src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEw_VdwLJ_GeomW4W4_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_single.h"
  48 #include "kernelutil_x86_avx_256_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_avx_256_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrE,jnrF,jnrG,jnrH;
  76     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  77     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  84     real             scratch[4*DIM];
  85     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 101     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 119     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 120     __m256i          ewitab;
 121     __m128i          ewitab_lo,ewitab_hi;
 122     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 123     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 124     real             *ewtab;
 125     __m256           dummy_mask,cutoff_mask;
 126     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 127     __m256           one     = _mm256_set1_ps(1.0);
 128     __m256           two     = _mm256_set1_ps(2.0);
 129     x                = xx[0];
 130     f                = ff[0];
 131
 132     nri              = nlist->nri;
 133     iinr             = nlist->iinr;
 134     jindex           = nlist->jindex;
 135     jjnr             = nlist->jjnr;
 136     shiftidx         = nlist->shift;
 137     gid              = nlist->gid;
 138     shiftvec         = fr->shift_vec[0];
 139     fshift           = fr->fshift[0];
 140     facel            = _mm256_set1_ps(fr->epsfac);
 141     charge           = mdatoms->chargeA;
 142     nvdwtype         = fr->ntype;
 143     vdwparam         = fr->nbfp;
 144     vdwtype          = mdatoms->typeA;
 145
 146     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 147     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 148     beta2            = _mm256_mul_ps(beta,beta);
 149     beta3            = _mm256_mul_ps(beta,beta2);
 150
 151     ewtab            = fr->ic->tabq_coul_FDV0;
 152     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 153     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 154
 155     /* Setup water-specific parameters */
 156     inr              = nlist->iinr[0];
 157     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 158     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 159     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 160     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 161
 162     jq1              = _mm256_set1_ps(charge[inr+1]);
 163     jq2              = _mm256_set1_ps(charge[inr+2]);
 164     jq3              = _mm256_set1_ps(charge[inr+3]);
 165     vdwjidx0A        = 2*vdwtype[inr+0];
 166     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 167     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 168     qq11             = _mm256_mul_ps(iq1,jq1);
 169     qq12             = _mm256_mul_ps(iq1,jq2);
 170     qq13             = _mm256_mul_ps(iq1,jq3);
 171     qq21             = _mm256_mul_ps(iq2,jq1);
 172     qq22             = _mm256_mul_ps(iq2,jq2);
 173     qq23             = _mm256_mul_ps(iq2,jq3);
 174     qq31             = _mm256_mul_ps(iq3,jq1);
 175     qq32             = _mm256_mul_ps(iq3,jq2);
 176     qq33             = _mm256_mul_ps(iq3,jq3);
 177
 178     /* Avoid stupid compiler warnings */
 179     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 180     j_coord_offsetA = 0;
 181     j_coord_offsetB = 0;
 182     j_coord_offsetC = 0;
 183     j_coord_offsetD = 0;
 184     j_coord_offsetE = 0;
 185     j_coord_offsetF = 0;
 186     j_coord_offsetG = 0;
 187     j_coord_offsetH = 0;
 188
 189     outeriter        = 0;
 190     inneriter        = 0;
 191
 192     for(iidx=0;iidx<4*DIM;iidx++)
 193     {
 194         scratch[iidx] = 0.0;
 195     }
 196
 197     /* Start outer loop over neighborlists */
 198     for(iidx=0; iidx<nri; iidx++)
 199     {
 200         /* Load shift vector for this list */
 201         i_shift_offset   = DIM*shiftidx[iidx];
 202
 203         /* Load limits for loop over neighbors */
 204         j_index_start    = jindex[iidx];
 205         j_index_end      = jindex[iidx+1];
 206
 207         /* Get outer coordinate index */
 208         inr              = iinr[iidx];
 209         i_coord_offset   = DIM*inr;
 210
 211         /* Load i particle coords and add shift vector */
 212         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 213                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 214
 215         fix0             = _mm256_setzero_ps();
 216         fiy0             = _mm256_setzero_ps();
 217         fiz0             = _mm256_setzero_ps();
 218         fix1             = _mm256_setzero_ps();
 219         fiy1             = _mm256_setzero_ps();
 220         fiz1             = _mm256_setzero_ps();
 221         fix2             = _mm256_setzero_ps();
 222         fiy2             = _mm256_setzero_ps();
 223         fiz2             = _mm256_setzero_ps();
 224         fix3             = _mm256_setzero_ps();
 225         fiy3             = _mm256_setzero_ps();
 226         fiz3             = _mm256_setzero_ps();
 227
 228         /* Reset potential sums */
 229         velecsum         = _mm256_setzero_ps();
 230         vvdwsum          = _mm256_setzero_ps();
 231
 232         /* Start inner kernel loop */
 233         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 234         {
 235
 236             /* Get j neighbor index, and coordinate index */
 237             jnrA             = jjnr[jidx];
 238             jnrB             = jjnr[jidx+1];
 239             jnrC             = jjnr[jidx+2];
 240             jnrD             = jjnr[jidx+3];
 241             jnrE             = jjnr[jidx+4];
 242             jnrF             = jjnr[jidx+5];
 243             jnrG             = jjnr[jidx+6];
 244             jnrH             = jjnr[jidx+7];
 245             j_coord_offsetA  = DIM*jnrA;
 246             j_coord_offsetB  = DIM*jnrB;
 247             j_coord_offsetC  = DIM*jnrC;
 248             j_coord_offsetD  = DIM*jnrD;
 249             j_coord_offsetE  = DIM*jnrE;
 250             j_coord_offsetF  = DIM*jnrF;
 251             j_coord_offsetG  = DIM*jnrG;
 252             j_coord_offsetH  = DIM*jnrH;
 253
 254             /* load j atom coordinates */
 255             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 256                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 257                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 258                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 259                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 260                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 261
 262             /* Calculate displacement vector */
 263             dx00             = _mm256_sub_ps(ix0,jx0);
 264             dy00             = _mm256_sub_ps(iy0,jy0);
 265             dz00             = _mm256_sub_ps(iz0,jz0);
 266             dx11             = _mm256_sub_ps(ix1,jx1);
 267             dy11             = _mm256_sub_ps(iy1,jy1);
 268             dz11             = _mm256_sub_ps(iz1,jz1);
 269             dx12             = _mm256_sub_ps(ix1,jx2);
 270             dy12             = _mm256_sub_ps(iy1,jy2);
 271             dz12             = _mm256_sub_ps(iz1,jz2);
 272             dx13             = _mm256_sub_ps(ix1,jx3);
 273             dy13             = _mm256_sub_ps(iy1,jy3);
 274             dz13             = _mm256_sub_ps(iz1,jz3);
 275             dx21             = _mm256_sub_ps(ix2,jx1);
 276             dy21             = _mm256_sub_ps(iy2,jy1);
 277             dz21             = _mm256_sub_ps(iz2,jz1);
 278             dx22             = _mm256_sub_ps(ix2,jx2);
 279             dy22             = _mm256_sub_ps(iy2,jy2);
 280             dz22             = _mm256_sub_ps(iz2,jz2);
 281             dx23             = _mm256_sub_ps(ix2,jx3);
 282             dy23             = _mm256_sub_ps(iy2,jy3);
 283             dz23             = _mm256_sub_ps(iz2,jz3);
 284             dx31             = _mm256_sub_ps(ix3,jx1);
 285             dy31             = _mm256_sub_ps(iy3,jy1);
 286             dz31             = _mm256_sub_ps(iz3,jz1);
 287             dx32             = _mm256_sub_ps(ix3,jx2);
 288             dy32             = _mm256_sub_ps(iy3,jy2);
 289             dz32             = _mm256_sub_ps(iz3,jz2);
 290             dx33             = _mm256_sub_ps(ix3,jx3);
 291             dy33             = _mm256_sub_ps(iy3,jy3);
 292             dz33             = _mm256_sub_ps(iz3,jz3);
 293
 294             /* Calculate squared distance and things based on it */
 295             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 296             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 297             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 298             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 299             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 300             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 301             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 302             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 303             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 304             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 305
 306             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 307             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 308             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 309             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 310             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 311             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 312             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 313             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 314             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 315
 316             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 317             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 318             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 319             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 320             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 321             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 322             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 323             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 324             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 325             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 326
 327             fjx0             = _mm256_setzero_ps();
 328             fjy0             = _mm256_setzero_ps();
 329             fjz0             = _mm256_setzero_ps();
 330             fjx1             = _mm256_setzero_ps();
 331             fjy1             = _mm256_setzero_ps();
 332             fjz1             = _mm256_setzero_ps();
 333             fjx2             = _mm256_setzero_ps();
 334             fjy2             = _mm256_setzero_ps();
 335             fjz2             = _mm256_setzero_ps();
 336             fjx3             = _mm256_setzero_ps();
 337             fjy3             = _mm256_setzero_ps();
 338             fjz3             = _mm256_setzero_ps();
 339
 340             /**************************
 341              * CALCULATE INTERACTIONS *
 342              **************************/
 343
 344             /* LENNARD-JONES DISPERSION/REPULSION */
 345
 346             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 347             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 348             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 349             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 350             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 354
 355             fscal            = fvdw;
 356
 357             /* Calculate temporary vectorial force */
 358             tx               = _mm256_mul_ps(fscal,dx00);
 359             ty               = _mm256_mul_ps(fscal,dy00);
 360             tz               = _mm256_mul_ps(fscal,dz00);
 361
 362             /* Update vectorial force */
 363             fix0             = _mm256_add_ps(fix0,tx);
 364             fiy0             = _mm256_add_ps(fiy0,ty);
 365             fiz0             = _mm256_add_ps(fiz0,tz);
 366
 367             fjx0             = _mm256_add_ps(fjx0,tx);
 368             fjy0             = _mm256_add_ps(fjy0,ty);
 369             fjz0             = _mm256_add_ps(fjz0,tz);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             r11              = _mm256_mul_ps(rsq11,rinv11);
 376
 377             /* EWALD ELECTROSTATICS */
 378
 379             /* Analytical PME correction */
 380             zeta2            = _mm256_mul_ps(beta2,rsq11);
 381             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 382             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 383             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 384             felec            = _mm256_mul_ps(qq11,felec);
 385             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 386             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 387             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 388             velec            = _mm256_mul_ps(qq11,velec);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velecsum         = _mm256_add_ps(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395             /* Calculate temporary vectorial force */
 396             tx               = _mm256_mul_ps(fscal,dx11);
 397             ty               = _mm256_mul_ps(fscal,dy11);
 398             tz               = _mm256_mul_ps(fscal,dz11);
 399
 400             /* Update vectorial force */
 401             fix1             = _mm256_add_ps(fix1,tx);
 402             fiy1             = _mm256_add_ps(fiy1,ty);
 403             fiz1             = _mm256_add_ps(fiz1,tz);
 404
 405             fjx1             = _mm256_add_ps(fjx1,tx);
 406             fjy1             = _mm256_add_ps(fjy1,ty);
 407             fjz1             = _mm256_add_ps(fjz1,tz);
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             r12              = _mm256_mul_ps(rsq12,rinv12);
 414
 415             /* EWALD ELECTROSTATICS */
 416
 417             /* Analytical PME correction */
 418             zeta2            = _mm256_mul_ps(beta2,rsq12);
 419             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 420             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 421             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 422             felec            = _mm256_mul_ps(qq12,felec);
