy2             = _mm256_add_ps(fiy2,ty);
1717             fiz2             = _mm256_add_ps(fiz2,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             r22              = _mm256_mul_ps(rsq22,rinv22);
1728
1729             /* EWALD ELECTROSTATICS */
1730
1731             /* Analytical PME correction */
1732             zeta2            = _mm256_mul_ps(beta2,rsq22);
1733             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1734             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1735             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1736             felec            = _mm256_mul_ps(qq22,felec);
1737
1738             fscal            = felec;
1739
1740             /* Calculate temporary vectorial force */
1741             tx               = _mm256_mul_ps(fscal,dx22);
1742             ty               = _mm256_mul_ps(fscal,dy22);
1743             tz               = _mm256_mul_ps(fscal,dz22);
1744
1745             /* Update vectorial force */
1746             fix2             = _mm256_add_ps(fix2,tx);
1747             fiy2             = _mm256_add_ps(fiy2,ty);
1748             fiz2             = _mm256_add_ps(fiz2,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             r23              = _mm256_mul_ps(rsq23,rinv23);
1759
1760             /* EWALD ELECTROSTATICS */
1761
1762             /* Analytical PME correction */
1763             zeta2            = _mm256_mul_ps(beta2,rsq23);
1764             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1765             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1766             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1767             felec            = _mm256_mul_ps(qq23,felec);
1768
1769             fscal            = felec;
1770
1771             /* Calculate temporary vectorial force */
1772             tx               = _mm256_mul_ps(fscal,dx23);
1773             ty               = _mm256_mul_ps(fscal,dy23);
1774             tz               = _mm256_mul_ps(fscal,dz23);
1775
1776             /* Update vectorial force */
1777             fix2             = _mm256_add_ps(fix2,tx);
1778             fiy2             = _mm256_add_ps(fiy2,ty);
1779             fiz2             = _mm256_add_ps(fiz2,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             r31              = _mm256_mul_ps(rsq31,rinv31);
1790
1791             /* EWALD ELECTROSTATICS */
1792
1793             /* Analytical PME correction */
1794             zeta2            = _mm256_mul_ps(beta2,rsq31);
1795             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1796             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1797             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1798             felec            = _mm256_mul_ps(qq31,felec);
1799
1800             fscal            = felec;
1801
1802             /* Calculate temporary vectorial force */
1803             tx               = _mm256_mul_ps(fscal,dx31);
1804             ty               = _mm256_mul_ps(fscal,dy31);
1805             tz               = _mm256_mul_ps(fscal,dz31);
1806
1807             /* Update vectorial force */
1808             fix3             = _mm256_add_ps(fix3,tx);
1809             fiy3             = _mm256_add_ps(fiy3,ty);
1810             fiz3             = _mm256_add_ps(fiz3,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             r32              = _mm256_mul_ps(rsq32,rinv32);
1821
1822             /* EWALD ELECTROSTATICS */
1823
1824             /* Analytical PME correction */
1825             zeta2            = _mm256_mul_ps(beta2,rsq32);
1826             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1827             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1828             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1829             felec            = _mm256_mul_ps(qq32,felec);
1830
1831             fscal            = felec;
1832
1833             /* Calculate temporary vectorial force */
1834             tx               = _mm256_mul_ps(fscal,dx32);
1835             ty               = _mm256_mul_ps(fscal,dy32);
1836             tz               = _mm256_mul_ps(fscal,dz32);
1837
1838             /* Update vectorial force */
1839             fix3             = _mm256_add_ps(fix3,tx);
1840             fiy3             = _mm256_add_ps(fiy3,ty);
1841             fiz3             = _mm256_add_ps(fiz3,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             r33              = _mm256_mul_ps(rsq33,rinv33);
1852
1853             /* EWALD ELECTROSTATICS */
1854
1855             /* Analytical PME correction */
1856             zeta2            = _mm256_mul_ps(beta2,rsq33);
1857             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1858             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1859             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1860             felec            = _mm256_mul_ps(qq33,felec);
1861
1862             fscal            = felec;
1863
1864             /* Calculate temporary vectorial force */
1865             tx               = _mm256_mul_ps(fscal,dx33);
1866             ty               = _mm256_mul_ps(fscal,dy33);
1867             tz               = _mm256_mul_ps(fscal,dz33);
1868
1869             /* Update vectorial force */
1870             fix3             = _mm256_add_ps(fix3,tx);
1871             fiy3             = _mm256_add_ps(fiy3,ty);
1872             fiz3             = _mm256_add_ps(fiz3,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             fjptrA             = f+j_coord_offsetA;
1879             fjptrB             = f+j_coord_offsetB;
1880             fjptrC             = f+j_coord_offsetC;
1881             fjptrD             = f+j_coord_offsetD;
1882             fjptrE             = f+j_coord_offsetE;
1883             fjptrF             = f+j_coord_offsetF;
1884             fjptrG             = f+j_coord_offsetG;
1885             fjptrH             = f+j_coord_offsetH;
1886
1887             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1888                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1889                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1890
1891             /* Inner loop uses 534 flops */
1892         }
1893
1894         if(jidx<j_index_end)
1895         {
1896
1897             /* Get j neighbor index, and coordinate index */
1898             jnrlistA         = jjnr[jidx];
1899             jnrlistB         = jjnr[jidx+1];
1900             jnrlistC         = jjnr[jidx+2];
1901             jnrlistD         = jjnr[jidx+3];
1902             jnrlistE         = jjnr[jidx+4];
1903             jnrlistF         = jjnr[jidx+5];
1904             jnrlistG         = jjnr[jidx+6];
1905             jnrlistH         = jjnr[jidx+7];
1906             /* Sign of each element will be negative for non-real atoms.
