src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_avx_256_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_VF_avx_256_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LJEwald
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_VF_avx_256_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  75     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  83     real *           vdwioffsetptr0;
  84     real *           vdwgridioffsetptr0;
  85     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  86     real *           vdwioffsetptr1;
  87     real *           vdwgridioffsetptr1;
  88     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     real *           vdwioffsetptr2;
  90     real *           vdwgridioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 100     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 101     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 102     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 107     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 108     real             *charge;
 109     int              nvdwtype;
 110     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 111     int              *vdwtype;
 112     real             *vdwparam;
 113     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 114     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 115     __m256d           c6grid_00;
 116     __m256d           c6grid_01;
 117     __m256d           c6grid_02;
 118     __m256d           c6grid_10;
 119     __m256d           c6grid_11;
 120     __m256d           c6grid_12;
 121     __m256d           c6grid_20;
 122     __m256d           c6grid_21;
 123     __m256d           c6grid_22;
 124     real             *vdwgridparam;
 125     __m256d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 126     __m256d           one_half  = _mm256_set1_pd(0.5);
 127     __m256d           minus_one = _mm256_set1_pd(-1.0);
 128     __m128i          ewitab;
 129     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 130     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 131     real             *ewtab;
 132     __m256d          dummy_mask,cutoff_mask;
 133     __m128           tmpmask0,tmpmask1;
 134     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 135     __m256d          one     = _mm256_set1_pd(1.0);
 136     __m256d          two     = _mm256_set1_pd(2.0);
 137     x                = xx[0];
 138     f                = ff[0];
 139
 140     nri              = nlist->nri;
 141     iinr             = nlist->iinr;
 142     jindex           = nlist->jindex;
 143     jjnr             = nlist->jjnr;
 144     shiftidx         = nlist->shift;
 145     gid              = nlist->gid;
 146     shiftvec         = fr->shift_vec[0];
 147     fshift           = fr->fshift[0];
 148     facel            = _mm256_set1_pd(fr->ic->epsfac);
 149     charge           = mdatoms->chargeA;
 150     nvdwtype         = fr->ntype;
 151     vdwparam         = fr->nbfp;
 152     vdwtype          = mdatoms->typeA;
 153     vdwgridparam     = fr->ljpme_c6grid;
 154     sh_lj_ewald      = _mm256_set1_pd(fr->ic->sh_lj_ewald);
 155     ewclj            = _mm256_set1_pd(fr->ic->ewaldcoeff_lj);
 156     ewclj2           = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
 157
 158     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 159     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 160     beta2            = _mm256_mul_pd(beta,beta);
 161     beta3            = _mm256_mul_pd(beta,beta2);
 162
 163     ewtab            = fr->ic->tabq_coul_FDV0;
 164     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 165     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 166
 167     /* Setup water-specific parameters */
 168     inr              = nlist->iinr[0];
 169     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 170     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 171     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 172     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 173     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
 174
 175     jq0              = _mm256_set1_pd(charge[inr+0]);
 176     jq1              = _mm256_set1_pd(charge[inr+1]);
 177     jq2              = _mm256_set1_pd(charge[inr+2]);
 178     vdwjidx0A        = 2*vdwtype[inr+0];
 179     qq00             = _mm256_mul_pd(iq0,jq0);
 180     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 181     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 182     c6grid_00        = _mm256_set1_pd(vdwgridioffsetptr0[vdwjidx0A]);
 183     qq01             = _mm256_mul_pd(iq0,jq1);
 184     qq02             = _mm256_mul_pd(iq0,jq2);
 185     qq10             = _mm256_mul_pd(iq1,jq0);
 186     qq11             = _mm256_mul_pd(iq1,jq1);
 187     qq12             = _mm256_mul_pd(iq1,jq2);
 188     qq20             = _mm256_mul_pd(iq2,jq0);
 189     qq21             = _mm256_mul_pd(iq2,jq1);
 190     qq22             = _mm256_mul_pd(iq2,jq2);
 191
 192     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 193     rcutoff_scalar   = fr->ic->rcoulomb;
 194     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 195     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 196
 197     sh_vdw_invrcut6  = _mm256_set1_pd(fr->ic->sh_invrc6);
 198     rvdw             = _mm256_set1_pd(fr->ic->rvdw);
 199
 200     /* Avoid stupid compiler warnings */
 201     jnrA = jnrB = jnrC = jnrD = 0;
 202     j_coord_offsetA = 0;
 203     j_coord_offsetB = 0;
 204     j_coord_offsetC = 0;
 205     j_coord_offsetD = 0;
 206
 207     outeriter        = 0;
 208     inneriter        = 0;
 209
 210     for(iidx=0;iidx<4*DIM;iidx++)
 211     {
 212         scratch[iidx] = 0.0;
 213     }
 214
 215     /* Start outer loop over neighborlists */
 216     for(iidx=0; iidx<nri; iidx++)
 217     {
 218         /* Load shift vector for this list */
 219         i_shift_offset   = DIM*shiftidx[iidx];
 220
 221         /* Load limits for loop over neighbors */
 222         j_index_start    = jindex[iidx];
 223         j_index_end      = jindex[iidx+1];
 224
 225         /* Get outer coordinate index */
 226         inr              = iinr[iidx];
 227         i_coord_offset   = DIM*inr;
 228
 229         /* Load i particle coords and add shift vector */
 230         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 231                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 232
 233         fix0             = _mm256_setzero_pd();
 234         fiy0             = _mm256_setzero_pd();
 235         fiz0             = _mm256_setzero_pd();
 236         fix1             = _mm256_setzero_pd();
 237         fiy1             = _mm256_setzero_pd();
 238         fiz1             = _mm256_setzero_pd();
 239         fix2             = _mm256_setzero_pd();
 240         fiy2             = _mm256_setzero_pd();
 241         fiz2             = _mm256_setzero_pd();
 242
 243         /* Reset potential sums */
 244         velecsum         = _mm256_setzero_pd();
 245         vvdwsum          = _mm256_setzero_pd();
 246
 247         /* Start inner kernel loop */
 248         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 249         {
 250
 251             /* Get j neighbor index, and coordinate index */
 252             jnrA             = jjnr[jidx];
 253             jnrB             = jjnr[jidx+1];
 254             jnrC             = jjnr[jidx+2];
 255             jnrD             = jjnr[jidx+3];
 256             j_coord_offsetA  = DIM*jnrA;
 257             j_coord_offsetB  = DIM*jnrB;
 258             j_coord_offsetC  = DIM*jnrC;
 259             j_coord_offsetD  = DIM*jnrD;
 260
 261             /* load j atom coordinates */
 262             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 263                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 264                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 265
 266             /* Calculate displacement vector */
 267             dx00             = _mm256_sub_pd(ix0,jx0);
 268             dy00             = _mm256_sub_pd(iy0,jy0);
 269             dz00             = _mm256_sub_pd(iz0,jz0);
 270             dx01             = _mm256_sub_pd(ix0,jx1);
 271             dy01             = _mm256_sub_pd(iy0,jy1);
 272             dz01             = _mm256_sub_pd(iz0,jz1);
 273             dx02             = _mm256_sub_pd(ix0,jx2);
 274             dy02             = _mm256_sub_pd(iy0,jy2);
 275             dz02             = _mm256_sub_pd(iz0,jz2);
 276             dx10             = _mm256_sub_pd(ix1,jx0);
 277             dy10             = _mm256_sub_pd(iy1,jy0);
 278             dz10             = _mm256_sub_pd(iz1,jz0);
 279             dx11             = _mm256_sub_pd(ix1,jx1);
 280             dy11             = _mm256_sub_pd(iy1,jy1);
 281             dz11             = _mm256_sub_pd(iz1,jz1);
 282             dx12             = _mm256_sub_pd(ix1,jx2);
 283             dy12             = _mm256_sub_pd(iy1,jy2);
 284             dz12             = _mm256_sub_pd(iz1,jz2);
 285             dx20             = _mm256_sub_pd(ix2,jx0);
 286             dy20             = _mm256_sub_pd(iy2,jy0);
 287             dz20             = _mm256_sub_pd(iz2,jz0);
 288             dx21             = _mm256_sub_pd(ix2,jx1);
 289             dy21             = _mm256_sub_pd(iy2,jy1);
 290             dz21             = _mm256_sub_pd(iz2,jz1);
 291             dx22             = _mm256_sub_pd(ix2,jx2);
 292             dy22             = _mm256_sub_pd(iy2,jy2);
 293             dz22             = _mm256_sub_pd(iz2,jz2);
 294
 295             /* Calculate squared distance and things based on it */
 296             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 297             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 298             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 299             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 300             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 301             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 302             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 303             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 304             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 305
 306             rinv00           = avx256_invsqrt_d(rsq00);
 307             rinv01           = avx256_invsqrt_d(rsq01);
 308             rinv02           = avx256_invsqrt_d(rsq02);
 309             rinv10           = avx256_invsqrt_d(rsq10);
 310             rinv11           = avx256_invsqrt_d(rsq11);
 311             rinv12           = avx256_invsqrt_d(rsq12);
 312             rinv20           = avx256_invsqrt_d(rsq20);
 313             rinv21           = avx256_invsqrt_d(rsq21);
 314             rinv22           = avx256_invsqrt_d(rsq22);
 315
 316             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 317             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 318             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 319             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 320             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 321             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 322             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 323             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 324             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 325
 326             fjx0             = _mm256_setzero_pd();
 327             fjy0             = _mm256_setzero_pd();
 328             fjz0             = _mm256_setzero_pd();
 329             fjx1             = _mm256_setzero_pd();
 330             fjy1             = _mm256_setzero_pd();
 331             fjz1             = _mm256_setzero_pd();
 332             fjx2             = _mm256_setzero_pd();
 333             fjy2             = _mm256_setzero_pd();
 334             fjz2             = _mm256_setzero_pd();
 335
 336             /**************************
 337              * CALCULATE INTERACTIONS *
 338              **************************/
 339
 340             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 341             {
 342
 343             r00              = _mm256_mul_pd(rsq00,rinv00);
 344
 345             /* EWALD ELECTROSTATICS */
 346
 347             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 348             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 349             ewitab           = _mm256_cvttpd_epi32(ewrt);
 350             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 351             ewitab           = _mm_slli_epi32(ewitab,2);
 352             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 353             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 354             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 355             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 356             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 357             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 358             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 359             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 360             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 361
 362             /* Analytical LJ-PME */
 363             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 364             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 365             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 366             exponent         = avx256_exp_d(ewcljrsq);
 367             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 368             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 369             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 370             vvdw6            = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
 371             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 372             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 373                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_00,sh_vdw_invrcut6),_mm256_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
 374             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 375             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