 423             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 424             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 425             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 426             velec            = _mm256_mul_ps(qq12,velec);
 427
 428             /* Update potential sum for this i atom from the interaction with this j atom. */
 429             velecsum         = _mm256_add_ps(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             /* Calculate temporary vectorial force */
 434             tx               = _mm256_mul_ps(fscal,dx12);
 435             ty               = _mm256_mul_ps(fscal,dy12);
 436             tz               = _mm256_mul_ps(fscal,dz12);
 437
 438             /* Update vectorial force */
 439             fix1             = _mm256_add_ps(fix1,tx);
 440             fiy1             = _mm256_add_ps(fiy1,ty);
 441             fiz1             = _mm256_add_ps(fiz1,tz);
 442
 443             fjx2             = _mm256_add_ps(fjx2,tx);
 444             fjy2             = _mm256_add_ps(fjy2,ty);
 445             fjz2             = _mm256_add_ps(fjz2,tz);
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             r13              = _mm256_mul_ps(rsq13,rinv13);
 452
 453             /* EWALD ELECTROSTATICS */
 454
 455             /* Analytical PME correction */
 456             zeta2            = _mm256_mul_ps(beta2,rsq13);
 457             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
 458             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 459             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 460             felec            = _mm256_mul_ps(qq13,felec);
 461             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 462             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 463             velec            = _mm256_sub_ps(rinv13,pmecorrV);
 464             velec            = _mm256_mul_ps(qq13,velec);
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velecsum         = _mm256_add_ps(velecsum,velec);
 468
 469             fscal            = felec;
 470
 471             /* Calculate temporary vectorial force */
 472             tx               = _mm256_mul_ps(fscal,dx13);
 473             ty               = _mm256_mul_ps(fscal,dy13);
 474             tz               = _mm256_mul_ps(fscal,dz13);
 475
 476             /* Update vectorial force */
 477             fix1             = _mm256_add_ps(fix1,tx);
 478             fiy1             = _mm256_add_ps(fiy1,ty);
 479             fiz1             = _mm256_add_ps(fiz1,tz);
 480
 481             fjx3             = _mm256_add_ps(fjx3,tx);
 482             fjy3             = _mm256_add_ps(fjy3,ty);
 483             fjz3             = _mm256_add_ps(fjz3,tz);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             r21              = _mm256_mul_ps(rsq21,rinv21);
 490
 491             /* EWALD ELECTROSTATICS */
 492
 493             /* Analytical PME correction */
 494             zeta2            = _mm256_mul_ps(beta2,rsq21);
 495             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
 496             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 497             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 498             felec            = _mm256_mul_ps(qq21,felec);
 499             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 500             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 501             velec            = _mm256_sub_ps(rinv21,pmecorrV);
 502             velec            = _mm256_mul_ps(qq21,velec);
 503
 504             /* Update potential sum for this i atom from the interaction with this j atom. */
 505             velecsum         = _mm256_add_ps(velecsum,velec);
 506
 507             fscal            = felec;
 508
 509             /* Calculate temporary vectorial force */
 510             tx               = _mm256_mul_ps(fscal,dx21);
 511             ty               = _mm256_mul_ps(fscal,dy21);
 512             tz               = _mm256_mul_ps(fscal,dz21);
 513
 514             /* Update vectorial force */
 515             fix2             = _mm256_add_ps(fix2,tx);
 516             fiy2             = _mm256_add_ps(fiy2,ty);
 517             fiz2             = _mm256_add_ps(fiz2,tz);
 518
 519             fjx1             = _mm256_add_ps(fjx1,tx);
 520             fjy1             = _mm256_add_ps(fjy1,ty);
 521             fjz1             = _mm256_add_ps(fjz1,tz);
 522
 523             /**************************
 524              * CALCULATE INTERACTIONS *
 525              **************************/
 526
 527             r22              = _mm256_mul_ps(rsq22,rinv22);
 528
 529             /* EWALD ELECTROSTATICS */
 530
 531             /* Analytical PME correction */
 532             zeta2            = _mm256_mul_ps(beta2,rsq22);
 533             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
 534             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 535             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 536             felec            = _mm256_mul_ps(qq22,felec);
 537             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 538             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 539             velec            = _mm256_sub_ps(rinv22,pmecorrV);
 540             velec            = _mm256_mul_ps(qq22,velec);
 541
 542             /* Update potential sum for this i atom from the interaction with this j atom. */
 543             velecsum         = _mm256_add_ps(velecsum,velec);
 544
 545             fscal            = felec;
 546
 547             /* Calculate temporary vectorial force */
 548             tx               = _mm256_mul_ps(fscal,dx22);
 549             ty               = _mm256_mul_ps(fscal,dy22);
 550             tz               = _mm256_mul_ps(fscal,dz22);
 551
 552             /* Update vectorial force */
 553             fix2             = _mm256_add_ps(fix2,tx);
 554             fiy2             = _mm256_add_ps(fiy2,ty);
 555             fiz2             = _mm256_add_ps(fiz2,tz);
 556
 557             fjx2             = _mm256_add_ps(fjx2,tx);
 558             fjy2             = _mm256_add_ps(fjy2,ty);
 559             fjz2             = _mm256_add_ps(fjz2,tz);
 560
 561             /**************************
 562              * CALCULATE INTERACTIONS *
 563              **************************/
 564
 565             r23              = _mm256_mul_ps(rsq23,rinv23);
 566
 567             /* EWALD ELECTROSTATICS */
 568
 569             /* Analytical PME correction */
 570             zeta2            = _mm256_mul_ps(beta2,rsq23);
 571             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
 572             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 573             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 574             felec            = _mm256_mul_ps(qq23,felec);
 575             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 576             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 577             velec            = _mm256_sub_ps(rinv23,pmecorrV);
 578             velec            = _mm256_mul_ps(qq23,velec);
 579
 580             /* Update potential sum for this i atom from the interaction with this j atom. */
 581             velecsum         = _mm256_add_ps(velecsum,velec);
 582
 583             fscal            = felec;
 584
 585             /* Calculate temporary vectorial force */
 586             tx               = _mm256_mul_ps(fscal,dx23);
 587             ty               = _mm256_mul_ps(fscal,dy23);
 588             tz               = _mm256_mul_ps(fscal,dz23);
 589
 590             /* Update vectorial force */
 591             fix2             = _mm256_add_ps(fix2,tx);
 592             fiy2             = _mm256_add_ps(fiy2,ty);
 593             fiz2             = _mm256_add_ps(fiz2,tz);
 594
 595             fjx3             = _mm256_add_ps(fjx3,tx);
 596             fjy3             = _mm256_add_ps(fjy3,ty);
 597             fjz3             = _mm256_add_ps(fjz3,tz);
 598
 599             /**************************
 600              * CALCULATE INTERACTIONS *
 601              **************************/
 602
 603             r31              = _mm256_mul_ps(rsq31,rinv31);
 604
 605             /* EWALD ELECTROSTATICS */
 606
 607             /* Analytical PME correction */
 608             zeta2            = _mm256_mul_ps(beta2,rsq31);
 609             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
 610             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 611             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 612             felec            = _mm256_mul_ps(qq31,felec);
 613             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 614             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 615             velec            = _mm256_sub_ps(rinv31,pmecorrV);
 616             velec            = _mm256_mul_ps(qq31,velec);
 617
 618             /* Update potential sum for this i atom from the interaction with this j atom. */
 619             velecsum         = _mm256_add_ps(velecsum,velec);
 620
 621             fscal            = felec;
 622
 623             /* Calculate temporary vectorial force */
 624             tx               = _mm256_mul_ps(fscal,dx31);
 625             ty               = _mm256_mul_ps(fscal,dy31);
 626             tz               = _mm256_mul_ps(fscal,dz31);
 627
 628             /* Update vectorial force */
 629             fix3             = _mm256_add_ps(fix3,tx);
 630             fiy3             = _mm256_add_ps(fiy3,ty);
 631             fiz3             = _mm256_add_ps(fiz3,tz);
 632
 633             fjx1             = _mm256_add_ps(fjx1,tx);
 634             fjy1             = _mm256_add_ps(fjy1,ty);
 635             fjz1             = _mm256_add_ps(fjz1,tz);
 636
 637             /**************************
 638              * CALCULATE INTERACTIONS *
 639              **************************/
 640
 641             r32              = _mm256_mul_ps(rsq32,rinv32);
 642
 643             /* EWALD ELECTROSTATICS */
 644
 645             /* Analytical PME correction */
 646             zeta2            = _mm256_mul_ps(beta2,rsq32);
 647             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
 648             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 649             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 650             felec            = _mm256_mul_ps(qq32,felec);
 651             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 652             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 653             velec            = _mm256_sub_ps(rinv32,pmecorrV);
 654             velec            = _mm256_mul_ps(qq32,velec);
 655
 656             /* Update potential sum for this i atom from the interaction with this j atom. */
 657             velecsum         = _mm256_add_ps(velecsum,velec);
 658
 659             fscal            = felec;
 660
 661             /* Calculate temporary vectorial force */
 662             tx               = _mm256_mul_ps(fscal,dx32);
 663             ty               = _mm256_mul_ps(fscal,dy32);
 664             tz               = _mm256_mul_ps(fscal,dz32);
 665
 666             /* Update vectorial force */
 667             fix3             = _mm256_add_ps(fix3,tx);
 668             fiy3             = _mm256_add_ps(fiy3,ty);
 669             fiz3             = _mm256_add_ps(fiz3,tz);
 670
 671             fjx2             = _mm256_add_ps(fjx2,tx);
 672             fjy2             = _mm256_add_ps(fjy2,ty);
 673             fjz2             = _mm256_add_ps(fjz2,tz);
 674
 675             /**************************
 676              * CALCULATE INTERACTIONS *
 677              **************************/
 678
 679             r33              = _mm256_mul_ps(rsq33,rinv33);
 680
 681             /* EWALD ELECTROSTATICS */
 682
 683             /* Analytical PME correction */
 684             zeta2            = _mm256_mul_ps(beta2,rsq33);
 685             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
 686             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 687             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 688             felec            = _mm256_mul_ps(qq33,felec);
 689             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 690             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 691             velec            = _mm256_sub_ps(rinv33,pmecorrV);
 692             velec            = _mm256_mul_ps(qq33,velec);
 693
 694             /* Update potential sum for this i atom from the interaction with this j atom. */
 695             velecsum         = _mm256_add_ps(velecsum,velec);
 696
 697             fscal            = felec;
 698
 699             /* Calculate temporary vectorial force */
 700             tx               = _mm256_mul_ps(fscal,dx33);
 701             ty               = _mm256_mul_ps(fscal,dy33);
 702             tz               = _mm256_mul_ps(fscal,dz33);
 703
 704             /* Update vectorial force */
 705             fix3             = _mm256_add_ps(fix3,tx);
 706             fiy3             = _mm256_add_ps(fiy3,ty);
 707             fiz3             = _mm256_add_ps(fiz3,tz);
 708
 709             fjx3             = _mm256_add_ps(fjx3,tx);
 710             fjy3             = _mm256_add_ps(fjy3,ty);
 711             fjz3             = _mm256_add_ps(fjz3,tz);
 712
 713             fjptrA             = f+j_coord_offsetA;
 714             fjptrB             = f+j_coord_offsetB;
 715             fjptrC             = f+j_coord_offsetC;
 716             fjptrD             = f+j_coord_offsetD;
 717             fjptrE             = f+j_coord_offsetE;
 718             fjptrF             = f+j_coord_offsetF;
 719             fjptrG             = f+j_coord_offsetG;
 720             fjptrH             = f+j_coord_offsetH;
 721
 722             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 723                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 724                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 725
 726             /* Inner loop uses 791 flops */
 727         }
 728
 729         if(jidx<j_index_end)
 730         {
 731
 732             /* Get j neighbor index, and coordinate index */
 733             jnrlistA         = jjnr[jidx];
 734             jnrlistB         = jjnr[jidx+1];
 735             jnrlistC         = jjnr[jidx+2];
 736             jnrlistD         = jjnr[jidx+3];
 737             jnrlistE         = jjnr[jidx+4];
 738             jnrlistF         = jjnr[jidx+5];
 739             jnrlistG         = jjnr[jidx+6];
 740             jnrlistH         = jjnr[jidx+7];
 741             /* Sign of each element will be negative for non-real atoms.