1907              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1908              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1909              */
1910             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1911                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1912
1913             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1914             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1915             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1916             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1917             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1918             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1919             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1920             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1921             j_coord_offsetA  = DIM*jnrA;
1922             j_coord_offsetB  = DIM*jnrB;
1923             j_coord_offsetC  = DIM*jnrC;
1924             j_coord_offsetD  = DIM*jnrD;
1925             j_coord_offsetE  = DIM*jnrE;
1926             j_coord_offsetF  = DIM*jnrF;
1927             j_coord_offsetG  = DIM*jnrG;
1928             j_coord_offsetH  = DIM*jnrH;
1929
1930             /* load j atom coordinates */
1931             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1932                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1933                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1934                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1935                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1936                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1937
1938             /* Calculate displacement vector */
1939             dx00             = _mm256_sub_ps(ix0,jx0);
1940             dy00             = _mm256_sub_ps(iy0,jy0);
1941             dz00             = _mm256_sub_ps(iz0,jz0);
1942             dx11             = _mm256_sub_ps(ix1,jx1);
1943             dy11             = _mm256_sub_ps(iy1,jy1);
1944             dz11             = _mm256_sub_ps(iz1,jz1);
1945             dx12             = _mm256_sub_ps(ix1,jx2);
1946             dy12             = _mm256_sub_ps(iy1,jy2);
1947             dz12             = _mm256_sub_ps(iz1,jz2);
1948             dx13             = _mm256_sub_ps(ix1,jx3);
1949             dy13             = _mm256_sub_ps(iy1,jy3);
1950             dz13             = _mm256_sub_ps(iz1,jz3);
1951             dx21             = _mm256_sub_ps(ix2,jx1);
1952             dy21             = _mm256_sub_ps(iy2,jy1);
1953             dz21             = _mm256_sub_ps(iz2,jz1);
1954             dx22             = _mm256_sub_ps(ix2,jx2);
1955             dy22             = _mm256_sub_ps(iy2,jy2);
1956             dz22             = _mm256_sub_ps(iz2,jz2);
1957             dx23             = _mm256_sub_ps(ix2,jx3);
1958             dy23             = _mm256_sub_ps(iy2,jy3);
1959             dz23             = _mm256_sub_ps(iz2,jz3);
1960             dx31             = _mm256_sub_ps(ix3,jx1);
1961             dy31             = _mm256_sub_ps(iy3,jy1);
1962             dz31             = _mm256_sub_ps(iz3,jz1);
1963             dx32             = _mm256_sub_ps(ix3,jx2);
1964             dy32             = _mm256_sub_ps(iy3,jy2);
1965             dz32             = _mm256_sub_ps(iz3,jz2);
1966             dx33             = _mm256_sub_ps(ix3,jx3);
1967             dy33             = _mm256_sub_ps(iy3,jy3);
1968             dz33             = _mm256_sub_ps(iz3,jz3);
1969
1970             /* Calculate squared distance and things based on it */
1971             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1972             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1973             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1974             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1975             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1976             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1977             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1978             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1979             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1980             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1981
1982             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1983             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1984             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1985             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1986             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1987             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1988             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1989             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1990             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1991
1992             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1993             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1994             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1995             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1996             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1997             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1998             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1999             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
2000             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