 376
 377             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             velec            = _mm256_and_pd(velec,cutoff_mask);
 381             velecsum         = _mm256_add_pd(velecsum,velec);
 382             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 383             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 384
 385             fscal            = _mm256_add_pd(felec,fvdw);
 386
 387             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 388
 389             /* Calculate temporary vectorial force */
 390             tx               = _mm256_mul_pd(fscal,dx00);
 391             ty               = _mm256_mul_pd(fscal,dy00);
 392             tz               = _mm256_mul_pd(fscal,dz00);
 393
 394             /* Update vectorial force */
 395             fix0             = _mm256_add_pd(fix0,tx);
 396             fiy0             = _mm256_add_pd(fiy0,ty);
 397             fiz0             = _mm256_add_pd(fiz0,tz);
 398
 399             fjx0             = _mm256_add_pd(fjx0,tx);
 400             fjy0             = _mm256_add_pd(fjy0,ty);
 401             fjz0             = _mm256_add_pd(fjz0,tz);
 402
 403             }
 404
 405             /**************************
 406              * CALCULATE INTERACTIONS *
 407              **************************/
 408
 409             if (gmx_mm256_any_lt(rsq01,rcutoff2))
 410             {
 411
 412             r01              = _mm256_mul_pd(rsq01,rinv01);
 413
 414             /* EWALD ELECTROSTATICS */
 415
 416             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 417             ewrt             = _mm256_mul_pd(r01,ewtabscale);
 418             ewitab           = _mm256_cvttpd_epi32(ewrt);
 419             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 420             ewitab           = _mm_slli_epi32(ewitab,2);
 421             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 422             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 423             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 424             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 425             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 426             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 427             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 428             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_sub_pd(rinv01,sh_ewald),velec));
 429             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
 430
 431             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
 432
 433             /* Update potential sum for this i atom from the interaction with this j atom. */
 434             velec            = _mm256_and_pd(velec,cutoff_mask);
 435             velecsum         = _mm256_add_pd(velecsum,velec);
 436
 437             fscal            = felec;
 438
 439             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 440
 441             /* Calculate temporary vectorial force */
 442             tx               = _mm256_mul_pd(fscal,dx01);
 443             ty               = _mm256_mul_pd(fscal,dy01);
 444             tz               = _mm256_mul_pd(fscal,dz01);
 445
 446             /* Update vectorial force */
 447             fix0             = _mm256_add_pd(fix0,tx);
 448             fiy0             = _mm256_add_pd(fiy0,ty);
 449             fiz0             = _mm256_add_pd(fiz0,tz);
 450
 451             fjx1             = _mm256_add_pd(fjx1,tx);
 452             fjy1             = _mm256_add_pd(fjy1,ty);
 453             fjz1             = _mm256_add_pd(fjz1,tz);
 454
 455             }
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             if (gmx_mm256_any_lt(rsq02,rcutoff2))
 462             {
 463
 464             r02              = _mm256_mul_pd(rsq02,rinv02);
 465
 466             /* EWALD ELECTROSTATICS */
 467
 468             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 469             ewrt             = _mm256_mul_pd(r02,ewtabscale);
 470             ewitab           = _mm256_cvttpd_epi32(ewrt);
 471             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 472             ewitab           = _mm_slli_epi32(ewitab,2);
 473             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 474             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 475             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 476             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 477             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 478             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 479             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 480             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_sub_pd(rinv02,sh_ewald),velec));
 481             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
 482
 483             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
 484
 485             /* Update potential sum for this i atom from the interaction with this j atom. */
 486             velec            = _mm256_and_pd(velec,cutoff_mask);
 487             velecsum         = _mm256_add_pd(velecsum,velec);
 488
 489             fscal            = felec;
 490
 491             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 492
 493             /* Calculate temporary vectorial force */
 494             tx               = _mm256_mul_pd(fscal,dx02);
 495             ty               = _mm256_mul_pd(fscal,dy02);
 496             tz               = _mm256_mul_pd(fscal,dz02);
 497
 498             /* Update vectorial force */
 499             fix0             = _mm256_add_pd(fix0,tx);
 500             fiy0             = _mm256_add_pd(fiy0,ty);
 501             fiz0             = _mm256_add_pd(fiz0,tz);
 502
 503             fjx2             = _mm256_add_pd(fjx2,tx);
 504             fjy2             = _mm256_add_pd(fjy2,ty);
 505             fjz2             = _mm256_add_pd(fjz2,tz);
 506
 507             }
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 514             {
 515
 516             r10              = _mm256_mul_pd(rsq10,rinv10);
 517
 518             /* EWALD ELECTROSTATICS */
 519
 520             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 521             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 522             ewitab           = _mm256_cvttpd_epi32(ewrt);
 523             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 524             ewitab           = _mm_slli_epi32(ewitab,2);
 525             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 526             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 527             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 528             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 529             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 530             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 531             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 532             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_sub_pd(rinv10,sh_ewald),velec));
 533             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 534
 535             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 536
 537             /* Update potential sum for this i atom from the interaction with this j atom. */
 538             velec            = _mm256_and_pd(velec,cutoff_mask);
 539             velecsum         = _mm256_add_pd(velecsum,velec);
 540
 541             fscal            = felec;
 542
 543             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 544
 545             /* Calculate temporary vectorial force */
 546             tx               = _mm256_mul_pd(fscal,dx10);
 547             ty               = _mm256_mul_pd(fscal,dy10);
 548             tz               = _mm256_mul_pd(fscal,dz10);
 549
 550             /* Update vectorial force */
 551             fix1             = _mm256_add_pd(fix1,tx);
 552             fiy1             = _mm256_add_pd(fiy1,ty);
 553             fiz1             = _mm256_add_pd(fiz1,tz);
 554
 555             fjx0             = _mm256_add_pd(fjx0,tx);
 556             fjy0             = _mm256_add_pd(fjy0,ty);
 557             fjz0             = _mm256_add_pd(fjz0,tz);
 558
 559             }
 560
 561             /**************************
 562              * CALCULATE INTERACTIONS *
 563              **************************/
 564
 565             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 566             {
 567
 568             r11              = _mm256_mul_pd(rsq11,rinv11);
 569
 570             /* EWALD ELECTROSTATICS */
 571
 572             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 573             ewrt             = _mm256_mul_pd(r11,ewtabscale);
 574             ewitab           = _mm256_cvttpd_epi32(ewrt);
 575             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 576             ewitab           = _mm_slli_epi32(ewitab,2);
 577             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 578             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 579             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 580             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 581             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 582             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 583             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 584             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
 585             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
 586
 587             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
 588
 589             /* Update potential sum for this i atom from the interaction with this j atom. */
 590             velec            = _mm256_and_pd(velec,cutoff_mask);
 591             velecsum         = _mm256_add_pd(velecsum,velec);
 592
 593             fscal            = felec;
 594
 595             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 596
 597             /* Calculate temporary vectorial force */
 598             tx               = _mm256_mul_pd(fscal,dx11);
 599             ty               = _mm256_mul_pd(fscal,dy11);
 600             tz               = _mm256_mul_pd(fscal,dz11);
 601
 602             /* Update vectorial force */
 603             fix1             = _mm256_add_pd(fix1,tx);
 604             fiy1             = _mm256_add_pd(fiy1,ty);
 605             fiz1             = _mm256_add_pd(fiz1,tz);
 606
 607             fjx1             = _mm256_add_pd(fjx1,tx);
 608             fjy1             = _mm256_add_pd(fjy1,ty);
 609             fjz1             = _mm256_add_pd(fjz1,tz);
 610
 611             }
 612
 613             /**************************
 614              * CALCULATE INTERACTIONS *
 615              **************************/
 616
 617             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 618             {
 619
 620             r12              = _mm256_mul_pd(rsq12,rinv12);
 621
 622             /* EWALD ELECTROSTATICS */
 623
 624             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 625             ewrt             = _mm256_mul_pd(r12,ewtabscale);
 626             ewitab           = _mm256_cvttpd_epi32(ewrt);
 627             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 628             ewitab           = _mm_slli_epi32(ewitab,2);
 629             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 630             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 631             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 632             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 633             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 634             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 635             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 636             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
 637             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
 638
 639             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velec            = _mm256_and_pd(velec,cutoff_mask);
 643             velecsum         = _mm256_add_pd(velecsum,velec);
 644
 645             fscal            = felec;
 646
 647             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 648
 649             /* Calculate temporary vectorial force */
 650             tx               = _mm256_mul_pd(fscal,dx12);
 651             ty               = _mm256_mul_pd(fscal,dy12);
 652             tz               = _mm256_mul_pd(fscal,dz12);
 653
 654             /* Update vectorial force */
 655             fix1             = _mm256_add_pd(fix1,tx);
 656             fiy1             = _mm256_add_pd(fiy1,ty);
 657             fiz1             = _mm256_add_pd(fiz1,tz);
 658
 659             fjx2             = _mm256_add_pd(fjx2,tx);
 660             fjy2             = _mm256_add_pd(fjy2,ty);
 661             fjz2             = _mm256_add_pd(fjz2,tz);
 662
 663             }
 664
 665             /**************************
 666              * CALCULATE INTERACTIONS *
 667              **************************/
 668
 669             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 670             {
 671
 672             r20              = _mm256_mul_pd(rsq20,rinv20);
 673
 674             /* EWALD ELECTROSTATICS */
 675
 676             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 677             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 678             ewitab           = _mm256_cvttpd_epi32(ewrt);
 679             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 680             ewitab           = _mm_slli_epi32(ewitab,2);
 681             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 682             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 683             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 684             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 685             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 686             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 687             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 688             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_sub_pd(rinv20,sh_ewald),velec));