 742              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 743              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 744              */
 745             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 746                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 747
 748             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 749             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 750             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 751             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 752             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 753             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 754             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 755             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 756             j_coord_offsetA  = DIM*jnrA;
 757             j_coord_offsetB  = DIM*jnrB;
 758             j_coord_offsetC  = DIM*jnrC;
 759             j_coord_offsetD  = DIM*jnrD;
 760             j_coord_offsetE  = DIM*jnrE;
 761             j_coord_offsetF  = DIM*jnrF;
 762             j_coord_offsetG  = DIM*jnrG;
 763             j_coord_offsetH  = DIM*jnrH;
 764
 765             /* load j atom coordinates */
 766             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 767                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 768                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 769                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 770                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 771                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 772
 773             /* Calculate displacement vector */
 774             dx00             = _mm256_sub_ps(ix0,jx0);
 775             dy00             = _mm256_sub_ps(iy0,jy0);
 776             dz00             = _mm256_sub_ps(iz0,jz0);
 777             dx11             = _mm256_sub_ps(ix1,jx1);
 778             dy11             = _mm256_sub_ps(iy1,jy1);
 779             dz11             = _mm256_sub_ps(iz1,jz1);
 780             dx12             = _mm256_sub_ps(ix1,jx2);
 781             dy12             = _mm256_sub_ps(iy1,jy2);
 782             dz12             = _mm256_sub_ps(iz1,jz2);
 783             dx13             = _mm256_sub_ps(ix1,jx3);
 784             dy13             = _mm256_sub_ps(iy1,jy3);
 785             dz13             = _mm256_sub_ps(iz1,jz3);
 786             dx21             = _mm256_sub_ps(ix2,jx1);
 787             dy21             = _mm256_sub_ps(iy2,jy1);
 788             dz21             = _mm256_sub_ps(iz2,jz1);
 789             dx22             = _mm256_sub_ps(ix2,jx2);
 790             dy22             = _mm256_sub_ps(iy2,jy2);
 791             dz22             = _mm256_sub_ps(iz2,jz2);
 792             dx23             = _mm256_sub_ps(ix2,jx3);
 793             dy23             = _mm256_sub_ps(iy2,jy3);
 794             dz23             = _mm256_sub_ps(iz2,jz3);
 795             dx31             = _mm256_sub_ps(ix3,jx1);
 796             dy31             = _mm256_sub_ps(iy3,jy1);
 797             dz31             = _mm256_sub_ps(iz3,jz1);
 798             dx32             = _mm256_sub_ps(ix3,jx2);
 799             dy32             = _mm256_sub_ps(iy3,jy2);
 800             dz32             = _mm256_sub_ps(iz3,jz2);
 801             dx33             = _mm256_sub_ps(ix3,jx3);
 802             dy33             = _mm256_sub_ps(iy3,jy3);
 803             dz33             = _mm256_sub_ps(iz3,jz3);
 804
 805             /* Calculate squared distance and things based on it */
 806             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 807             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 808             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 809             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 810             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 811             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 812             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 813             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 814             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 815             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 816
 817             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 818             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 819             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 820             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 821             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 822             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 823             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 824             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 825             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 826
 827             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 828             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 829             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 830             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 831             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 832             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 833             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 834             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 835             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 836             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 837
 838             fjx0             = _mm256_setzero_ps();
 839             fjy0             = _mm256_setzero_ps();
 840             fjz0             = _mm256_setzero_ps();
 841             fjx1             = _mm256_setzero_ps();
 842             fjy1             = _mm256_setzero_ps();
 843             fjz1             = _mm256_setzero_ps();
 844             fjx2             = _mm256_setzero_ps();
 845             fjy2             = _mm256_setzero_ps();
 846             fjz2             = _mm256_setzero_ps();
 847             fjx3             = _mm256_setzero_ps();
 848             fjy3             = _mm256_setzero_ps();
 849             fjz3             = _mm256_setzero_ps();
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             /* LENNARD-JONES DISPERSION/REPULSION */
 856
 857             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 858             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 859             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 860             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 861             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 862
 863             /* Update potential sum for this i atom from the interaction with this j atom. */
 864             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 865             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 866
 867             fscal            = fvdw;
 868
 869             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 870
 871             /* Calculate temporary vectorial force */
 872             tx               = _mm256_mul_ps(fscal,dx00);
 873             ty               = _mm256_mul_ps(fscal,dy00);
 874             tz               = _mm256_mul_ps(fscal,dz00);
 875
 876             /* Update vectorial force */
 877             fix0             = _mm256_add_ps(fix0,tx);
 878             fiy0             = _mm256_add_ps(fiy0,ty);
 879             fiz0             = _mm256_add_ps(fiz0,tz);
 880
 881             fjx0             = _mm256_add_ps(fjx0,tx);
 882             fjy0             = _mm256_add_ps(fjy0,ty);
 883             fjz0             = _mm256_add_ps(fjz0,tz);
 884
 885             /**************************
 886              * CALCULATE INTERACTIONS *
 887              **************************/
 888
 889             r11              = _mm256_mul_ps(rsq11,rinv11);
 890             r11              = _mm256_andnot_ps(dummy_mask,r11);
 891
 892             /* EWALD ELECTROSTATICS */
 893
 894             /* Analytical PME correction */
 895             zeta2            = _mm256_mul_ps(beta2,rsq11);
 896             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 897             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 898             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 899             felec            = _mm256_mul_ps(qq11,felec);
 900             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 901             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 902             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 903             velec            = _mm256_mul_ps(qq11,velec);
 904
 905             /* Update potential sum for this i atom from the interaction with this j atom. */
 906             velec            = _mm256_andnot_ps(dummy_mask,velec);
 907             velecsum         = _mm256_add_ps(velecsum,velec);
 908
 909             fscal            = felec;
 910
 911             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 912
 913             /* Calculate temporary vectorial force */
 914             tx               = _mm256_mul_ps(fscal,dx11);
 915             ty               = _mm256_mul_ps(fscal,dy11);
 916             tz               = _mm256_mul_ps(fscal,dz11);
 917
 918             /* Update vectorial force */
 919             fix1             = _mm256_add_ps(fix1,tx);
 920             fiy1             = _mm256_add_ps(fiy1,ty);
 921             fiz1             = _mm256_add_ps(fiz1,tz);
 922
 923             fjx1             = _mm256_add_ps(fjx1,tx);
 924             fjy1             = _mm256_add_ps(fjy1,ty);
 925             fjz1             = _mm256_add_ps(fjz1,tz);
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             r12              = _mm256_mul_ps(rsq12,rinv12);
 932             r12              = _mm256_andnot_ps(dummy_mask,r12);
 933
 934             /* EWALD ELECTROSTATICS */
 935
 936             /* Analytical PME correction */
 937             zeta2            = _mm256_mul_ps(beta2,rsq12);
 938             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 939             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 940             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 941             felec            = _mm256_mul_ps(qq12,felec);
 942             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 943             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 944             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 945             velec            = _mm256_mul_ps(qq12,velec);
 946
 947             /* Update potential sum for this i atom from the interaction with this j atom. */
 948             velec            = _mm256_andnot_ps(dummy_mask,velec);
 949             velecsum         = _mm256_add_ps(velecsum,velec);
 950
 951             fscal            = felec;
 952
 953             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 954
 955             /* Calculate temporary vectorial force */
 956             tx               = _mm256_mul_ps(fscal,dx12);
 957             ty               = _mm256_mul_ps(fscal,dy12);
 958             tz               = _mm256_mul_ps(fscal,dz12);
 959
 960             /* Update vectorial force */
 961             fix1             = _mm256_add_ps(fix1,tx);
 962             fiy1             = _mm256_add_ps(fiy1,ty);
 963             fiz1             = _mm256_add_ps(fiz1,tz);
 964
 965             fjx2             = _mm256_add_ps(fjx2,tx);
 966             fjy2             = _mm256_add_ps(fjy2,ty);
 967             fjz2             = _mm256_add_ps(fjz2,tz);
 968
 969             /**************************
 970              * CALCULATE INTERACTIONS *
 971              **************************/
 972
 973             r13              = _mm256_mul_ps(rsq13,rinv13);
 974             r13              = _mm256_andnot_ps(dummy_mask,r13);
 975
 976             /* EWALD ELECTROSTATICS */
 977
 978             /* Analytical PME correction */
 979             zeta2            = _mm256_mul_ps(beta2,rsq13);
 980             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
 981             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 982             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 983             felec            = _mm256_mul_ps(qq13,felec);
 984             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 985             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 986             velec            = _mm256_sub_ps(rinv13,pmecorrV);
 987             velec            = _mm256_mul_ps(qq13,velec);
 988
 989             /* Update potential sum for this i atom from the interaction with this j atom. */
 990             velec            = _mm256_andnot_ps(dummy_mask,velec);
 991             velecsum         = _mm256_add_ps(velecsum,velec);
 992
 993             fscal            = felec;
 994
 995             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 996
 997             /* Calculate temporary vectorial force */
 998             tx               = _mm256_mul_ps(fscal,dx13);
 999             ty               = _mm256_mul_ps(fscal,dy13);
1000             tz               = _mm256_mul_ps(fscal,dz13);
1001
1002             /* Update vectorial force */
1003             fix1             = _mm256_add_ps(fix1,tx);
1004             fiy1             = _mm256_add_ps(fiy1,ty);
1005             fiz1             = _mm256_add_ps(fiz1,tz);
1006
1007             fjx3             = _mm256_add_ps(fjx3,tx);
1008             fjy3             = _mm256_add_ps(fjy3,ty);
1009             fjz3             = _mm256_add_ps(fjz3,tz);
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             r21              = _mm256_mul_ps(rsq21,rinv21);
1016             r21              = _mm256_andnot_ps(dummy_mask,r21);
1017
1018             /* EWALD ELECTROSTATICS */
1019
1020             /* Analytical PME correction */
1021             zeta2            = _mm256_mul_ps(beta2,rsq21);