2001             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
2002
2003             fjx0             = _mm256_setzero_ps();
2004             fjy0             = _mm256_setzero_ps();
2005             fjz0             = _mm256_setzero_ps();
2006             fjx1             = _mm256_setzero_ps();
2007             fjy1             = _mm256_setzero_ps();
2008             fjz1             = _mm256_setzero_ps();
2009             fjx2             = _mm256_setzero_ps();
2010             fjy2             = _mm256_setzero_ps();
2011             fjz2             = _mm256_setzero_ps();
2012             fjx3             = _mm256_setzero_ps();
2013             fjy3             = _mm256_setzero_ps();
2014             fjz3             = _mm256_setzero_ps();
2015
2016             /**************************
2017              * CALCULATE INTERACTIONS *
2018              **************************/
2019
2020             /* LENNARD-JONES DISPERSION/REPULSION */
2021
2022             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2023             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
2024
2025             fscal            = fvdw;
2026
2027             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2028
2029             /* Calculate temporary vectorial force */
2030             tx               = _mm256_mul_ps(fscal,dx00);
2031             ty               = _mm256_mul_ps(fscal,dy00);
2032             tz               = _mm256_mul_ps(fscal,dz00);
2033
2034             /* Update vectorial force */
2035             fix0             = _mm256_add_ps(fix0,tx);
2036             fiy0             = _mm256_add_ps(fiy0,ty);
2037             fiz0             = _mm256_add_ps(fiz0,tz);
2038
2039             fjx0             = _mm256_add_ps(fjx0,tx);
2040             fjy0             = _mm256_add_ps(fjy0,ty);
2041             fjz0             = _mm256_add_ps(fjz0,tz);
2042
2043             /**************************
2044              * CALCULATE INTERACTIONS *
2045              **************************/
2046
2047             r11              = _mm256_mul_ps(rsq11,rinv11);
2048             r11              = _mm256_andnot_ps(dummy_mask,r11);
2049
2050             /* EWALD ELECTROSTATICS */
2051
2052             /* Analytical PME correction */
2053             zeta2            = _mm256_mul_ps(beta2,rsq11);
2054             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
2055             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2056             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2057             felec            = _mm256_mul_ps(qq11,felec);
2058
2059             fscal            = felec;
2060
2061             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2062
2063             /* Calculate temporary vectorial force */
2064             tx               = _mm256_mul_ps(fscal,dx11);
2065             ty               = _mm256_mul_ps(fscal,dy11);
2066             tz               = _mm256_mul_ps(fscal,dz11);
2067
2068             /* Update vectorial force */
2069             fix1             = _mm256_add_ps(fix1,tx);
2070             fiy1             = _mm256_add_ps(fiy1,ty);
2071             fiz1             = _mm256_add_ps(fiz1,tz);
2072
2073             fjx1             = _mm256_add_ps(fjx1,tx);
2074             fjy1             = _mm256_add_ps(fjy1,ty);
2075             fjz1             = _mm256_add_ps(fjz1,tz);
2076
2077             /**************************
2078              * CALCULATE INTERACTIONS *
2079              **************************/
2080
2081             r12              = _mm256_mul_ps(rsq12,rinv12);
2082             r12              = _mm256_andnot_ps(dummy_mask,r12);
2083
2084             /* EWALD ELECTROSTATICS */
2085
2086             /* Analytical PME correction */
2087             zeta2            = _mm256_mul_ps(beta2,rsq12);
2088             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
2089             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2090             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2091             felec            = _mm256_mul_ps(qq12,felec);
2092
2093             fscal            = felec;
2094
2095             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2096
2097             /* Calculate temporary vectorial force */
2098             tx               = _mm256_mul_ps(fscal,dx12);
2099             ty               = _mm256_mul_ps(fscal,dy12);
2100             tz               = _mm256_mul_ps(fscal,dz12);
2101
2102             /* Update vectorial force */
2103             fix1             = _mm256_add_ps(fix1,tx);
2104             fiy1             = _mm256_add_ps(fiy1,ty);
2105             fiz1             = _mm256_add_ps(fiz1,tz);
2106
2107             fjx2             = _mm256_add_ps(fjx2,tx);
2108             fjy2             = _mm256_add_ps(fjy2,ty);
2109             fjz2             = _mm256_add_ps(fjz2,tz);
2110
2111             /**************************
2112              * CALCULATE INTERACTIONS *
2113              **************************/
2114
2115             r13              = _mm256_mul_ps(rsq13,rinv13);
2116             r13              = _mm256_andnot_ps(dummy_mask,r13);
2117
2118             /* EWALD ELECTROSTATICS */
2119
2120             /* Analytical PME correction */
2121             zeta2            = _mm256_mul_ps(beta2,rsq13);
2122             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
2123             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2124             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2125             felec            = _mm256_mul_ps(qq13,felec);
2126
2127             fscal            = felec;