 689             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 690
 691             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 692
 693             /* Update potential sum for this i atom from the interaction with this j atom. */
 694             velec            = _mm256_and_pd(velec,cutoff_mask);
 695             velecsum         = _mm256_add_pd(velecsum,velec);
 696
 697             fscal            = felec;
 698
 699             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 700
 701             /* Calculate temporary vectorial force */
 702             tx               = _mm256_mul_pd(fscal,dx20);
 703             ty               = _mm256_mul_pd(fscal,dy20);
 704             tz               = _mm256_mul_pd(fscal,dz20);
 705
 706             /* Update vectorial force */
 707             fix2             = _mm256_add_pd(fix2,tx);
 708             fiy2             = _mm256_add_pd(fiy2,ty);
 709             fiz2             = _mm256_add_pd(fiz2,tz);
 710
 711             fjx0             = _mm256_add_pd(fjx0,tx);
 712             fjy0             = _mm256_add_pd(fjy0,ty);
 713             fjz0             = _mm256_add_pd(fjz0,tz);
 714
 715             }
 716
 717             /**************************
 718              * CALCULATE INTERACTIONS *
 719              **************************/
 720
 721             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 722             {
 723
 724             r21              = _mm256_mul_pd(rsq21,rinv21);
 725
 726             /* EWALD ELECTROSTATICS */
 727
 728             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 729             ewrt             = _mm256_mul_pd(r21,ewtabscale);
 730             ewitab           = _mm256_cvttpd_epi32(ewrt);
 731             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 732             ewitab           = _mm_slli_epi32(ewitab,2);
 733             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 734             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 735             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 736             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 737             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 738             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 739             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 740             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
 741             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
 742
 743             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
 744
 745             /* Update potential sum for this i atom from the interaction with this j atom. */
 746             velec            = _mm256_and_pd(velec,cutoff_mask);
 747             velecsum         = _mm256_add_pd(velecsum,velec);
 748
 749             fscal            = felec;
 750
 751             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 752
 753             /* Calculate temporary vectorial force */
 754             tx               = _mm256_mul_pd(fscal,dx21);
 755             ty               = _mm256_mul_pd(fscal,dy21);
 756             tz               = _mm256_mul_pd(fscal,dz21);
 757
 758             /* Update vectorial force */
 759             fix2             = _mm256_add_pd(fix2,tx);
 760             fiy2             = _mm256_add_pd(fiy2,ty);
 761             fiz2             = _mm256_add_pd(fiz2,tz);
 762
 763             fjx1             = _mm256_add_pd(fjx1,tx);
 764             fjy1             = _mm256_add_pd(fjy1,ty);
 765             fjz1             = _mm256_add_pd(fjz1,tz);
 766
 767             }
 768
 769             /**************************
 770              * CALCULATE INTERACTIONS *
 771              **************************/
 772
 773             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 774             {
 775
 776             r22              = _mm256_mul_pd(rsq22,rinv22);
 777
 778             /* EWALD ELECTROSTATICS */
 779
 780             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 781             ewrt             = _mm256_mul_pd(r22,ewtabscale);
 782             ewitab           = _mm256_cvttpd_epi32(ewrt);
 783             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 784             ewitab           = _mm_slli_epi32(ewitab,2);
 785             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 786             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 787             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 788             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 789             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 790             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 791             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 792             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
 793             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
 794
 795             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
 796
 797             /* Update potential sum for this i atom from the interaction with this j atom. */
 798             velec            = _mm256_and_pd(velec,cutoff_mask);
 799             velecsum         = _mm256_add_pd(velecsum,velec);
 800
 801             fscal            = felec;
 802
 803             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 804
 805             /* Calculate temporary vectorial force */
 806             tx               = _mm256_mul_pd(fscal,dx22);
 807             ty               = _mm256_mul_pd(fscal,dy22);
 808             tz               = _mm256_mul_pd(fscal,dz22);
 809
 810             /* Update vectorial force */
 811             fix2             = _mm256_add_pd(fix2,tx);
 812             fiy2             = _mm256_add_pd(fiy2,ty);
 813             fiz2             = _mm256_add_pd(fiz2,tz);
 814
 815             fjx2             = _mm256_add_pd(fjx2,tx);
 816             fjy2             = _mm256_add_pd(fjy2,ty);
 817             fjz2             = _mm256_add_pd(fjz2,tz);
 818
 819             }
 820
 821             fjptrA             = f+j_coord_offsetA;
 822             fjptrB             = f+j_coord_offsetB;
 823             fjptrC             = f+j_coord_offsetC;
 824             fjptrD             = f+j_coord_offsetD;
 825
 826             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 827                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 828
 829             /* Inner loop uses 450 flops */
 830         }
 831
 832         if(jidx<j_index_end)
 833         {
 834
 835             /* Get j neighbor index, and coordinate index */
 836             jnrlistA         = jjnr[jidx];
 837             jnrlistB         = jjnr[jidx+1];
 838             jnrlistC         = jjnr[jidx+2];
 839             jnrlistD         = jjnr[jidx+3];
 840             /* Sign of each element will be negative for non-real atoms.
 841              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 842              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 843              */
 844             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 845
 846             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 847             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 848             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 849
 850             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 851             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 852             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 853             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 854             j_coord_offsetA  = DIM*jnrA;
 855             j_coord_offsetB  = DIM*jnrB;
 856             j_coord_offsetC  = DIM*jnrC;
 857             j_coord_offsetD  = DIM*jnrD;
 858
 859             /* load j atom coordinates */
 860             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 861                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 862                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 863
 864             /* Calculate displacement vector */
 865             dx00             = _mm256_sub_pd(ix0,jx0);
 866             dy00             = _mm256_sub_pd(iy0,jy0);
 867             dz00             = _mm256_sub_pd(iz0,jz0);
 868             dx01             = _mm256_sub_pd(ix0,jx1);
 869             dy01             = _mm256_sub_pd(iy0,jy1);
 870             dz01             = _mm256_sub_pd(iz0,jz1);
 871             dx02             = _mm256_sub_pd(ix0,jx2);
 872             dy02             = _mm256_sub_pd(iy0,jy2);
 873             dz02             = _mm256_sub_pd(iz0,jz2);
 874             dx10             = _mm256_sub_pd(ix1,jx0);
 875             dy10             = _mm256_sub_pd(iy1,jy0);
 876             dz10             = _mm256_sub_pd(iz1,jz0);
 877             dx11             = _mm256_sub_pd(ix1,jx1);
 878             dy11             = _mm256_sub_pd(iy1,jy1);
 879             dz11             = _mm256_sub_pd(iz1,jz1);
 880             dx12             = _mm256_sub_pd(ix1,jx2);
 881             dy12             = _mm256_sub_pd(iy1,jy2);
 882             dz12             = _mm256_sub_pd(iz1,jz2);
 883             dx20             = _mm256_sub_pd(ix2,jx0);
 884             dy20             = _mm256_sub_pd(iy2,jy0);
 885             dz20             = _mm256_sub_pd(iz2,jz0);
 886             dx21             = _mm256_sub_pd(ix2,jx1);
 887             dy21             = _mm256_sub_pd(iy2,jy1);
 888             dz21             = _mm256_sub_pd(iz2,jz1);
 889             dx22             = _mm256_sub_pd(ix2,jx2);
 890             dy22             = _mm256_sub_pd(iy2,jy2);
 891             dz22             = _mm256_sub_pd(iz2,jz2);
 892
 893             /* Calculate squared distance and things based on it */
 894             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 895             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 896             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 897             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 898             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 899             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 900             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 901             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 902             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 903
 904             rinv00           = avx256_invsqrt_d(rsq00);
 905             rinv01           = avx256_invsqrt_d(rsq01);
 906             rinv02           = avx256_invsqrt_d(rsq02);
 907             rinv10           = avx256_invsqrt_d(rsq10);
 908             rinv11           = avx256_invsqrt_d(rsq11);
 909             rinv12           = avx256_invsqrt_d(rsq12);
 910             rinv20           = avx256_invsqrt_d(rsq20);
 911             rinv21           = avx256_invsqrt_d(rsq21);
 912             rinv22           = avx256_invsqrt_d(rsq22);
 913
 914             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 915             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 916             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 917             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 918             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 919             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 920             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 921             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 922             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 923
 924             fjx0             = _mm256_setzero_pd();
 925             fjy0             = _mm256_setzero_pd();
 926             fjz0             = _mm256_setzero_pd();
 927             fjx1             = _mm256_setzero_pd();
 928             fjy1             = _mm256_setzero_pd();
 929             fjz1             = _mm256_setzero_pd();
 930             fjx2             = _mm256_setzero_pd();
 931             fjy2             = _mm256_setzero_pd();
 932             fjz2             = _mm256_setzero_pd();
 933
 934             /**************************
 935              * CALCULATE INTERACTIONS *
 936              **************************/
 937
 938             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 939             {
 940
 941             r00              = _mm256_mul_pd(rsq00,rinv00);
 942             r00              = _mm256_andnot_pd(dummy_mask,r00);
 943
 944             /* EWALD ELECTROSTATICS */
 945
 946             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 947             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 948             ewitab           = _mm256_cvttpd_epi32(ewrt);
 949             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 950             ewitab           = _mm_slli_epi32(ewitab,2);
 951             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 952             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 953             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 954             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 955             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 956             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 957             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 958             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 959             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 960
 961             /* Analytical LJ-PME */
 962             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 963             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 964             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 965             exponent         = avx256_exp_d(ewcljrsq);
 966             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 967             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 968             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 969             vvdw6            = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