1022             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1023             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1024             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1025             felec            = _mm256_mul_ps(qq21,felec);
1026             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1027             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1028             velec            = _mm256_sub_ps(rinv21,pmecorrV);
1029             velec            = _mm256_mul_ps(qq21,velec);
1030
1031             /* Update potential sum for this i atom from the interaction with this j atom. */
1032             velec            = _mm256_andnot_ps(dummy_mask,velec);
1033             velecsum         = _mm256_add_ps(velecsum,velec);
1034
1035             fscal            = felec;
1036
1037             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1038
1039             /* Calculate temporary vectorial force */
1040             tx               = _mm256_mul_ps(fscal,dx21);
1041             ty               = _mm256_mul_ps(fscal,dy21);
1042             tz               = _mm256_mul_ps(fscal,dz21);
1043
1044             /* Update vectorial force */
1045             fix2             = _mm256_add_ps(fix2,tx);
1046             fiy2             = _mm256_add_ps(fiy2,ty);
1047             fiz2             = _mm256_add_ps(fiz2,tz);
1048
1049             fjx1             = _mm256_add_ps(fjx1,tx);
1050             fjy1             = _mm256_add_ps(fjy1,ty);
1051             fjz1             = _mm256_add_ps(fjz1,tz);
1052
1053             /**************************
1054              * CALCULATE INTERACTIONS *
1055              **************************/
1056
1057             r22              = _mm256_mul_ps(rsq22,rinv22);
1058             r22              = _mm256_andnot_ps(dummy_mask,r22);
1059
1060             /* EWALD ELECTROSTATICS */
1061
1062             /* Analytical PME correction */
1063             zeta2            = _mm256_mul_ps(beta2,rsq22);
1064             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1065             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1066             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1067             felec            = _mm256_mul_ps(qq22,felec);
1068             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1069             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1070             velec            = _mm256_sub_ps(rinv22,pmecorrV);
1071             velec            = _mm256_mul_ps(qq22,velec);
1072
1073             /* Update potential sum for this i atom from the interaction with this j atom. */
1074             velec            = _mm256_andnot_ps(dummy_mask,velec);
1075             velecsum         = _mm256_add_ps(velecsum,velec);
1076
1077             fscal            = felec;
1078
1079             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1080
1081             /* Calculate temporary vectorial force */
1082             tx               = _mm256_mul_ps(fscal,dx22);
1083             ty               = _mm256_mul_ps(fscal,dy22);
1084             tz               = _mm256_mul_ps(fscal,dz22);
1085
1086             /* Update vectorial force */
1087             fix2             = _mm256_add_ps(fix2,tx);
1088             fiy2             = _mm256_add_ps(fiy2,ty);
1089             fiz2             = _mm256_add_ps(fiz2,tz);
1090
1091             fjx2             = _mm256_add_ps(fjx2,tx);
1092             fjy2             = _mm256_add_ps(fjy2,ty);
1093             fjz2             = _mm256_add_ps(fjz2,tz);
1094
1095             /**************************
1096              * CALCULATE INTERACTIONS *
1097              **************************/
1098
1099             r23              = _mm256_mul_ps(rsq23,rinv23);
1100             r23              = _mm256_andnot_ps(dummy_mask,r23);
1101
1102             /* EWALD ELECTROSTATICS */
1103
1104             /* Analytical PME correction */
1105             zeta2            = _mm256_mul_ps(beta2,rsq23);
1106             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1107             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1108             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1109             felec            = _mm256_mul_ps(qq23,felec);
1110             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1111             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1112             velec            = _mm256_sub_ps(rinv23,pmecorrV);
1113             velec            = _mm256_mul_ps(qq23,velec);
1114
1115             /* Update potential sum for this i atom from the interaction with this j atom. */
1116             velec            = _mm256_andnot_ps(dummy_mask,velec);
1117             velecsum         = _mm256_add_ps(velecsum,velec);
1118
1119             fscal            = felec;
1120
1121             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1122
1123             /* Calculate temporary vectorial force */
1124             tx               = _mm256_mul_ps(fscal,dx23);
1125             ty               = _mm256_mul_ps(fscal,dy23);
1126             tz               = _mm256_mul_ps(fscal,dz23);
1127
1128             /* Update vectorial force */
1129             fix2             = _mm256_add_ps(fix2,tx);
1130             fiy2             = _mm256_add_ps(fiy2,ty);
1131             fiz2             = _mm256_add_ps(fiz2,tz);
1132
1133             fjx3             = _mm256_add_ps(fjx3,tx);
1134             fjy3             = _mm256_add_ps(fjy3,ty);
1135             fjz3             = _mm256_add_ps(fjz3,tz);
1136
1137             /**************************
1138              * CALCULATE INTERACTIONS *
1139              **************************/
1140
1141             r31              = _mm256_mul_ps(rsq31,rinv31);
1142             r31              = _mm256_andnot_ps(dummy_mask,r31);
1143
1144             /* EWALD ELECTROSTATICS */
1145
1146             /* Analytical PME correction */
1147             zeta2            = _mm256_mul_ps(beta2,rsq31);
1148             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1149             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1150             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1151             felec            = _mm256_mul_ps(qq31,felec);
1152             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1153             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1154             velec            = _mm256_sub_ps(rinv31,pmecorrV);
1155             velec            = _mm256_mul_ps(qq31,velec);
1156
1157             /* Update potential sum for this i atom from the interaction with this j atom. */
1158             velec            = _mm256_andnot_ps(dummy_mask,velec);
1159             velecsum         = _mm256_add_ps(velecsum,velec);
1160
1161             fscal            = felec;
1162
1163             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1164
1165             /* Calculate temporary vectorial force */
1166             tx               = _mm256_mul_ps(fscal,dx31);
1167             ty               = _mm256_mul_ps(fscal,dy31);
1168             tz               = _mm256_mul_ps(fscal,dz31);
1169
1170             /* Update vectorial force */
1171             fix3             = _mm256_add_ps(fix3,tx);
1172             fiy3             = _mm256_add_ps(fiy3,ty);
1173             fiz3             = _mm256_add_ps(fiz3,tz);
1174
1175             fjx1             = _mm256_add_ps(fjx1,tx);
1176             fjy1             = _mm256_add_ps(fjy1,ty);
1177             fjz1             = _mm256_add_ps(fjz1,tz);
1178
1179             /**************************
1180              * CALCULATE INTERACTIONS *
1181              **************************/
1182
1183             r32              = _mm256_mul_ps(rsq32,rinv32);
1184             r32              = _mm256_andnot_ps(dummy_mask,r32);
1185
1186             /* EWALD ELECTROSTATICS */
1187
1188             /* Analytical PME correction */
1189             zeta2            = _mm256_mul_ps(beta2,rsq32);
1190             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1191             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1192             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1193             felec            = _mm256_mul_ps(qq32,felec);
1194             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1195             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1196             velec            = _mm256_sub_ps(rinv32,pmecorrV);
1197             velec            = _mm256_mul_ps(qq32,velec);
1198
1199             /* Update potential sum for this i atom from the interaction with this j atom. */
1200             velec            = _mm256_andnot_ps(dummy_mask,velec);
1201             velecsum         = _mm256_add_ps(velecsum,velec);
1202
1203             fscal            = felec;
1204
1205             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1206
1207             /* Calculate temporary vectorial force */
1208             tx               = _mm256_mul_ps(fscal,dx32);
1209             ty               = _mm256_mul_ps(fscal,dy32);
1210             tz               = _mm256_mul_ps(fscal,dz32);
1211
1212             /* Update vectorial force */
1213             fix3             = _mm256_add_ps(fix3,tx);
1214             fiy3             = _mm256_add_ps(fiy3,ty);
1215             fiz3             = _mm256_add_ps(fiz3,tz);
1216
1217             fjx2             = _mm256_add_ps(fjx2,tx);
1218             fjy2             = _mm256_add_ps(fjy2,ty);
1219             fjz2             = _mm256_add_ps(fjz2,tz);
1220
1221             /**************************
1222              * CALCULATE INTERACTIONS *
1223              **************************/
1224
1225             r33              = _mm256_mul_ps(rsq33,rinv33);
1226             r33              = _mm256_andnot_ps(dummy_mask,r33);
1227
1228             /* EWALD ELECTROSTATICS */
1229
1230             /* Analytical PME correction */
1231             zeta2            = _mm256_mul_ps(beta2,rsq33);
1232             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1233             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1234             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1235             felec            = _mm256_mul_ps(qq33,felec);
1236             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1237             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1238             velec            = _mm256_sub_ps(rinv33,pmecorrV);
1239             velec            = _mm256_mul_ps(qq33,velec);
1240
1241             /* Update potential sum for this i atom from the interaction with this j atom. */
1242             velec            = _mm256_andnot_ps(dummy_mask,velec);
1243             velecsum         = _mm256_add_ps(velecsum,velec);
1244
1245             fscal            = felec;
1246
1247             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1248
1249             /* Calculate temporary vectorial force */
1250             tx               = _mm256_mul_ps(fscal,dx33);
1251             ty               = _mm256_mul_ps(fscal,dy33);
1252             tz               = _mm256_mul_ps(fscal,dz33);
1253
1254             /* Update vectorial force */
1255             fix3             = _mm256_add_ps(fix3,tx);
1256             fiy3             = _mm256_add_ps(fiy3,ty);
1257             fiz3             = _mm256_add_ps(fiz3,tz);
1258
1259             fjx3             = _mm256_add_ps(fjx3,tx);
1260             fjy3             = _mm256_add_ps(fjy3,ty);
1261             fjz3             = _mm256_add_ps(fjz3,tz);
1262
1263             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1264             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1265             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1266             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1267             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1268             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1269             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1270             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1271
1272             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1273                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1274                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1275
1276             /* Inner loop uses 800 flops */
1277         }
1278
1279         /* End of innermost loop */
1280
1281         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1282                                                  f+i_coord_offset,fshift+i_shift_offset);
1283
1284         ggid                        = gid[iidx];
1285         /* Update potential energies */
1286         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1287         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1288
1289         /* Increment number of inner iterations */
1290         inneriter                  += j_index_end - j_index_start;
1291
1292         /* Outer loop uses 26 flops */
1293     }
1294
1295     /* Increment number of outer iterations */
1296     outeriter        += nri;
1297
1298     /* Update outer/inner flops */
1299
1300     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*800);
1301 }
1302 /*
1303  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_single
1304  * Electrostatics interaction: Ewald
1305  * VdW interaction:            LennardJones
1306  * Geometry:                   Water4-Water4
1307  * Calculate force/pot:        Force
1308  */
1309 void
1310 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_avx_256_single
1311                     (t_nblist                    * gmx_restrict       nlist,
1312                      rvec                        * gmx_restrict          xx,
1313                      rvec                        * gmx_restrict          ff,
1314                      t_forcerec                  * gmx_restrict          fr,
1315                      t_mdatoms                   * gmx_restrict     mdatoms,
1316                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1317                      t_nrnb                      * gmx_restrict        nrnb)
1318 {
1319     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1320      * just 0 for non-waters.