2128
2129             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2130
2131             /* Calculate temporary vectorial force */
2132             tx               = _mm256_mul_ps(fscal,dx13);
2133             ty               = _mm256_mul_ps(fscal,dy13);
2134             tz               = _mm256_mul_ps(fscal,dz13);
2135
2136             /* Update vectorial force */
2137             fix1             = _mm256_add_ps(fix1,tx);
2138             fiy1             = _mm256_add_ps(fiy1,ty);
2139             fiz1             = _mm256_add_ps(fiz1,tz);
2140
2141             fjx3             = _mm256_add_ps(fjx3,tx);
2142             fjy3             = _mm256_add_ps(fjy3,ty);
2143             fjz3             = _mm256_add_ps(fjz3,tz);
2144
2145             /**************************
2146              * CALCULATE INTERACTIONS *
2147              **************************/
2148
2149             r21              = _mm256_mul_ps(rsq21,rinv21);
2150             r21              = _mm256_andnot_ps(dummy_mask,r21);
2151
2152             /* EWALD ELECTROSTATICS */
2153
2154             /* Analytical PME correction */
2155             zeta2            = _mm256_mul_ps(beta2,rsq21);
2156             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
2157             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2158             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2159             felec            = _mm256_mul_ps(qq21,felec);
2160
2161             fscal            = felec;
2162
2163             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2164
2165             /* Calculate temporary vectorial force */
2166             tx               = _mm256_mul_ps(fscal,dx21);
2167             ty               = _mm256_mul_ps(fscal,dy21);
2168             tz               = _mm256_mul_ps(fscal,dz21);
2169
2170             /* Update vectorial force */
2171             fix2             = _mm256_add_ps(fix2,tx);
2172             fiy2             = _mm256_add_ps(fiy2,ty);
2173             fiz2             = _mm256_add_ps(fiz2,tz);
2174
2175             fjx1             = _mm256_add_ps(fjx1,tx);
2176             fjy1             = _mm256_add_ps(fjy1,ty);
2177             fjz1             = _mm256_add_ps(fjz1,tz);
2178
2179             /**************************
2180              * CALCULATE INTERACTIONS *
2181              **************************/
2182
2183             r22              = _mm256_mul_ps(rsq22,rinv22);
2184             r22              = _mm256_andnot_ps(dummy_mask,r22);
2185
2186             /* EWALD ELECTROSTATICS */
2187
2188             /* Analytical PME correction */
2189             zeta2            = _mm256_mul_ps(beta2,rsq22);
2190             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
2191             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2192             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2193             felec            = _mm256_mul_ps(qq22,felec);
2194
2195             fscal            = felec;
2196
2197             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2198
2199             /* Calculate temporary vectorial force */
2200             tx               = _mm256_mul_ps(fscal,dx22);
2201             ty               = _mm256_mul_ps(fscal,dy22);
2202             tz               = _mm256_mul_ps(fscal,dz22);
2203
2204             /* Update vectorial force */
2205             fix2             = _mm256_add_ps(fix2,tx);
2206             fiy2             = _mm256_add_ps(fiy2,ty);
2207             fiz2             = _mm256_add_ps(fiz2,tz);
2208
2209             fjx2             = _mm256_add_ps(fjx2,tx);
2210             fjy2             = _mm256_add_ps(fjy2,ty);
2211             fjz2             = _mm256_add_ps(fjz2,tz);
2212
2213             /**************************
2214              * CALCULATE INTERACTIONS *
2215              **************************/
2216
2217             r23              = _mm256_mul_ps(rsq23,rinv23);
2218             r23              = _mm256_andnot_ps(dummy_mask,r23);
2219
2220             /* EWALD ELECTROSTATICS */
2221
2222             /* Analytical PME correction */
2223             zeta2            = _mm256_mul_ps(beta2,rsq23);
2224             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
2225             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2226             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2227             felec            = _mm256_mul_ps(qq23,felec);
2228
2229             fscal            = felec;
2230
2231             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2232
2233             /* Calculate temporary vectorial force */
2234             tx               = _mm256_mul_ps(fscal,dx23);
2235             ty               = _mm256_mul_ps(fscal,dy23);
2236             tz               = _mm256_mul_ps(fscal,dz23);
2237
2238             /* Update vectorial force */
2239             fix2             = _mm256_add_ps(fix2,tx);
2240             fiy2             = _mm256_add_ps(fiy2,ty);
2241             fiz2             = _mm256_add_ps(fiz2,tz);
2242
2243             fjx3             = _mm256_add_ps(fjx3,tx);
2244             fjy3             = _mm256_add_ps(fjy3,ty);
2245             fjz3             = _mm256_add_ps(fjz3,tz);
2246
2247             /**************************
2248              * CALCULATE INTERACTIONS *
2249              **************************/
2250
2251             r31              = _mm256_mul_ps(rsq31,rinv31);
2252             r31              = _mm256_andnot_ps(dummy_mask,r31);
2253
2254             /* EWALD ELECTROSTATICS */
2255