 970             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 971             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 972                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_00,sh_vdw_invrcut6),_mm256_mul_pd(c6grid_00,sh_lj_ewald))),one_sixth));
 973             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 974             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
 975
 976             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 977
 978             /* Update potential sum for this i atom from the interaction with this j atom. */
 979             velec            = _mm256_and_pd(velec,cutoff_mask);
 980             velec            = _mm256_andnot_pd(dummy_mask,velec);
 981             velecsum         = _mm256_add_pd(velecsum,velec);
 982             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 983             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 984             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 985
 986             fscal            = _mm256_add_pd(felec,fvdw);
 987
 988             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 989
 990             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 991
 992             /* Calculate temporary vectorial force */
 993             tx               = _mm256_mul_pd(fscal,dx00);
 994             ty               = _mm256_mul_pd(fscal,dy00);
 995             tz               = _mm256_mul_pd(fscal,dz00);
 996
 997             /* Update vectorial force */
 998             fix0             = _mm256_add_pd(fix0,tx);
 999             fiy0             = _mm256_add_pd(fiy0,ty);
1000             fiz0             = _mm256_add_pd(fiz0,tz);
1001
1002             fjx0             = _mm256_add_pd(fjx0,tx);
1003             fjy0             = _mm256_add_pd(fjy0,ty);
1004             fjz0             = _mm256_add_pd(fjz0,tz);
1005
1006             }
1007
1008             /**************************
1009              * CALCULATE INTERACTIONS *
1010              **************************/
1011
1012             if (gmx_mm256_any_lt(rsq01,rcutoff2))
1013             {
1014
1015             r01              = _mm256_mul_pd(rsq01,rinv01);
1016             r01              = _mm256_andnot_pd(dummy_mask,r01);
1017
1018             /* EWALD ELECTROSTATICS */
1019
1020             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1021             ewrt             = _mm256_mul_pd(r01,ewtabscale);
1022             ewitab           = _mm256_cvttpd_epi32(ewrt);
1023             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1024             ewitab           = _mm_slli_epi32(ewitab,2);
1025             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1026             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1027             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1028             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1029             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1030             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1031             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1032             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_sub_pd(rinv01,sh_ewald),velec));
1033             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1034
1035             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
1036
1037             /* Update potential sum for this i atom from the interaction with this j atom. */
1038             velec            = _mm256_and_pd(velec,cutoff_mask);
1039             velec            = _mm256_andnot_pd(dummy_mask,velec);
1040             velecsum         = _mm256_add_pd(velecsum,velec);
1041
1042             fscal            = felec;
1043
1044             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1045
1046             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1047
1048             /* Calculate temporary vectorial force */
1049             tx               = _mm256_mul_pd(fscal,dx01);
1050             ty               = _mm256_mul_pd(fscal,dy01);
1051             tz               = _mm256_mul_pd(fscal,dz01);
1052
1053             /* Update vectorial force */
1054             fix0             = _mm256_add_pd(fix0,tx);
1055             fiy0             = _mm256_add_pd(fiy0,ty);
1056             fiz0             = _mm256_add_pd(fiz0,tz);
1057
1058             fjx1             = _mm256_add_pd(fjx1,tx);
1059             fjy1             = _mm256_add_pd(fjy1,ty);
1060             fjz1             = _mm256_add_pd(fjz1,tz);
1061
1062             }
1063
1064             /**************************
1065              * CALCULATE INTERACTIONS *
1066              **************************/
1067
1068             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1069             {
1070
1071             r02              = _mm256_mul_pd(rsq02,rinv02);
1072             r02              = _mm256_andnot_pd(dummy_mask,r02);
1073
1074             /* EWALD ELECTROSTATICS */
1075
1076             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1077             ewrt             = _mm256_mul_pd(r02,ewtabscale);
1078             ewitab           = _mm256_cvttpd_epi32(ewrt);
1079             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1080             ewitab           = _mm_slli_epi32(ewitab,2);
1081             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1082             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1083             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1084             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1085             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1086             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1087             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1088             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_sub_pd(rinv02,sh_ewald),velec));
1089             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1090
1091             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
1092
1093             /* Update potential sum for this i atom from the interaction with this j atom. */
1094             velec            = _mm256_and_pd(velec,cutoff_mask);
1095             velec            = _mm256_andnot_pd(dummy_mask,velec);
1096             velecsum         = _mm256_add_pd(velecsum,velec);
1097
1098             fscal            = felec;
1099
1100             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1101
1102             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1103
1104             /* Calculate temporary vectorial force */
1105             tx               = _mm256_mul_pd(fscal,dx02);
1106             ty               = _mm256_mul_pd(fscal,dy02);
1107             tz               = _mm256_mul_pd(fscal,dz02);
1108
1109             /* Update vectorial force */
1110             fix0             = _mm256_add_pd(fix0,tx);
1111             fiy0             = _mm256_add_pd(fiy0,ty);
1112             fiz0             = _mm256_add_pd(fiz0,tz);
1113
1114             fjx2             = _mm256_add_pd(fjx2,tx);
1115             fjy2             = _mm256_add_pd(fjy2,ty);
1116             fjz2             = _mm256_add_pd(fjz2,tz);
1117
1118             }
1119
1120             /**************************
1121              * CALCULATE INTERACTIONS *
1122              **************************/
1123
1124             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1125             {
1126
1127             r10              = _mm256_mul_pd(rsq10,rinv10);
1128             r10              = _mm256_andnot_pd(dummy_mask,r10);
1129
1130             /* EWALD ELECTROSTATICS */
1131
1132             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1133             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1134             ewitab           = _mm256_cvttpd_epi32(ewrt);
1135             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1136             ewitab           = _mm_slli_epi32(ewitab,2);
1137             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1138             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1139             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1140             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1141             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1142             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1143             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1144             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_sub_pd(rinv10,sh_ewald),velec));
1145             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1146
1147             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1148
1149             /* Update potential sum for this i atom from the interaction with this j atom. */
1150             velec            = _mm256_and_pd(velec,cutoff_mask);
1151             velec            = _mm256_andnot_pd(dummy_mask,velec);
1152             velecsum         = _mm256_add_pd(velecsum,velec);
1153
1154             fscal            = felec;
1155
1156             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1157
1158             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1159
1160             /* Calculate temporary vectorial force */
1161             tx               = _mm256_mul_pd(fscal,dx10);
1162             ty               = _mm256_mul_pd(fscal,dy10);
1163             tz               = _mm256_mul_pd(fscal,dz10);
1164
1165             /* Update vectorial force */
1166             fix1             = _mm256_add_pd(fix1,tx);
1167             fiy1             = _mm256_add_pd(fiy1,ty);
1168             fiz1             = _mm256_add_pd(fiz1,tz);
1169
1170             fjx0             = _mm256_add_pd(fjx0,tx);
1171             fjy0             = _mm256_add_pd(fjy0,ty);
1172             fjz0             = _mm256_add_pd(fjz0,tz);
1173
1174             }
1175
1176             /**************************
1177              * CALCULATE INTERACTIONS *
1178              **************************/
1179
1180             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1181             {
1182
1183             r11              = _mm256_mul_pd(rsq11,rinv11);
1184             r11              = _mm256_andnot_pd(dummy_mask,r11);
1185
1186             /* EWALD ELECTROSTATICS */
1187
1188             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1189             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1190             ewitab           = _mm256_cvttpd_epi32(ewrt);
1191             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1192             ewitab           = _mm_slli_epi32(ewitab,2);
1193             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1194             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1195             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1196             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1197             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1198             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1199             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1200             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
1201             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1202
1203             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1204
1205             /* Update potential sum for this i atom from the interaction with this j atom. */
1206             velec            = _mm256_and_pd(velec,cutoff_mask);
1207             velec            = _mm256_andnot_pd(dummy_mask,velec);
1208             velecsum         = _mm256_add_pd(velecsum,velec);
1209
1210             fscal            = felec;
1211
1212             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1213
1214             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1215
1216             /* Calculate temporary vectorial force */
1217             tx               = _mm256_mul_pd(fscal,dx11);
1218             ty               = _mm256_mul_pd(fscal,dy11);
1219             tz               = _mm256_mul_pd(fscal,dz11);
1220
1221             /* Update vectorial force */
1222             fix1             = _mm256_add_pd(fix1,tx);
1223             fiy1             = _mm256_add_pd(fiy1,ty);
1224             fiz1             = _mm256_add_pd(fiz1,tz);
1225
1226             fjx1             = _mm256_add_pd(fjx1,tx);
1227             fjy1             = _mm256_add_pd(fjy1,ty);
1228             fjz1             = _mm256_add_pd(fjz1,tz);
1229
1230             }
1231
1232             /**************************
1233              * CALCULATE INTERACTIONS *
1234              **************************/
1235
1236             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1237             {
1238
1239             r12              = _mm256_mul_pd(rsq12,rinv12);
1240             r12              = _mm256_andnot_pd(dummy_mask,r12);
1241
1242             /* EWALD ELECTROSTATICS */
1243
1244             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1245             ewrt             = _mm256_mul_pd(r12,ewtabscale);
1246             ewitab           = _mm256_cvttpd_epi32(ewrt);
1247             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1248             ewitab           = _mm_slli_epi32(ewitab,2);
1249             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1250             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1251             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1252             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1253             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1254             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1255             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1256             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
1257             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1258
1259             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
1260
1261             /* Update potential sum for this i atom from the interaction with this j atom. */
1262             velec            = _mm256_and_pd(velec,cutoff_mask);
1263             velec            = _mm256_andnot_pd(dummy_mask,velec);