1321      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1322      * jnr indices corresponding to data put in the four positions in the SIMD register.
1323      */
1324     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1325     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1326     int              jnrA,jnrB,jnrC,jnrD;
1327     int              jnrE,jnrF,jnrG,jnrH;
1328     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1329     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1330     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1331     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1332     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1333     real             rcutoff_scalar;
1334     real             *shiftvec,*fshift,*x,*f;
1335     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1336     real             scratch[4*DIM];
1337     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1338     real *           vdwioffsetptr0;
1339     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1340     real *           vdwioffsetptr1;
1341     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1342     real *           vdwioffsetptr2;
1343     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1344     real *           vdwioffsetptr3;
1345     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1346     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1347     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1348     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1349     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1350     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1351     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1352     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1353     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1354     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1355     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1356     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1357     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1358     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1359     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1360     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1361     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1362     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1363     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1364     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1365     real             *charge;
1366     int              nvdwtype;
1367     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1368     int              *vdwtype;
1369     real             *vdwparam;
1370     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1371     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1372     __m256i          ewitab;
1373     __m128i          ewitab_lo,ewitab_hi;
1374     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1375     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1376     real             *ewtab;
1377     __m256           dummy_mask,cutoff_mask;
1378     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1379     __m256           one     = _mm256_set1_ps(1.0);
1380     __m256           two     = _mm256_set1_ps(2.0);
1381     x                = xx[0];
1382     f                = ff[0];
1383
1384     nri              = nlist->nri;
1385     iinr             = nlist->iinr;
1386     jindex           = nlist->jindex;
1387     jjnr             = nlist->jjnr;
1388     shiftidx         = nlist->shift;
1389     gid              = nlist->gid;
1390     shiftvec         = fr->shift_vec[0];
1391     fshift           = fr->fshift[0];
1392     facel            = _mm256_set1_ps(fr->epsfac);
1393     charge           = mdatoms->chargeA;
1394     nvdwtype         = fr->ntype;
1395     vdwparam         = fr->nbfp;
1396     vdwtype          = mdatoms->typeA;
1397
1398     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
1399     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1400     beta2            = _mm256_mul_ps(beta,beta);
1401     beta3            = _mm256_mul_ps(beta,beta2);
1402
1403     ewtab            = fr->ic->tabq_coul_F;
1404     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
1405     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1406
1407     /* Setup water-specific parameters */
1408     inr              = nlist->iinr[0];
1409     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1410     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1411     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1412     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1413
1414     jq1              = _mm256_set1_ps(charge[inr+1]);
1415     jq2              = _mm256_set1_ps(charge[inr+2]);
1416     jq3              = _mm256_set1_ps(charge[inr+3]);
1417     vdwjidx0A        = 2*vdwtype[inr+0];
1418     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1419     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1420     qq11             = _mm256_mul_ps(iq1,jq1);
1421     qq12             = _mm256_mul_ps(iq1,jq2);
1422     qq13             = _mm256_mul_ps(iq1,jq3);
1423     qq21             = _mm256_mul_ps(iq2,jq1);
1424     qq22             = _mm256_mul_ps(iq2,jq2);
1425     qq23             = _mm256_mul_ps(iq2,jq3);
1426     qq31             = _mm256_mul_ps(iq3,jq1);
1427     qq32             = _mm256_mul_ps(iq3,jq2);
1428     qq33             = _mm256_mul_ps(iq3,jq3);
1429
1430     /* Avoid stupid compiler warnings */
1431     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1432     j_coord_offsetA = 0;
1433     j_coord_offsetB = 0;
1434     j_coord_offsetC = 0;
1435     j_coord_offsetD = 0;
1436     j_coord_offsetE = 0;
1437     j_coord_offsetF = 0;
1438     j_coord_offsetG = 0;
1439     j_coord_offsetH = 0;
1440
1441     outeriter        = 0;
1442     inneriter        = 0;
1443
1444     for(iidx=0;iidx<4*DIM;iidx++)
1445     {
1446         scratch[iidx] = 0.0;
1447     }
1448
1449     /* Start outer loop over neighborlists */
1450     for(iidx=0; iidx<nri; iidx++)
1451     {
1452         /* Load shift vector for this list */
1453         i_shift_offset   = DIM*shiftidx[iidx];
1454
1455         /* Load limits for loop over neighbors */
1456         j_index_start    = jindex[iidx];
1457         j_index_end      = jindex[iidx+1];
1458
1459         /* Get outer coordinate index */
1460         inr              = iinr[iidx];
1461         i_coord_offset   = DIM*inr;
1462
1463         /* Load i particle coords and add shift vector */
1464         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1465                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1466
1467         fix0             = _mm256_setzero_ps();
1468         fiy0             = _mm256_setzero_ps();
1469         fiz0             = _mm256_setzero_ps();
1470         fix1             = _mm256_setzero_ps();
1471         fiy1             = _mm256_setzero_ps();
1472         fiz1             = _mm256_setzero_ps();
1473         fix2             = _mm256_setzero_ps();
1474         fiy2             = _mm256_setzero_ps();
1475         fiz2             = _mm256_setzero_ps();
1476         fix3             = _mm256_setzero_ps();
1477         fiy3             = _mm256_setzero_ps();
1478         fiz3             = _mm256_setzero_ps();
1479
1480         /* Start inner kernel loop */
1481         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1482         {
1483
1484             /* Get j neighbor index, and coordinate index */
1485             jnrA             = jjnr[jidx];
1486             jnrB             = jjnr[jidx+1];
1487             jnrC             = jjnr[jidx+2];
1488             jnrD             = jjnr[jidx+3];
1489             jnrE             = jjnr[jidx+4];
1490             jnrF             = jjnr[jidx+5];
1491             jnrG             = jjnr[jidx+6];
1492             jnrH             = jjnr[jidx+7];
1493             j_coord_offsetA  = DIM*jnrA;
1494             j_coord_offsetB  = DIM*jnrB;
1495             j_coord_offsetC  = DIM*jnrC;
1496             j_coord_offsetD  = DIM*jnrD;
1497             j_coord_offsetE  = DIM*jnrE;
1498             j_coord_offsetF  = DIM*jnrF;
1499             j_coord_offsetG  = DIM*jnrG;
1500             j_coord_offsetH  = DIM*jnrH;
1501
1502             /* load j atom coordinates */
1503             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1504                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1505                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1506                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1507                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1508                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1509
1510             /* Calculate displacement vector */
1511             dx00             = _mm256_sub_ps(ix0,jx0);
1512             dy00             = _mm256_sub_ps(iy0,jy0);
1513             dz00             = _mm256_sub_ps(iz0,jz0);
1514             dx11             = _mm256_sub_ps(ix1,jx1);
1515             dy11             = _mm256_sub_ps(iy1,jy1);
1516             dz11             = _mm256_sub_ps(iz1,jz1);
1517             dx12             = _mm256_sub_ps(ix1,jx2);
1518             dy12             = _mm256_sub_ps(iy1,jy2);
1519             dz12             = _mm256_sub_ps(iz1,jz2);
1520             dx13             = _mm256_sub_ps(ix1,jx3);
1521             dy13             = _mm256_sub_ps(iy1,jy3);
1522             dz13             = _mm256_sub_ps(iz1,jz3);
1523             dx21             = _mm256_sub_ps(ix2,jx1);
1524             dy21             = _mm256_sub_ps(iy2,jy1);
1525             dz21             = _mm256_sub_ps(iz2,jz1);
1526             dx22             = _mm256_sub_ps(ix2,jx2);
1527             dy22             = _mm256_sub_ps(iy2,jy2);
1528             dz22             = _mm256_sub_ps(iz2,jz2);
1529             dx23             = _mm256_sub_ps(ix2,jx3);
1530             dy23             = _mm256_sub_ps(iy2,jy3);
1531             dz23             = _mm256_sub_ps(iz2,jz3);
1532             dx31             = _mm256_sub_ps(ix3,jx1);
1533             dy31             = _mm256_sub_ps(iy3,jy1);
1534             dz31             = _mm256_sub_ps(iz3,jz1);
1535             dx32             = _mm256_sub_ps(ix3,jx2);
1536             dy32             = _mm256_sub_ps(iy3,jy2);
1537             dz32             = _mm256_sub_ps(iz3,jz2);
1538             dx33             = _mm256_sub_ps(ix3,jx3);
1539             dy33             = _mm256_sub_ps(iy3,jy3);
1540             dz33             = _mm256_sub_ps(iz3,jz3);
1541
1542             /* Calculate squared distance and things based on it */
1543             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1544             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1545             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1546             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1547             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1548             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1549             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1550             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1551             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1552             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1553
1554             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1555             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1556             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1557             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1558             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1559             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1560             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1561             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1562             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1563
1564             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1565             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1566             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1567             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1568             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1569             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1570             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1571             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1572             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1573             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1574