2256             /* Analytical PME correction */
2257             zeta2            = _mm256_mul_ps(beta2,rsq31);
2258             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
2259             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2260             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2261             felec            = _mm256_mul_ps(qq31,felec);
2262
2263             fscal            = felec;
2264
2265             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2266
2267             /* Calculate temporary vectorial force */
2268             tx               = _mm256_mul_ps(fscal,dx31);
2269             ty               = _mm256_mul_ps(fscal,dy31);
2270             tz               = _mm256_mul_ps(fscal,dz31);
2271
2272             /* Update vectorial force */
2273             fix3             = _mm256_add_ps(fix3,tx);
2274             fiy3             = _mm256_add_ps(fiy3,ty);
2275             fiz3             = _mm256_add_ps(fiz3,tz);
2276
2277             fjx1             = _mm256_add_ps(fjx1,tx);
2278             fjy1             = _mm256_add_ps(fjy1,ty);
2279             fjz1             = _mm256_add_ps(fjz1,tz);
2280
2281             /**************************
2282              * CALCULATE INTERACTIONS *
2283              **************************/
2284
2285             r32              = _mm256_mul_ps(rsq32,rinv32);
2286             r32              = _mm256_andnot_ps(dummy_mask,r32);
2287
2288             /* EWALD ELECTROSTATICS */
2289
2290             /* Analytical PME correction */
2291             zeta2            = _mm256_mul_ps(beta2,rsq32);
2292             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
2293             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2294             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2295             felec            = _mm256_mul_ps(qq32,felec);
2296
2297             fscal            = felec;
2298
2299             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2300
2301             /* Calculate temporary vectorial force */
2302             tx               = _mm256_mul_ps(fscal,dx32);
2303             ty               = _mm256_mul_ps(fscal,dy32);
2304             tz               = _mm256_mul_ps(fscal,dz32);
2305
2306             /* Update vectorial force */
2307             fix3             = _mm256_add_ps(fix3,tx);
2308             fiy3             = _mm256_add_ps(fiy3,ty);
2309             fiz3             = _mm256_add_ps(fiz3,tz);
2310
2311             fjx2             = _mm256_add_ps(fjx2,tx);
2312             fjy2             = _mm256_add_ps(fjy2,ty);
2313             fjz2             = _mm256_add_ps(fjz2,tz);
2314
2315             /**************************
2316              * CALCULATE INTERACTIONS *
2317              **************************/
2318
2319             r33              = _mm256_mul_ps(rsq33,rinv33);
2320             r33              = _mm256_andnot_ps(dummy_mask,r33);
2321
2322             /* EWALD ELECTROSTATICS */
2323
2324             /* Analytical PME correction */
2325             zeta2            = _mm256_mul_ps(beta2,rsq33);
2326             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
2327             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2328             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2329             felec            = _mm256_mul_ps(qq33,felec);
2330
2331             fscal            = felec;
2332
2333             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2334
2335             /* Calculate temporary vectorial force */
2336             tx               = _mm256_mul_ps(fscal,dx33);
2337             ty               = _mm256_mul_ps(fscal,dy33);
2338             tz               = _mm256_mul_ps(fscal,dz33);
2339
2340             /* Update vectorial force */
2341             fix3             = _mm256_add_ps(fix3,tx);
2342             fiy3             = _mm256_add_ps(fiy3,ty);
2343             fiz3             = _mm256_add_ps(fiz3,tz);
2344
2345             fjx3             = _mm256_add_ps(fjx3,tx);
2346             fjy3             = _mm256_add_ps(fjy3,ty);
2347             fjz3             = _mm256_add_ps(fjz3,tz);
2348
2349             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2350             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2351             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2352             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2353             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2354             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2355             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2356             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2357
2358             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2359                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2360                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2361
2362             /* Inner loop uses 543 flops */
2363         }
2364
2365         /* End of innermost loop */
2366
2367         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2368                                                  f+i_coord_offset,fshift+i_shift_offset);
2369
2370         /* Increment number of inner iterations */
2371         inneriter                  += j_index_end - j_index_start;
2372
2373         /* Outer loop uses 24 flops */
2374     }
2375
2376     /* Increment number of outer iterations */
2377     outeriter        += nri;
2378
2379     /* Update outer/inner flops */
2380
2381     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*543);
2382 }