1264             velecsum         = _mm256_add_pd(velecsum,velec);
1265
1266             fscal            = felec;
1267
1268             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1269
1270             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1271
1272             /* Calculate temporary vectorial force */
1273             tx               = _mm256_mul_pd(fscal,dx12);
1274             ty               = _mm256_mul_pd(fscal,dy12);
1275             tz               = _mm256_mul_pd(fscal,dz12);
1276
1277             /* Update vectorial force */
1278             fix1             = _mm256_add_pd(fix1,tx);
1279             fiy1             = _mm256_add_pd(fiy1,ty);
1280             fiz1             = _mm256_add_pd(fiz1,tz);
1281
1282             fjx2             = _mm256_add_pd(fjx2,tx);
1283             fjy2             = _mm256_add_pd(fjy2,ty);
1284             fjz2             = _mm256_add_pd(fjz2,tz);
1285
1286             }
1287
1288             /**************************
1289              * CALCULATE INTERACTIONS *
1290              **************************/
1291
1292             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1293             {
1294
1295             r20              = _mm256_mul_pd(rsq20,rinv20);
1296             r20              = _mm256_andnot_pd(dummy_mask,r20);
1297
1298             /* EWALD ELECTROSTATICS */
1299
1300             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1301             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1302             ewitab           = _mm256_cvttpd_epi32(ewrt);
1303             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1304             ewitab           = _mm_slli_epi32(ewitab,2);
1305             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1306             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1307             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1308             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1309             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1310             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1311             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1312             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_sub_pd(rinv20,sh_ewald),velec));
1313             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1314
1315             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1316
1317             /* Update potential sum for this i atom from the interaction with this j atom. */
1318             velec            = _mm256_and_pd(velec,cutoff_mask);
1319             velec            = _mm256_andnot_pd(dummy_mask,velec);
1320             velecsum         = _mm256_add_pd(velecsum,velec);
1321
1322             fscal            = felec;
1323
1324             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1325
1326             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1327
1328             /* Calculate temporary vectorial force */
1329             tx               = _mm256_mul_pd(fscal,dx20);
1330             ty               = _mm256_mul_pd(fscal,dy20);
1331             tz               = _mm256_mul_pd(fscal,dz20);
1332
1333             /* Update vectorial force */
1334             fix2             = _mm256_add_pd(fix2,tx);
1335             fiy2             = _mm256_add_pd(fiy2,ty);
1336             fiz2             = _mm256_add_pd(fiz2,tz);
1337
1338             fjx0             = _mm256_add_pd(fjx0,tx);
1339             fjy0             = _mm256_add_pd(fjy0,ty);
1340             fjz0             = _mm256_add_pd(fjz0,tz);
1341
1342             }
1343
1344             /**************************
1345              * CALCULATE INTERACTIONS *
1346              **************************/
1347
1348             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1349             {
1350
1351             r21              = _mm256_mul_pd(rsq21,rinv21);
1352             r21              = _mm256_andnot_pd(dummy_mask,r21);
1353
1354             /* EWALD ELECTROSTATICS */
1355
1356             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1357             ewrt             = _mm256_mul_pd(r21,ewtabscale);
1358             ewitab           = _mm256_cvttpd_epi32(ewrt);
1359             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1360             ewitab           = _mm_slli_epi32(ewitab,2);
1361             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1362             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1363             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1364             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1365             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1366             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1367             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1368             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
1369             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1370
1371             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
1372
1373             /* Update potential sum for this i atom from the interaction with this j atom. */
1374             velec            = _mm256_and_pd(velec,cutoff_mask);
1375             velec            = _mm256_andnot_pd(dummy_mask,velec);
1376             velecsum         = _mm256_add_pd(velecsum,velec);
1377
1378             fscal            = felec;
1379
1380             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1381
1382             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1383
1384             /* Calculate temporary vectorial force */
1385             tx               = _mm256_mul_pd(fscal,dx21);
1386             ty               = _mm256_mul_pd(fscal,dy21);
1387             tz               = _mm256_mul_pd(fscal,dz21);
1388
1389             /* Update vectorial force */
1390             fix2             = _mm256_add_pd(fix2,tx);
1391             fiy2             = _mm256_add_pd(fiy2,ty);
1392             fiz2             = _mm256_add_pd(fiz2,tz);
1393
1394             fjx1             = _mm256_add_pd(fjx1,tx);
1395             fjy1             = _mm256_add_pd(fjy1,ty);
1396             fjz1             = _mm256_add_pd(fjz1,tz);
1397
1398             }
1399
1400             /**************************
1401              * CALCULATE INTERACTIONS *
1402              **************************/
1403
1404             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1405             {
1406
1407             r22              = _mm256_mul_pd(rsq22,rinv22);
1408             r22              = _mm256_andnot_pd(dummy_mask,r22);
1409
1410             /* EWALD ELECTROSTATICS */
1411
1412             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1413             ewrt             = _mm256_mul_pd(r22,ewtabscale);
1414             ewitab           = _mm256_cvttpd_epi32(ewrt);
1415             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1416             ewitab           = _mm_slli_epi32(ewitab,2);
1417             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1418             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1419             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1420             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1421             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1422             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1423             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1424             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
1425             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1426
1427             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
1428
1429             /* Update potential sum for this i atom from the interaction with this j atom. */
1430             velec            = _mm256_and_pd(velec,cutoff_mask);
1431             velec            = _mm256_andnot_pd(dummy_mask,velec);
1432             velecsum         = _mm256_add_pd(velecsum,velec);
1433
1434             fscal            = felec;
1435
1436             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1437
1438             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1439
1440             /* Calculate temporary vectorial force */
1441             tx               = _mm256_mul_pd(fscal,dx22);
1442             ty               = _mm256_mul_pd(fscal,dy22);
1443             tz               = _mm256_mul_pd(fscal,dz22);
1444
1445             /* Update vectorial force */
1446             fix2             = _mm256_add_pd(fix2,tx);
1447             fiy2             = _mm256_add_pd(fiy2,ty);
1448             fiz2             = _mm256_add_pd(fiz2,tz);
1449
1450             fjx2             = _mm256_add_pd(fjx2,tx);
1451             fjy2             = _mm256_add_pd(fjy2,ty);
1452             fjz2             = _mm256_add_pd(fjz2,tz);
1453
1454             }
1455
1456             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1457             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1458             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1459             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1460
1461             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1462                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1463
1464             /* Inner loop uses 459 flops */
1465         }
1466
1467         /* End of innermost loop */
1468
1469         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1470                                                  f+i_coord_offset,fshift+i_shift_offset);
1471
1472         ggid                        = gid[iidx];
1473         /* Update potential energies */
1474         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1475         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1476
1477         /* Increment number of inner iterations */
1478         inneriter                  += j_index_end - j_index_start;
1479
1480         /* Outer loop uses 20 flops */
1481     }
1482
1483     /* Increment number of outer iterations */
1484     outeriter        += nri;
1485
1486     /* Update outer/inner flops */
1487
1488     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*459);
1489 }
1490 /*
1491  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_F_avx_256_double
1492  * Electrostatics interaction: Ewald
1493  * VdW interaction:            LJEwald
1494  * Geometry:                   Water3-Water3
1495  * Calculate force/pot:        Force
1496  */
1497 void
1498 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW3W3_F_avx_256_double
1499                     (t_nblist                    * gmx_restrict       nlist,
1500                      rvec                        * gmx_restrict          xx,
1501                      rvec                        * gmx_restrict          ff,
1502                      struct t_forcerec           * gmx_restrict          fr,
1503                      t_mdatoms                   * gmx_restrict     mdatoms,
1504                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1505                      t_nrnb                      * gmx_restrict        nrnb)
1506 {
1507     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1508      * just 0 for non-waters.
1509      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1510      * jnr indices corresponding to data put in the four positions in the SIMD register.
1511      */
1512     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1513     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1514     int              jnrA,jnrB,jnrC,jnrD;
1515     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1516     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1517     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1518     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1519     real             rcutoff_scalar;
1520     real             *shiftvec,*fshift,*x,*f;
1521     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1522     real             scratch[4*DIM];
1523     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1524     real *           vdwioffsetptr0;
1525     real *           vdwgridioffsetptr0;
1526     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1527     real *           vdwioffsetptr1;
1528     real *           vdwgridioffsetptr1;
1529     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1530     real *           vdwioffsetptr2;
1531     real *           vdwgridioffsetptr2;
1532     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1533     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1534     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1535     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1536     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1537     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1538     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1539     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1540     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1541     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1542     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1543     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1544     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1545     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1546     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1547     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1548     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1549     real             *charge;
1550     int              nvdwtype;
1551     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1552     int              *vdwtype;
1553     real             *vdwparam;
1554     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1555     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1556     __m256d           c6grid_00;
1557     __m256d           c6grid_01;
1558     __m256d           c6grid_02;
1559     __m256d           c6grid_10;
1560     __m256d           c6grid_11;
1561     __m256d           c6grid_12;
1562     __m256d           c6grid_20;
1563     __m256d           c6grid_21;
1564     __m256d           c6grid_22;