1575             fjx0             = _mm256_setzero_ps();
1576             fjy0             = _mm256_setzero_ps();
1577             fjz0             = _mm256_setzero_ps();
1578             fjx1             = _mm256_setzero_ps();
1579             fjy1             = _mm256_setzero_ps();
1580             fjz1             = _mm256_setzero_ps();
1581             fjx2             = _mm256_setzero_ps();
1582             fjy2             = _mm256_setzero_ps();
1583             fjz2             = _mm256_setzero_ps();
1584             fjx3             = _mm256_setzero_ps();
1585             fjy3             = _mm256_setzero_ps();
1586             fjz3             = _mm256_setzero_ps();
1587
1588             /**************************
1589              * CALCULATE INTERACTIONS *
1590              **************************/
1591
1592             /* LENNARD-JONES DISPERSION/REPULSION */
1593
1594             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1595             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1596
1597             fscal            = fvdw;
1598
1599             /* Calculate temporary vectorial force */
1600             tx               = _mm256_mul_ps(fscal,dx00);
1601             ty               = _mm256_mul_ps(fscal,dy00);
1602             tz               = _mm256_mul_ps(fscal,dz00);
1603
1604             /* Update vectorial force */
1605             fix0             = _mm256_add_ps(fix0,tx);
1606             fiy0             = _mm256_add_ps(fiy0,ty);
1607             fiz0             = _mm256_add_ps(fiz0,tz);
1608
1609             fjx0             = _mm256_add_ps(fjx0,tx);
1610             fjy0             = _mm256_add_ps(fjy0,ty);
1611             fjz0             = _mm256_add_ps(fjz0,tz);
1612
1613             /**************************
1614              * CALCULATE INTERACTIONS *
1615              **************************/
1616
1617             r11              = _mm256_mul_ps(rsq11,rinv11);
1618
1619             /* EWALD ELECTROSTATICS */
1620
1621             /* Analytical PME correction */
1622             zeta2            = _mm256_mul_ps(beta2,rsq11);
1623             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
1624             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1625             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1626             felec            = _mm256_mul_ps(qq11,felec);
1627
1628             fscal            = felec;
1629
1630             /* Calculate temporary vectorial force */
1631             tx               = _mm256_mul_ps(fscal,dx11);
1632             ty               = _mm256_mul_ps(fscal,dy11);
1633             tz               = _mm256_mul_ps(fscal,dz11);
1634
1635             /* Update vectorial force */
1636             fix1             = _mm256_add_ps(fix1,tx);
1637             fiy1             = _mm256_add_ps(fiy1,ty);
1638             fiz1             = _mm256_add_ps(fiz1,tz);
1639
1640             fjx1             = _mm256_add_ps(fjx1,tx);
1641             fjy1             = _mm256_add_ps(fjy1,ty);
1642             fjz1             = _mm256_add_ps(fjz1,tz);
1643
1644             /**************************
1645              * CALCULATE INTERACTIONS *
1646              **************************/
1647
1648             r12              = _mm256_mul_ps(rsq12,rinv12);
1649
1650             /* EWALD ELECTROSTATICS */
1651
1652             /* Analytical PME correction */
1653             zeta2            = _mm256_mul_ps(beta2,rsq12);
1654             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
1655             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1656             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1657             felec            = _mm256_mul_ps(qq12,felec);
1658
1659             fscal            = felec;
1660
1661             /* Calculate temporary vectorial force */
1662             tx               = _mm256_mul_ps(fscal,dx12);
1663             ty               = _mm256_mul_ps(fscal,dy12);
1664             tz               = _mm256_mul_ps(fscal,dz12);
1665
1666             /* Update vectorial force */
1667             fix1             = _mm256_add_ps(fix1,tx);
1668             fiy1             = _mm256_add_ps(fiy1,ty);
1669             fiz1             = _mm256_add_ps(fiz1,tz);
1670
1671             fjx2             = _mm256_add_ps(fjx2,tx);
1672             fjy2             = _mm256_add_ps(fjy2,ty);
1673             fjz2             = _mm256_add_ps(fjz2,tz);
1674
1675             /**************************
1676              * CALCULATE INTERACTIONS *
1677              **************************/
1678
1679             r13              = _mm256_mul_ps(rsq13,rinv13);
1680
1681             /* EWALD ELECTROSTATICS */
1682
1683             /* Analytical PME correction */
1684             zeta2            = _mm256_mul_ps(beta2,rsq13);
1685             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
1686             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1687             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1688             felec            = _mm256_mul_ps(qq13,felec);
1689
1690             fscal            = felec;
1691
1692             /* Calculate temporary vectorial force */
1693             tx               = _mm256_mul_ps(fscal,dx13);
1694             ty               = _mm256_mul_ps(fscal,dy13);
1695             tz               = _mm256_mul_ps(fscal,dz13);
1696
1697             /* Update vectorial force */
1698             fix1             = _mm256_add_ps(fix1,tx);
1699             fiy1             = _mm256_add_ps(fiy1,ty);
1700             fiz1             = _mm256_add_ps(fiz1,tz);
1701
1702             fjx3             = _mm256_add_ps(fjx3,tx);
1703             fjy3             = _mm256_add_ps(fjy3,ty);
1704             fjz3             = _mm256_add_ps(fjz3,tz);
1705
1706             /**************************
1707              * CALCULATE INTERACTIONS *
1708              **************************/
1709
1710             r21              = _mm256_mul_ps(rsq21,rinv21);
1711
1712             /* EWALD ELECTROSTATICS */
1713
1714             /* Analytical PME correction */
1715             zeta2            = _mm256_mul_ps(beta2,rsq21);
1716             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1717             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1718             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1719             felec            = _mm256_mul_ps(qq21,felec);
1720
1721             fscal            = felec;
1722
1723             /* Calculate temporary vectorial force */
1724             tx               = _mm256_mul_ps(fscal,dx21);
1725             ty               = _mm256_mul_ps(fscal,dy21);
1726             tz               = _mm256_mul_ps(fscal,dz21);
1727
1728             /* Update vectorial force */
1729             fix2             = _mm256_add_ps(fix2,tx);
1730             fiy2             = _mm256_add_ps(fiy2,ty);
1731             fiz2             = _mm256_add_ps(fiz2,tz);
1732
1733             fjx1             = _mm256_add_ps(fjx1,tx);
1734             fjy1             = _mm256_add_ps(fjy1,ty);
1735             fjz1             = _mm256_add_ps(fjz1,tz);
1736
1737             /**************************
1738              * CALCULATE INTERACTIONS *
1739              **************************/
1740
1741             r22              = _mm256_mul_ps(rsq22,rinv22);
1742
1743             /* EWALD ELECTROSTATICS */
1744
1745             /* Analytical PME correction */
1746             zeta2            = _mm256_mul_ps(beta2,rsq22);
1747             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1748             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1749             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1750             felec            = _mm256_mul_ps(qq22,felec);
1751
1752             fscal            = felec;
1753
1754             /* Calculate temporary vectorial force */
1755             tx               = _mm256_mul_ps(fscal,dx22);
1756             ty               = _mm256_mul_ps(fscal,dy22);
1757             tz               = _mm256_mul_ps(fscal,dz22);
1758
1759             /* Update vectorial force */
1760             fix2             = _mm256_add_ps(fix2,tx);
1761             fiy2             = _mm256_add_ps(fiy2,ty);
1762             fiz2             = _mm256_add_ps(fiz2,tz);
1763
1764             fjx2             = _mm256_add_ps(fjx2,tx);
1765             fjy2             = _mm256_add_ps(fjy2,ty);
1766             fjz2             = _mm256_add_ps(fjz2,tz);
1767
1768             /**************************
1769              * CALCULATE INTERACTIONS *
1770              **************************/
1771
1772             r23              = _mm256_mul_ps(rsq23,rinv23);
1773
1774             /* EWALD ELECTROSTATICS */
1775
1776             /* Analytical PME correction */
1777             zeta2            = _mm256_mul_ps(beta2,rsq23);
1778             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
1779             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1780             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1781             felec            = _mm256_mul_ps(qq23,felec);
1782
1783             fscal            = felec;
1784
1785             /* Calculate temporary vectorial force */
1786             tx               = _mm256_mul_ps(fscal,dx23);
1787             ty               = _mm256_mul_ps(fscal,dy23);
1788             tz               = _mm256_mul_ps(fscal,dz23);
1789
1790             /* Update vectorial force */
1791             fix2             = _mm256_add_ps(fix2,tx);
1792             fiy2             = _mm256_add_ps(fiy2,ty);
1793             fiz2             = _mm256_add_ps(fiz2,tz);
1794
1795             fjx3             = _mm256_add_ps(fjx3,tx);
1796             fjy3             = _mm256_add_ps(fjy3,ty);
1797             fjz3             = _mm256_add_ps(fjz3,tz);
1798
1799             /**************************
1800              * CALCULATE INTERACTIONS *
1801              **************************/
1802
1803             r31              = _mm256_mul_ps(rsq31,rinv31);
1804
1805             /* EWALD ELECTROSTATICS */
1806
1807             /* Analytical PME correction */
1808             zeta2            = _mm256_mul_ps(beta2,rsq31);
1809             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
1810             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1811             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1812             felec            = _mm256_mul_ps(qq31,felec);
1813
1814             fscal            = felec;
1815
1816             /* Calculate temporary vectorial force */
1817             tx               = _mm256_mul_ps(fscal,dx31);
1818             ty               = _mm256_mul_ps(fscal,dy31);
1819             tz               = _mm256_mul_ps(fscal,dz31);
1820
1821             /* Update vectorial force */
1822             fix3             = _mm256_add_ps(fix3,tx);
1823             fiy3             = _mm256_add_ps(fiy3,ty);
1824             fiz3             = _mm256_add_ps(fiz3,tz);
1825
1826             fjx1             = _mm256_add_ps(fjx1,tx);
1827             fjy1             = _mm256_add_ps(fjy1,ty);
1828             fjz1             = _mm256_add_ps(fjz1,tz);
1829
1830             /**************************
1831              * CALCULATE INTERACTIONS *
1832              **************************/
1833
1834             r32              = _mm256_mul_ps(rsq32,rinv32);
1835
1836             /* EWALD ELECTROSTATICS */
1837
1838             /* Analytical PME correction */
1839             zeta2            = _mm256_mul_ps(beta2,rsq32);
1840             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
1841             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1842             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1843             felec            = _mm256_mul_ps(qq32,felec);
1844
1845             fscal            = felec;
1846
1847             /* Calculate temporary vectorial force */
1848             tx               = _mm256_mul_ps(fscal,dx32);
1849             ty               = _mm256_mul_ps(fscal,dy32);
1850             tz               = _mm256_mul_ps(fscal,dz32);
1851
1852             /* Update vectorial force */
1853             fix3             = _mm256_add_ps(fix3,tx);
1854             fiy3             = _mm256_add_ps(fiy3,ty);
1855             fiz3             = _mm256_add_ps(fiz3,tz);
1856
1857             fjx2             = _mm256_add_ps(fjx2,tx);
1858             fjy2             = _mm256_add_ps(fjy2,ty);
1859             fjz2             = _mm256_add_ps(fjz2,tz);
1860
1861             /**************************
1862              * CALCULATE INTERACTIONS *
1863              **************************/
1864
1865             r33              = _mm256_mul_ps(rsq33,rinv33);
1866
1867             /* EWALD ELECTROSTATICS */
1868
1869             /* Analytical PME correction */
1870             zeta2            = _mm256_mul_ps(beta2,rsq33);
1871             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
1872             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1873             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1874             felec            = _mm256_mul_ps(qq33,felec);
1875
1876             fscal            = felec;
1877
1878             /* Calculate temporary vectorial force */
1879             tx               = _mm256_mul_ps(fscal,dx33);
1880             ty               = _mm256_mul_ps(fscal,dy33);
1881             tz               = _mm256_mul_ps(fscal,dz33);
1882
1883             /* Update vectorial force */
1884             fix3             = _mm256_add_ps(fix3,tx);
1885             fiy3             = _mm256_add_ps(fiy3,ty);
1886             fiz3             = _mm256_add_ps(fiz3,tz);
1887
1888             fjx3             = _mm256_add_ps(fjx3,tx);
1889             fjy3             = _mm256_add_ps(fjy3,ty);
1890             fjz3             = _mm256_add_ps(fjz3,tz);
1891
1892             fjptrA             = f+j_coord_offsetA;
1893             fjptrB             = f+j_coord_offsetB;
1894             fjptrC             = f+j_coord_offsetC;
1895             fjptrD             = f+j_coord_offsetD;
1896             fjptrE             = f+j_coord_offsetE;
1897             fjptrF             = f+j_coord_offsetF;
1898             fjptrG             = f+j_coord_offsetG;
1899             fjptrH             = f+j_coord_offsetH;
1900
1901             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1902                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1903                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1904
1905             /* Inner loop uses 534 flops */
1906         }
1907
1908         if(jidx<j_index_end)
1909         {
1910
1911             /* Get j neighbor index, and coordinate index */
1912             jnrlistA         = jjnr[jidx];
1913             jnrlistB         = jjnr[jidx+1];
1914             jnrlistC         = jjnr[jidx+2];
1915             jnrlistD         = jjnr[jidx+3];
1916             jnrlistE         = jjnr[jidx+4];
1917             jnrlistF         = jjnr[jidx+5];
1918             jnrlistG         = jjnr[jidx+6];
1919             jnrlistH         = jjnr[jidx+7];
1920             /* Sign of each element will be negative for non-real atoms.
1921              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1922              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1923              */
1924             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1925                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1926
1927             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1928             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1929             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1930             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1931             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1932             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1933             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1934             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1935             j_coord_offsetA  = DIM*jnrA;
1936             j_coord_offsetB  = DIM*jnrB;
1937             j_coord_offsetC  = DIM*jnrC;
1938             j_coord_offsetD  = DIM*jnrD;
1939             j_coord_offsetE  = DIM*jnrE;
1940             j_coord_offsetF  = DIM*jnrF;
1941             j_coord_offsetG  = DIM*jnrG;
1942             j_coord_offsetH  = DIM*jnrH;
1943
1944             /* load j atom coordinates */
1945             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1946                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1947                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1948                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1949                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1950                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1951
1952             /* Calculate displacement vector */
1953             dx00             = _mm256_sub_ps(ix0,jx0);
1954             dy00             = _mm256_sub_ps(iy0,jy0);
1955             dz00             = _mm256_sub_ps(iz0,jz0);
1956             dx11             = _mm256_sub_ps(ix1,jx1);
1957             dy11             = _mm256_sub_ps(iy1,jy1);
1958             dz11             = _mm256_sub_ps(iz1,jz1);
1959             dx12             = _mm256_sub_ps(ix1,jx2);
1960             dy12             = _mm256_sub_ps(iy1,jy2);
1961             dz12             = _mm256_sub_ps(iz1,jz2);
1962             dx13             = _mm256_sub_ps(ix1,jx3);
1963             dy13             = _mm256_sub_ps(iy1,jy3);
1964             dz13             = _mm256_sub_ps(iz1,jz3);
1965             dx21             = _mm256_sub_ps(ix2,jx1);
1966             dy21             = _mm256_sub_ps(iy2,jy1);
1967             dz21             = _mm256_sub_ps(iz2,jz1);
1968             dx22             = _mm256_sub_ps(ix2,jx2);
1969             dy22             = _mm256_sub_ps(iy2,jy2);
1970             dz22             = _mm256_sub_ps(iz2,jz2);
1971             dx23             = _mm256_sub_ps(ix2,jx3);
1972             dy23             = _mm256_sub_ps(iy2,jy3);
1973             dz23             = _mm256_sub_ps(iz2,jz3);
1974             dx31             = _mm256_sub_ps(ix3,jx1);
1975             dy31             = _mm256_sub_ps(iy3,jy1);
1976             dz31             = _mm256_sub_ps(iz3,jz1);
1977             dx32             = _mm256_sub_ps(ix3,jx2);
1978             dy32             = _mm256_sub_ps(iy3,jy2);
1979             dz32             = _mm256_sub_ps(iz3,jz2);
1980             dx33             = _mm256_sub_ps(ix3,jx3);
1981             dy33             = _mm256_sub_ps(iy3,jy3);
1982             dz33             = _mm256_sub_ps(iz3,jz3);
1983
1984             /* Calculate squared distance and things based on it */
1985             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1986             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1987             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1988             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1989             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1990             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1991             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1992             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1993             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1994             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1995
1996             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1997             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1998             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1999             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
2000             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
2001             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
2002             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
2003             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
2004             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
2005
2006             rinvsq00         = gmx_mm256_inv_ps(rsq00);
2007             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
2008             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
2009             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
2010             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
2011             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
2012             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
2013             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
2014             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
2015             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
2016
2017             fjx0             = _mm256_setzero_ps();
2018             fjy0             = _mm256_setzero_ps();
2019             fjz0             = _mm256_setzero_ps();
2020             fjx1             = _mm256_setzero_ps();
2021             fjy1             = _mm256_setzero_ps();
2022             fjz1             = _mm256_setzero_ps();
2023             fjx2             = _mm256_setzero_ps();
2024             fjy2             = _mm256_setzero_ps();
2025             fjz2             = _mm256_setzero_ps();
2026             fjx3             = _mm256_setzero_ps();
2027             fjy3             = _mm256_setzero_ps();
2028             fjz3             = _mm256_setzero_ps();
2029
2030             /**************************
2031              * CALCULATE INTERACTIONS *
2032              **************************/
2033
2034             /* LENNARD-JONES DISPERSION/REPULSION */
2035
2036             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2037             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
2038
2039             fscal            = fvdw;
2040
2041             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2042
2043             /* Calculate temporary vectorial force */
2044             tx               = _mm256_mul_ps(fscal,dx00);
2045             ty               = _mm256_mul_ps(fscal,dy00);
2046             tz               = _mm256_mul_ps(fscal,dz00);
2047
2048             /* Update vectorial force */
2049             fix0             = _mm256_add_ps(fix0,tx);
2050             fiy0             = _mm256_add_ps(fiy0,ty);
2051             fiz0             = _mm256_add_ps(fiz0,tz);
2052
2053             fjx0             = _mm256_add_ps(fjx0,tx);
2054             fjy0             = _mm256_add_ps(fjy0,ty);
2055             fjz0             = _mm256_add_ps(fjz0,tz);
2056
2057             /**************************
2058              * CALCULATE INTERACTIONS *
2059              **************************/
2060
2061             r11              = _mm256_mul_ps(rsq11,rinv11);
2062             r11              = _mm256_andnot_ps(dummy_mask,r11);
2063
2064             /* EWALD ELECTROSTATICS */
2065
2066             /* Analytical PME correction */
2067             zeta2            = _mm256_mul_ps(beta2,rsq11);
2068             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
2069             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2070             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2071             felec            = _mm256_mul_ps(qq11,felec);
2072
2073             fscal            = felec;
2074
2075             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2076
2077             /* Calculate temporary vectorial force */
2078             tx               = _mm256_mul_ps(fscal,dx11);
2079             ty               = _mm256_mul_ps(fscal,dy11);
2080             tz               = _mm256_mul_ps(fscal,dz11);
2081
2082             /* Update vectorial force */
2083             fix1             = _mm256_add_ps(fix1,tx);
2084             fiy1             = _mm256_add_ps(fiy1,ty);
2085             fiz1             = _mm256_add_ps(fiz1,tz);
2086
2087             fjx1             = _mm256_add_ps(fjx1,tx);
2088             fjy1             = _mm256_add_ps(fjy1,ty);
2089             fjz1             = _mm256_add_ps(fjz1,tz);
2090
2091             /**************************
2092              * CALCULATE INTERACTIONS *
2093              **************************/
2094
2095             r12              = _mm256_mul_ps(rsq12,rinv12);
2096             r12              = _mm256_andnot_ps(dummy_mask,r12);
2097
2098             /* EWALD ELECTROSTATICS */
2099
2100             /* Analytical PME correction */
2101             zeta2            = _mm256_mul_ps(beta2,rsq12);
2102             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
2103             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2104             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2105             felec            = _mm256_mul_ps(qq12,felec);
2106
2107             fscal            = felec;
2108
2109             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2110