1565     real             *vdwgridparam;
1566     __m256d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1567     __m256d           one_half  = _mm256_set1_pd(0.5);
1568     __m256d           minus_one = _mm256_set1_pd(-1.0);
1569     __m128i          ewitab;
1570     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1571     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1572     real             *ewtab;
1573     __m256d          dummy_mask,cutoff_mask;
1574     __m128           tmpmask0,tmpmask1;
1575     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1576     __m256d          one     = _mm256_set1_pd(1.0);
1577     __m256d          two     = _mm256_set1_pd(2.0);
1578     x                = xx[0];
1579     f                = ff[0];
1580
1581     nri              = nlist->nri;
1582     iinr             = nlist->iinr;
1583     jindex           = nlist->jindex;
1584     jjnr             = nlist->jjnr;
1585     shiftidx         = nlist->shift;
1586     gid              = nlist->gid;
1587     shiftvec         = fr->shift_vec[0];
1588     fshift           = fr->fshift[0];
1589     facel            = _mm256_set1_pd(fr->ic->epsfac);
1590     charge           = mdatoms->chargeA;
1591     nvdwtype         = fr->ntype;
1592     vdwparam         = fr->nbfp;
1593     vdwtype          = mdatoms->typeA;
1594     vdwgridparam     = fr->ljpme_c6grid;
1595     sh_lj_ewald      = _mm256_set1_pd(fr->ic->sh_lj_ewald);
1596     ewclj            = _mm256_set1_pd(fr->ic->ewaldcoeff_lj);
1597     ewclj2           = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
1598
1599     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
1600     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1601     beta2            = _mm256_mul_pd(beta,beta);
1602     beta3            = _mm256_mul_pd(beta,beta2);
1603
1604     ewtab            = fr->ic->tabq_coul_F;
1605     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
1606     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1607
1608     /* Setup water-specific parameters */
1609     inr              = nlist->iinr[0];
1610     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1611     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1612     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1613     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1614     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1615
1616     jq0              = _mm256_set1_pd(charge[inr+0]);
1617     jq1              = _mm256_set1_pd(charge[inr+1]);
1618     jq2              = _mm256_set1_pd(charge[inr+2]);
1619     vdwjidx0A        = 2*vdwtype[inr+0];
1620     qq00             = _mm256_mul_pd(iq0,jq0);
1621     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1622     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1623     c6grid_00        = _mm256_set1_pd(vdwgridioffsetptr0[vdwjidx0A]);
1624     qq01             = _mm256_mul_pd(iq0,jq1);
1625     qq02             = _mm256_mul_pd(iq0,jq2);
1626     qq10             = _mm256_mul_pd(iq1,jq0);
1627     qq11             = _mm256_mul_pd(iq1,jq1);
1628     qq12             = _mm256_mul_pd(iq1,jq2);
1629     qq20             = _mm256_mul_pd(iq2,jq0);
1630     qq21             = _mm256_mul_pd(iq2,jq1);
1631     qq22             = _mm256_mul_pd(iq2,jq2);
1632
1633     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1634     rcutoff_scalar   = fr->ic->rcoulomb;
1635     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
1636     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
1637
1638     sh_vdw_invrcut6  = _mm256_set1_pd(fr->ic->sh_invrc6);
1639     rvdw             = _mm256_set1_pd(fr->ic->rvdw);
1640
1641     /* Avoid stupid compiler warnings */
1642     jnrA = jnrB = jnrC = jnrD = 0;
1643     j_coord_offsetA = 0;
1644     j_coord_offsetB = 0;
1645     j_coord_offsetC = 0;
1646     j_coord_offsetD = 0;
1647
1648     outeriter        = 0;
1649     inneriter        = 0;
1650
1651     for(iidx=0;iidx<4*DIM;iidx++)
1652     {
1653         scratch[iidx] = 0.0;
1654     }
1655
1656     /* Start outer loop over neighborlists */
1657     for(iidx=0; iidx<nri; iidx++)
1658     {
1659         /* Load shift vector for this list */
1660         i_shift_offset   = DIM*shiftidx[iidx];
1661
1662         /* Load limits for loop over neighbors */
1663         j_index_start    = jindex[iidx];
1664         j_index_end      = jindex[iidx+1];
1665
1666         /* Get outer coordinate index */
1667         inr              = iinr[iidx];
1668         i_coord_offset   = DIM*inr;
1669
1670         /* Load i particle coords and add shift vector */
1671         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1672                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1673
1674         fix0             = _mm256_setzero_pd();
1675         fiy0             = _mm256_setzero_pd();
1676         fiz0             = _mm256_setzero_pd();
1677         fix1             = _mm256_setzero_pd();
1678         fiy1             = _mm256_setzero_pd();
1679         fiz1             = _mm256_setzero_pd();
1680         fix2             = _mm256_setzero_pd();
1681         fiy2             = _mm256_setzero_pd();
1682         fiz2             = _mm256_setzero_pd();
1683
1684         /* Start inner kernel loop */
1685         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1686         {
1687
1688             /* Get j neighbor index, and coordinate index */
1689             jnrA             = jjnr[jidx];
1690             jnrB             = jjnr[jidx+1];
1691             jnrC             = jjnr[jidx+2];
1692             jnrD             = jjnr[jidx+3];
1693             j_coord_offsetA  = DIM*jnrA;
1694             j_coord_offsetB  = DIM*jnrB;
1695             j_coord_offsetC  = DIM*jnrC;
1696             j_coord_offsetD  = DIM*jnrD;
1697
1698             /* load j atom coordinates */
1699             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1700                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1701                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1702
1703             /* Calculate displacement vector */
1704             dx00             = _mm256_sub_pd(ix0,jx0);
1705             dy00             = _mm256_sub_pd(iy0,jy0);
1706             dz00             = _mm256_sub_pd(iz0,jz0);
1707             dx01             = _mm256_sub_pd(ix0,jx1);
1708             dy01             = _mm256_sub_pd(iy0,jy1);
1709             dz01             = _mm256_sub_pd(iz0,jz1);
1710             dx02             = _mm256_sub_pd(ix0,jx2);
1711             dy02             = _mm256_sub_pd(iy0,jy2);
1712             dz02             = _mm256_sub_pd(iz0,jz2);
1713             dx10             = _mm256_sub_pd(ix1,jx0);
1714             dy10             = _mm256_sub_pd(iy1,jy0);
1715             dz10             = _mm256_sub_pd(iz1,jz0);
1716             dx11             = _mm256_sub_pd(ix1,jx1);
1717             dy11             = _mm256_sub_pd(iy1,jy1);
1718             dz11             = _mm256_sub_pd(iz1,jz1);
1719             dx12             = _mm256_sub_pd(ix1,jx2);
1720             dy12             = _mm256_sub_pd(iy1,jy2);
1721             dz12             = _mm256_sub_pd(iz1,jz2);
1722             dx20             = _mm256_sub_pd(ix2,jx0);
1723             dy20             = _mm256_sub_pd(iy2,jy0);
1724             dz20             = _mm256_sub_pd(iz2,jz0);
1725             dx21             = _mm256_sub_pd(ix2,jx1);
1726             dy21             = _mm256_sub_pd(iy2,jy1);
1727             dz21             = _mm256_sub_pd(iz2,jz1);
1728             dx22             = _mm256_sub_pd(ix2,jx2);
1729             dy22             = _mm256_sub_pd(iy2,jy2);
1730             dz22             = _mm256_sub_pd(iz2,jz2);
1731
1732             /* Calculate squared distance and things based on it */
1733             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1734             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1735             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1736             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1737             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1738             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1739             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1740             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1741             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1742
1743             rinv00           = avx256_invsqrt_d(rsq00);
1744             rinv01           = avx256_invsqrt_d(rsq01);
1745             rinv02           = avx256_invsqrt_d(rsq02);
1746             rinv10           = avx256_invsqrt_d(rsq10);
1747             rinv11           = avx256_invsqrt_d(rsq11);
1748             rinv12           = avx256_invsqrt_d(rsq12);
1749             rinv20           = avx256_invsqrt_d(rsq20);
1750             rinv21           = avx256_invsqrt_d(rsq21);
1751             rinv22           = avx256_invsqrt_d(rsq22);
1752
1753             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1754             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1755             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1756             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1757             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1758             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1759             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1760             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1761             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1762
1763             fjx0             = _mm256_setzero_pd();
1764             fjy0             = _mm256_setzero_pd();
1765             fjz0             = _mm256_setzero_pd();
1766             fjx1             = _mm256_setzero_pd();
1767             fjy1             = _mm256_setzero_pd();
1768             fjz1             = _mm256_setzero_pd();
1769             fjx2             = _mm256_setzero_pd();
1770             fjy2             = _mm256_setzero_pd();
1771             fjz2             = _mm256_setzero_pd();
1772
1773             /**************************
1774              * CALCULATE INTERACTIONS *
1775              **************************/
1776
1777             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1778             {
1779
1780             r00              = _mm256_mul_pd(rsq00,rinv00);
1781
1782             /* EWALD ELECTROSTATICS */
1783
1784             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1785             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1786             ewitab           = _mm256_cvttpd_epi32(ewrt);
1787             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1788             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1789                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1790                                             &ewtabF,&ewtabFn);
1791             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1792             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1793
1794             /* Analytical LJ-PME */
1795             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1796             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
1797             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
1798             exponent         = avx256_exp_d(ewcljrsq);
1799             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1800             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
1801             /* f6A = 6 * C6grid * (1 - poly) */
1802             f6A              = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
1803             /* f6B = C6grid * exponent * beta^6 */
1804             f6B              = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
1805             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1806             fvdw              = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1807
1808             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1809
1810             fscal            = _mm256_add_pd(felec,fvdw);
1811
1812             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1813
1814             /* Calculate temporary vectorial force */
1815             tx               = _mm256_mul_pd(fscal,dx00);
1816             ty               = _mm256_mul_pd(fscal,dy00);
1817             tz               = _mm256_mul_pd(fscal,dz00);
1818
1819             /* Update vectorial force */
1820             fix0             = _mm256_add_pd(fix0,tx);
1821             fiy0             = _mm256_add_pd(fiy0,ty);
1822             fiz0             = _mm256_add_pd(fiz0,tz);
1823
1824             fjx0             = _mm256_add_pd(fjx0,tx);
1825             fjy0             = _mm256_add_pd(fjy0,ty);
1826             fjz0             = _mm256_add_pd(fjz0,tz);
1827
1828             }
1829
1830             /**************************
1831              * CALCULATE INTERACTIONS *
1832              **************************/
1833
1834             if (gmx_mm256_any_lt(rsq01,rcutoff2))
1835             {
1836
1837             r01              = _mm256_mul_pd(rsq01,rinv01);
1838
1839             /* EWALD ELECTROSTATICS */
1840
1841             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1842             ewrt             = _mm256_mul_pd(r01,ewtabscale);
1843             ewitab           = _mm256_cvttpd_epi32(ewrt);
1844             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1845             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1846                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1847                                             &ewtabF,&ewtabFn);
1848             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1849             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1850
1851             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
1852
1853             fscal            = felec;
1854
1855             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1856
1857             /* Calculate temporary vectorial force */
1858             tx               = _mm256_mul_pd(fscal,dx01);
1859             ty               = _mm256_mul_pd(fscal,dy01);
1860             tz               = _mm256_mul_pd(fscal,dz01);