2111             /* Calculate temporary vectorial force */
2112             tx               = _mm256_mul_ps(fscal,dx12);
2113             ty               = _mm256_mul_ps(fscal,dy12);
2114             tz               = _mm256_mul_ps(fscal,dz12);
2115
2116             /* Update vectorial force */
2117             fix1             = _mm256_add_ps(fix1,tx);
2118             fiy1             = _mm256_add_ps(fiy1,ty);
2119             fiz1             = _mm256_add_ps(fiz1,tz);
2120
2121             fjx2             = _mm256_add_ps(fjx2,tx);
2122             fjy2             = _mm256_add_ps(fjy2,ty);
2123             fjz2             = _mm256_add_ps(fjz2,tz);
2124
2125             /**************************
2126              * CALCULATE INTERACTIONS *
2127              **************************/
2128
2129             r13              = _mm256_mul_ps(rsq13,rinv13);
2130             r13              = _mm256_andnot_ps(dummy_mask,r13);
2131
2132             /* EWALD ELECTROSTATICS */
2133
2134             /* Analytical PME correction */
2135             zeta2            = _mm256_mul_ps(beta2,rsq13);
2136             rinv3            = _mm256_mul_ps(rinvsq13,rinv13);
2137             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2138             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2139             felec            = _mm256_mul_ps(qq13,felec);
2140
2141             fscal            = felec;
2142
2143             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2144
2145             /* Calculate temporary vectorial force */
2146             tx               = _mm256_mul_ps(fscal,dx13);
2147             ty               = _mm256_mul_ps(fscal,dy13);
2148             tz               = _mm256_mul_ps(fscal,dz13);
2149
2150             /* Update vectorial force */
2151             fix1             = _mm256_add_ps(fix1,tx);
2152             fiy1             = _mm256_add_ps(fiy1,ty);
2153             fiz1             = _mm256_add_ps(fiz1,tz);
2154
2155             fjx3             = _mm256_add_ps(fjx3,tx);
2156             fjy3             = _mm256_add_ps(fjy3,ty);
2157             fjz3             = _mm256_add_ps(fjz3,tz);
2158
2159             /**************************
2160              * CALCULATE INTERACTIONS *
2161              **************************/
2162
2163             r21              = _mm256_mul_ps(rsq21,rinv21);
2164             r21              = _mm256_andnot_ps(dummy_mask,r21);
2165
2166             /* EWALD ELECTROSTATICS */
2167
2168             /* Analytical PME correction */
2169             zeta2            = _mm256_mul_ps(beta2,rsq21);
2170             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
2171             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2172             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2173             felec            = _mm256_mul_ps(qq21,felec);
2174
2175             fscal            = felec;
2176
2177             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2178
2179             /* Calculate temporary vectorial force */
2180             tx               = _mm256_mul_ps(fscal,dx21);
2181             ty               = _mm256_mul_ps(fscal,dy21);
2182             tz               = _mm256_mul_ps(fscal,dz21);
2183
2184             /* Update vectorial force */
2185             fix2             = _mm256_add_ps(fix2,tx);
2186             fiy2             = _mm256_add_ps(fiy2,ty);
2187             fiz2             = _mm256_add_ps(fiz2,tz);
2188
2189             fjx1             = _mm256_add_ps(fjx1,tx);
2190             fjy1             = _mm256_add_ps(fjy1,ty);
2191             fjz1             = _mm256_add_ps(fjz1,tz);
2192
2193             /**************************
2194              * CALCULATE INTERACTIONS *
2195              **************************/
2196
2197             r22              = _mm256_mul_ps(rsq22,rinv22);
2198             r22              = _mm256_andnot_ps(dummy_mask,r22);
2199
2200             /* EWALD ELECTROSTATICS */
2201
2202             /* Analytical PME correction */
2203             zeta2            = _mm256_mul_ps(beta2,rsq22);
2204             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
2205             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2206             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2207             felec            = _mm256_mul_ps(qq22,felec);
2208
2209             fscal            = felec;
2210
2211             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2212
2213             /* Calculate temporary vectorial force */
2214             tx               = _mm256_mul_ps(fscal,dx22);
2215             ty               = _mm256_mul_ps(fscal,dy22);
2216             tz               = _mm256_mul_ps(fscal,dz22);
2217
2218             /* Update vectorial force */
2219             fix2             = _mm256_add_ps(fix2,tx);
2220             fiy2             = _mm256_add_ps(fiy2,ty);
2221             fiz2             = _mm256_add_ps(fiz2,tz);
2222
2223             fjx2             = _mm256_add_ps(fjx2,tx);
2224             fjy2             = _mm256_add_ps(fjy2,ty);
2225             fjz2             = _mm256_add_ps(fjz2,tz);
2226
2227             /**************************
2228              * CALCULATE INTERACTIONS *
2229              **************************/
2230
2231             r23              = _mm256_mul_ps(rsq23,rinv23);
2232             r23              = _mm256_andnot_ps(dummy_mask,r23);
2233
2234             /* EWALD ELECTROSTATICS */
2235
2236             /* Analytical PME correction */
2237             zeta2            = _mm256_mul_ps(beta2,rsq23);
2238             rinv3            = _mm256_mul_ps(rinvsq23,rinv23);
2239             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2240             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2241             felec            = _mm256_mul_ps(qq23,felec);
2242
2243             fscal            = felec;
2244
2245             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2246
2247             /* Calculate temporary vectorial force */
2248             tx               = _mm256_mul_ps(fscal,dx23);
2249             ty               = _mm256_mul_ps(fscal,dy23);
2250             tz               = _mm256_mul_ps(fscal,dz23);
2251
2252             /* Update vectorial force */
2253             fix2             = _mm256_add_ps(fix2,tx);
2254             fiy2             = _mm256_add_ps(fiy2,ty);
2255             fiz2             = _mm256_add_ps(fiz2,tz);
2256
2257             fjx3             = _mm256_add_ps(fjx3,tx);
2258             fjy3             = _mm256_add_ps(fjy3,ty);
2259             fjz3             = _mm256_add_ps(fjz3,tz);
2260
2261             /**************************
2262              * CALCULATE INTERACTIONS *
2263              **************************/
2264
2265             r31              = _mm256_mul_ps(rsq31,rinv31);
2266             r31              = _mm256_andnot_ps(dummy_mask,r31);
2267
2268             /* EWALD ELECTROSTATICS */
2269
2270             /* Analytical PME correction */
2271             zeta2            = _mm256_mul_ps(beta2,rsq31);
2272             rinv3            = _mm256_mul_ps(rinvsq31,rinv31);
2273             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2274             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2275             felec            = _mm256_mul_ps(qq31,felec);
2276
2277             fscal            = felec;
2278
2279             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2280
2281             /* Calculate temporary vectorial force */
2282             tx               = _mm256_mul_ps(fscal,dx31);
2283             ty               = _mm256_mul_ps(fscal,dy31);
2284             tz               = _mm256_mul_ps(fscal,dz31);
2285
2286             /* Update vectorial force */
2287             fix3             = _mm256_add_ps(fix3,tx);
2288             fiy3             = _mm256_add_ps(fiy3,ty);
2289             fiz3             = _mm256_add_ps(fiz3,tz);
2290
2291             fjx1             = _mm256_add_ps(fjx1,tx);
2292             fjy1             = _mm256_add_ps(fjy1,ty);
2293             fjz1             = _mm256_add_ps(fjz1,tz);
2294
2295             /**************************
2296              * CALCULATE INTERACTIONS *
2297              **************************/
2298
2299             r32              = _mm256_mul_ps(rsq32,rinv32);
2300             r32              = _mm256_andnot_ps(dummy_mask,r32);
2301
2302             /* EWALD ELECTROSTATICS */
2303
2304             /* Analytical PME correction */
2305             zeta2            = _mm256_mul_ps(beta2,rsq32);
2306             rinv3            = _mm256_mul_ps(rinvsq32,rinv32);
2307             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2308             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2309             felec            = _mm256_mul_ps(qq32,felec);
2310
2311             fscal            = felec;
2312
2313             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2314
2315             /* Calculate temporary vectorial force */
2316             tx               = _mm256_mul_ps(fscal,dx32);
2317             ty               = _mm256_mul_ps(fscal,dy32);
2318             tz               = _mm256_mul_ps(fscal,dz32);
2319
2320             /* Update vectorial force */
2321             fix3             = _mm256_add_ps(fix3,tx);
2322             fiy3             = _mm256_add_ps(fiy3,ty);
2323             fiz3             = _mm256_add_ps(fiz3,tz);
2324
2325             fjx2             = _mm256_add_ps(fjx2,tx);
2326             fjy2             = _mm256_add_ps(fjy2,ty);
2327             fjz2             = _mm256_add_ps(fjz2,tz);
2328
2329             /**************************
2330              * CALCULATE INTERACTIONS *
2331              **************************/
2332
2333             r33              = _mm256_mul_ps(rsq33,rinv33);
2334             r33              = _mm256_andnot_ps(dummy_mask,r33);
2335
2336             /* EWALD ELECTROSTATICS */
2337
2338             /* Analytical PME correction */
2339             zeta2            = _mm256_mul_ps(beta2,rsq33);
2340             rinv3            = _mm256_mul_ps(rinvsq33,rinv33);
2341             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2342             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2343             felec            = _mm256_mul_ps(qq33,felec);
2344
2345             fscal            = felec;
2346
2347             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2348
2349             /* Calculate temporary vectorial force */
2350             tx               = _mm256_mul_ps(fscal,dx33);
2351             ty               = _mm256_mul_ps(fscal,dy33);
2352             tz               = _mm256_mul_ps(fscal,dz33);
2353
2354             /* Update vectorial force */
2355             fix3             = _mm256_add_ps(fix3,tx);
2356             fiy3             = _mm256_add_ps(fiy3,ty);
2357             fiz3             = _mm256_add_ps(fiz3,tz);
2358
2359             fjx3             = _mm256_add_ps(fjx3,tx);
2360             fjy3             = _mm256_add_ps(fjy3,ty);
2361             fjz3             = _mm256_add_ps(fjz3,tz);
2362
2363             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2364             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2365             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2366             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2367             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2368             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2369             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2370             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2371
2372             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2373                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2374                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2375
2376             /* Inner loop uses 543 flops */
2377         }
2378
2379         /* End of innermost loop */
2380
2381         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2382                                                  f+i_coord_offset,fshift+i_shift_offset);
2383
2384         /* Increment number of inner iterations */
2385         inneriter                  += j_index_end - j_index_start;
2386
2387         /* Outer loop uses 24 flops */
2388     }
2389
2390     /* Increment number of outer iterations */
2391     outeriter        += nri;
2392
2393     /* Update outer/inner flops */
2394
2395     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*543);
2396 }