1861
1862             /* Update vectorial force */
1863             fix0             = _mm256_add_pd(fix0,tx);
1864             fiy0             = _mm256_add_pd(fiy0,ty);
1865             fiz0             = _mm256_add_pd(fiz0,tz);
1866
1867             fjx1             = _mm256_add_pd(fjx1,tx);
1868             fjy1             = _mm256_add_pd(fjy1,ty);
1869             fjz1             = _mm256_add_pd(fjz1,tz);
1870
1871             }
1872
1873             /**************************
1874              * CALCULATE INTERACTIONS *
1875              **************************/
1876
1877             if (gmx_mm256_any_lt(rsq02,rcutoff2))
1878             {
1879
1880             r02              = _mm256_mul_pd(rsq02,rinv02);
1881
1882             /* EWALD ELECTROSTATICS */
1883
1884             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1885             ewrt             = _mm256_mul_pd(r02,ewtabscale);
1886             ewitab           = _mm256_cvttpd_epi32(ewrt);
1887             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1888             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1889                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1890                                             &ewtabF,&ewtabFn);
1891             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1892             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1893
1894             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
1895
1896             fscal            = felec;
1897
1898             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1899
1900             /* Calculate temporary vectorial force */
1901             tx               = _mm256_mul_pd(fscal,dx02);
1902             ty               = _mm256_mul_pd(fscal,dy02);
1903             tz               = _mm256_mul_pd(fscal,dz02);
1904
1905             /* Update vectorial force */
1906             fix0             = _mm256_add_pd(fix0,tx);
1907             fiy0             = _mm256_add_pd(fiy0,ty);
1908             fiz0             = _mm256_add_pd(fiz0,tz);
1909
1910             fjx2             = _mm256_add_pd(fjx2,tx);
1911             fjy2             = _mm256_add_pd(fjy2,ty);
1912             fjz2             = _mm256_add_pd(fjz2,tz);
1913
1914             }
1915
1916             /**************************
1917              * CALCULATE INTERACTIONS *
1918              **************************/
1919
1920             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1921             {
1922
1923             r10              = _mm256_mul_pd(rsq10,rinv10);
1924
1925             /* EWALD ELECTROSTATICS */
1926
1927             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1928             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1929             ewitab           = _mm256_cvttpd_epi32(ewrt);
1930             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1931             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1932                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1933                                             &ewtabF,&ewtabFn);
1934             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1935             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1936
1937             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1938
1939             fscal            = felec;
1940
1941             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1942
1943             /* Calculate temporary vectorial force */
1944             tx               = _mm256_mul_pd(fscal,dx10);
1945             ty               = _mm256_mul_pd(fscal,dy10);
1946             tz               = _mm256_mul_pd(fscal,dz10);
1947
1948             /* Update vectorial force */
1949             fix1             = _mm256_add_pd(fix1,tx);
1950             fiy1             = _mm256_add_pd(fiy1,ty);
1951             fiz1             = _mm256_add_pd(fiz1,tz);
1952
1953             fjx0             = _mm256_add_pd(fjx0,tx);
1954             fjy0             = _mm256_add_pd(fjy0,ty);
1955             fjz0             = _mm256_add_pd(fjz0,tz);
1956
1957             }
1958
1959             /**************************
1960              * CALCULATE INTERACTIONS *
1961              **************************/
1962
1963             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1964             {
1965
1966             r11              = _mm256_mul_pd(rsq11,rinv11);
1967
1968             /* EWALD ELECTROSTATICS */
1969
1970             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1971             ewrt             = _mm256_mul_pd(r11,ewtabscale);
1972             ewitab           = _mm256_cvttpd_epi32(ewrt);
1973             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1974             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1975                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1976                                             &ewtabF,&ewtabFn);
1977             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1978             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1979
1980             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1981
1982             fscal            = felec;
1983
1984             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1985
1986             /* Calculate temporary vectorial force */
1987             tx               = _mm256_mul_pd(fscal,dx11);
1988             ty               = _mm256_mul_pd(fscal,dy11);
1989             tz               = _mm256_mul_pd(fscal,dz11);
1990
1991             /* Update vectorial force */
1992             fix1             = _mm256_add_pd(fix1,tx);
1993             fiy1             = _mm256_add_pd(fiy1,ty);
1994             fiz1             = _mm256_add_pd(fiz1,tz);
1995
1996             fjx1             = _mm256_add_pd(fjx1,tx);
1997             fjy1             = _mm256_add_pd(fjy1,ty);
1998             fjz1             = _mm256_add_pd(fjz1,tz);
1999
2000             }
2001
2002             /**************************
2003              * CALCULATE INTERACTIONS *
2004              **************************/
2005
2006             if (gmx_mm256_any_lt(rsq12,rcutoff2))
2007             {
2008
2009             r12              = _mm256_mul_pd(rsq12,rinv12);
2010
2011             /* EWALD ELECTROSTATICS */
2012
2013             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2014             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2015             ewitab           = _mm256_cvttpd_epi32(ewrt);
2016             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2017             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2018                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2019                                             &ewtabF,&ewtabFn);
2020             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2021             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2022
2023             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
2024
2025             fscal            = felec;
2026
2027             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2028
2029             /* Calculate temporary vectorial force */
2030             tx               = _mm256_mul_pd(fscal,dx12);
2031             ty               = _mm256_mul_pd(fscal,dy12);
2032             tz               = _mm256_mul_pd(fscal,dz12);
2033
2034             /* Update vectorial force */
2035             fix1             = _mm256_add_pd(fix1,tx);
2036             fiy1             = _mm256_add_pd(fiy1,ty);
2037             fiz1             = _mm256_add_pd(fiz1,tz);
2038
2039             fjx2             = _mm256_add_pd(fjx2,tx);
2040             fjy2             = _mm256_add_pd(fjy2,ty);
2041             fjz2             = _mm256_add_pd(fjz2,tz);
2042
2043             }
2044
2045             /**************************
2046              * CALCULATE INTERACTIONS *
2047              **************************/
2048
2049             if (gmx_mm256_any_lt(rsq20,rcutoff2))
2050             {
2051
2052             r20              = _mm256_mul_pd(rsq20,rinv20);
2053
2054             /* EWALD ELECTROSTATICS */
2055
2056             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2057             ewrt             = _mm256_mul_pd(r20,ewtabscale);
2058             ewitab           = _mm256_cvttpd_epi32(ewrt);
2059             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2060             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2061                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2062                                             &ewtabF,&ewtabFn);
2063             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2064             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2065
2066             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
2067
2068             fscal            = felec;
2069
2070             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2071
2072             /* Calculate temporary vectorial force */
2073             tx               = _mm256_mul_pd(fscal,dx20);
2074             ty               = _mm256_mul_pd(fscal,dy20);
2075             tz               = _mm256_mul_pd(fscal,dz20);
2076
2077             /* Update vectorial force */
2078             fix2             = _mm256_add_pd(fix2,tx);
2079             fiy2             = _mm256_add_pd(fiy2,ty);
2080             fiz2             = _mm256_add_pd(fiz2,tz);
2081
2082             fjx0             = _mm256_add_pd(fjx0,tx);
2083             fjy0             = _mm256_add_pd(fjy0,ty);
2084             fjz0             = _mm256_add_pd(fjz0,tz);
2085
2086             }
2087
2088             /**************************
2089              * CALCULATE INTERACTIONS *
2090              **************************/
2091
2092             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2093             {
2094
2095             r21              = _mm256_mul_pd(rsq21,rinv21);
2096
2097             /* EWALD ELECTROSTATICS */
2098
2099             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2100             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2101             ewitab           = _mm256_cvttpd_epi32(ewrt);
2102             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2103             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2104                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2105                                             &ewtabF,&ewtabFn);
2106             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2107             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2108
2109             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2110
2111             fscal            = felec;
2112
2113             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2114
2115             /* Calculate temporary vectorial force */
2116             tx               = _mm256_mul_pd(fscal,dx21);
2117             ty               = _mm256_mul_pd(fscal,dy21);
2118             tz               = _mm256_mul_pd(fscal,dz21);
2119
2120             /* Update vectorial force */
2121             fix2             = _mm256_add_pd(fix2,tx);
2122             fiy2             = _mm256_add_pd(fiy2,ty);
2123             fiz2             = _mm256_add_pd(fiz2,tz);
2124
2125             fjx1             = _mm256_add_pd(fjx1,tx);
2126             fjy1             = _mm256_add_pd(fjy1,ty);
2127             fjz1             = _mm256_add_pd(fjz1,tz);
2128
2129             }
2130
2131             /**************************
2132              * CALCULATE INTERACTIONS *
2133              **************************/
2134
2135             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2136             {
2137
2138             r22              = _mm256_mul_pd(rsq22,rinv22);
2139
2140             /* EWALD ELECTROSTATICS */
2141
2142             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2143             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2144             ewitab           = _mm256_cvttpd_epi32(ewrt);
2145             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2146             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2147                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2148                                             &ewtabF,&ewtabFn);
2149             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2150             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2151
2152             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2153
2154             fscal            = felec;
2155
2156             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2157
2158             /* Calculate temporary vectorial force */
2159             tx               = _mm256_mul_pd(fscal,dx22);
2160             ty               = _mm256_mul_pd(fscal,dy22);
2161             tz               = _mm256_mul_pd(fscal,dz22);
2162
2163             /* Update vectorial force */
2164             fix2             = _mm256_add_pd(fix2,tx);
2165             fiy2             = _mm256_add_pd(fiy2,ty);
2166             fiz2             = _mm256_add_pd(fiz2,tz);
2167
2168             fjx2             = _mm256_add_pd(fjx2,tx);
2169             fjy2             = _mm256_add_pd(fjy2,ty);
2170             fjz2             = _mm256_add_pd(fjz2,tz);
2171
2172             }
2173
2174             fjptrA             = f+j_coord_offsetA;
2175             fjptrB             = f+j_coord_offsetB;
2176             fjptrC             = f+j_coord_offsetC;
2177             fjptrD             = f+j_coord_offsetD;
2178
2179             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2180                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2181
2182             /* Inner loop uses 374 flops */
2183         }
2184
2185         if(jidx<j_index_end)
2186         {
2187
2188             /* Get j neighbor index, and coordinate index */
2189             jnrlistA         = jjnr[jidx];
2190             jnrlistB         = jjnr[jidx+1];
2191             jnrlistC         = jjnr[jidx+2];
2192             jnrlistD         = jjnr[jidx+3];
2193             /* Sign of each element will be negative for non-real atoms.
2194              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2195              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
2196              */
2197             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2198
2199             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
2200             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
2201             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
2202
2203             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2204             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2205             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2206             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2207             j_coord_offsetA  = DIM*jnrA;
2208             j_coord_offsetB  = DIM*jnrB;
2209             j_coord_offsetC  = DIM*jnrC;
2210             j_coord_offsetD  = DIM*jnrD;
2211
2212             /* load j atom coordinates */
2213             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
2214                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2215                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2216
2217             /* Calculate displacement vector */
2218             dx00             = _mm256_sub_pd(ix0,jx0);
2219             dy00             = _mm256_sub_pd(iy0,jy0);
2220             dz00             = _mm256_sub_pd(iz0,jz0);
2221             dx01             = _mm256_sub_pd(ix0,jx1);
2222             dy01             = _mm256_sub_pd(iy0,jy1);
2223             dz01             = _mm256_sub_pd(iz0,jz1);
2224             dx02             = _mm256_sub_pd(ix0,jx2);
2225             dy02             = _mm256_sub_pd(iy0,jy2);
2226             dz02             = _mm256_sub_pd(iz0,jz2);
2227             dx10             = _mm256_sub_pd(ix1,jx0);
2228             dy10             = _mm256_sub_pd(iy1,jy0);
2229             dz10             = _mm256_sub_pd(iz1,jz0);
2230             dx11             = _mm256_sub_pd(ix1,jx1);
2231             dy11             = _mm256_sub_pd(iy1,jy1);
2232             dz11             = _mm256_sub_pd(iz1,jz1);
2233             dx12             = _mm256_sub_pd(ix1,jx2);
2234             dy12             = _mm256_sub_pd(iy1,jy2);
2235             dz12             = _mm256_sub_pd(iz1,jz2);
2236             dx20             = _mm256_sub_pd(ix2,jx0);
2237             dy20             = _mm256_sub_pd(iy2,jy0);
2238             dz20             = _mm256_sub_pd(iz2,jz0);
2239             dx21             = _mm256_sub_pd(ix2,jx1);
2240             dy21             = _mm256_sub_pd(iy2,jy1);
2241             dz21             = _mm256_sub_pd(iz2,jz1);
2242             dx22             = _mm256_sub_pd(ix2,jx2);
2243             dy22             = _mm256_sub_pd(iy2,jy2);
2244             dz22             = _mm256_sub_pd(iz2,jz2);
2245
2246             /* Calculate squared distance and things based on it */
2247             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2248             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
2249             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
2250             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
2251             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2252             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2253             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
2254             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2255             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2256
2257             rinv00           = avx256_invsqrt_d(rsq00);
2258             rinv01           = avx256_invsqrt_d(rsq01);
2259             rinv02           = avx256_invsqrt_d(rsq02);
2260             rinv10           = avx256_invsqrt_d(rsq10);
2261             rinv11           = avx256_invsqrt_d(rsq11);
2262             rinv12           = avx256_invsqrt_d(rsq12);
2263             rinv20           = avx256_invsqrt_d(rsq20);
2264             rinv21           = avx256_invsqrt_d(rsq21);
2265             rinv22           = avx256_invsqrt_d(rsq22);
2266
2267             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
2268             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
2269             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
2270             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
2271             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
2272             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
2273             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
2274             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
2275             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
2276
2277             fjx0             = _mm256_setzero_pd();
2278             fjy0             = _mm256_setzero_pd();
2279             fjz0             = _mm256_setzero_pd();
2280             fjx1             = _mm256_setzero_pd();
2281             fjy1             = _mm256_setzero_pd();
2282             fjz1             = _mm256_setzero_pd();
2283             fjx2             = _mm256_setzero_pd();
2284             fjy2             = _mm256_setzero_pd();
2285             fjz2             = _mm256_setzero_pd();
2286
2287             /**************************
2288              * CALCULATE INTERACTIONS *
2289              **************************/
2290
2291             if (gmx_mm256_any_lt(rsq00,rcutoff2))
2292             {
2293
2294             r00              = _mm256_mul_pd(rsq00,rinv00);
2295             r00              = _mm256_andnot_pd(dummy_mask,r00);
2296
2297             /* EWALD ELECTROSTATICS */
2298
2299             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2300             ewrt             = _mm256_mul_pd(r00,ewtabscale);
2301             ewitab           = _mm256_cvttpd_epi32(ewrt);
2302             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2303             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2304                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2305                                             &ewtabF,&ewtabFn);
2306             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2307             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
2308
2309             /* Analytical LJ-PME */
2310             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2311             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
2312             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
2313             exponent         = avx256_exp_d(ewcljrsq);
2314             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2315             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
2316             /* f6A = 6 * C6grid * (1 - poly) */
2317             f6A              = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
2318             /* f6B = C6grid * exponent * beta^6 */
2319             f6B              = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
2320             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2321             fvdw              = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2322
2323             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
2324
2325             fscal            = _mm256_add_pd(felec,fvdw);
2326
2327             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2328
2329             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2330
2331             /* Calculate temporary vectorial force */
2332             tx               = _mm256_mul_pd(fscal,dx00);
2333             ty               = _mm256_mul_pd(fscal,dy00);
2334             tz               = _mm256_mul_pd(fscal,dz00);
2335
2336             /* Update vectorial force */
2337             fix0             = _mm256_add_pd(fix0,tx);
2338             fiy0             = _mm256_add_pd(fiy0,ty);
2339             fiz0             = _mm256_add_pd(fiz0,tz);
2340
2341             fjx0             = _mm256_add_pd(fjx0,tx);
2342             fjy0             = _mm256_add_pd(fjy0,ty);
2343             fjz0             = _mm256_add_pd(fjz0,tz);
2344
2345             }
2346
2347             /**************************
2348              * CALCULATE INTERACTIONS *
2349              **************************/
2350
2351             if (gmx_mm256_any_lt(rsq01,rcutoff2))
2352             {
2353
2354             r01              = _mm256_mul_pd(rsq01,rinv01);
2355             r01              = _mm256_andnot_pd(dummy_mask,r01);
2356
2357             /* EWALD ELECTROSTATICS */
2358
2359             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2360             ewrt             = _mm256_mul_pd(r01,ewtabscale);
2361             ewitab           = _mm256_cvttpd_epi32(ewrt);
2362             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2363             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2364                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2365                                             &ewtabF,&ewtabFn);
2366             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2367             felec            = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
2368
2369             cutoff_mask      = _mm256_cmp_pd(rsq01,rcutoff2,_CMP_LT_OQ);
2370
2371             fscal            = felec;
2372
2373             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2374
2375             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2376
2377             /* Calculate temporary vectorial force */
2378             tx               = _mm256_mul_pd(fscal,dx01);
2379             ty               = _mm256_mul_pd(fscal,dy01);
2380             tz               = _mm256_mul_pd(fscal,dz01);
2381
2382             /* Update vectorial force */
2383             fix0             = _mm256_add_pd(fix0,tx);
2384             fiy0             = _mm256_add_pd(fiy0,ty);
2385             fiz0             = _mm256_add_pd(fiz0,tz);
2386
2387             fjx1             = _mm256_add_pd(fjx1,tx);
2388             fjy1             = _mm256_add_pd(fjy1,ty);
2389             fjz1             = _mm256_add_pd(fjz1,tz);
2390
2391             }
2392
2393             /**************************
2394              * CALCULATE INTERACTIONS *
2395              **************************/
2396
2397             if (gmx_mm256_any_lt(rsq02,rcutoff2))
2398             {
2399
2400             r02              = _mm256_mul_pd(rsq02,rinv02);
2401             r02              = _mm256_andnot_pd(dummy_mask,r02);
2402
2403             /* EWALD ELECTROSTATICS */
2404
2405             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2406             ewrt             = _mm256_mul_pd(r02,ewtabscale);
2407             ewitab           = _mm256_cvttpd_epi32(ewrt);
2408             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2409             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2410                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2411                                             &ewtabF,&ewtabFn);
2412             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2413             felec            = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
2414
2415             cutoff_mask      = _mm256_cmp_pd(rsq02,rcutoff2,_CMP_LT_OQ);
2416
2417             fscal            = felec;
2418
2419             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2420
2421             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2422
2423             /* Calculate temporary vectorial force */
2424             tx               = _mm256_mul_pd(fscal,dx02);
2425             ty               = _mm256_mul_pd(fscal,dy02);
2426             tz               = _mm256_mul_pd(fscal,dz02);
2427
2428             /* Update vectorial force */
2429             fix0             = _mm256_add_pd(fix0,tx);
2430             fiy0             = _mm256_add_pd(fiy0,ty);
2431             fiz0             = _mm256_add_pd(fiz0,tz);
2432
2433             fjx2             = _mm256_add_pd(fjx2,tx);
2434             fjy2             = _mm256_add_pd(fjy2,ty);
2435             fjz2             = _mm256_add_pd(fjz2,tz);
2436
2437             }
2438
2439             /**************************
2440              * CALCULATE INTERACTIONS *
2441              **************************/
2442
2443             if (gmx_mm256_any_lt(rsq10,rcutoff2))
2444             {
2445
2446             r10              = _mm256_mul_pd(rsq10,rinv10);
2447             r10              = _mm256_andnot_pd(dummy_mask,r10);
2448
2449             /* EWALD ELECTROSTATICS */
2450
2451             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2452             ewrt             = _mm256_mul_pd(r10,ewtabscale);
2453             ewitab           = _mm256_cvttpd_epi32(ewrt);
2454             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2455             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2456                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2457                                             &ewtabF,&ewtabFn);
2458             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2459             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
2460
2461             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
2462
2463             fscal            = felec;
2464
2465             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2466
2467             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2468
2469             /* Calculate temporary vectorial force */
2470             tx               = _mm256_mul_pd(fscal,dx10);
2471             ty               = _mm256_mul_pd(fscal,dy10);
2472             tz               = _mm256_mul_pd(fscal,dz10);
2473
2474             /* Update vectorial force */
2475             fix1             = _mm256_add_pd(fix1,tx);
2476             fiy1             = _mm256_add_pd(fiy1,ty);
2477             fiz1             = _mm256_add_pd(fiz1,tz);
2478
2479             fjx0             = _mm256_add_pd(fjx0,tx);
2480             fjy0             = _mm256_add_pd(fjy0,ty);
2481             fjz0             = _mm256_add_pd(fjz0,tz);
2482
2483             }
2484
2485             /**************************
2486              * CALCULATE INTERACTIONS *
2487              **************************/
2488
2489             if (gmx_mm256_any_lt(rsq11,rcutoff2))
2490             {
2491
2492             r11              = _mm256_mul_pd(rsq11,rinv11);
2493             r11              = _mm256_andnot_pd(dummy_mask,r11);
2494
2495             /* EWALD ELECTROSTATICS */
2496
2497             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2498             ewrt             = _mm256_mul_pd(r11,ewtabscale);
2499             ewitab           = _mm256_cvttpd_epi32(ewrt);
2500             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2501             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2502                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2503                                             &ewtabF,&ewtabFn);
2504             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2505             felec            = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2506
2507             cutoff_mask      = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
2508
2509             fscal            = felec;
2510
2511             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2512
2513             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2514
2515             /* Calculate temporary vectorial force */
2516             tx               = _mm256_mul_pd(fscal,dx11);
2517             ty               = _mm256_mul_pd(fscal,dy11);
2518             tz               = _mm256_mul_pd(fscal,dz11);
2519
2520             /* Update vectorial force */
2521             fix1             = _mm256_add_pd(fix1,tx);
2522             fiy1             = _mm256_add_pd(fiy1,ty);
2523             fiz1             = _mm256_add_pd(fiz1,tz);
2524
2525             fjx1             = _mm256_add_pd(fjx1,tx);
2526             fjy1             = _mm256_add_pd(fjy1,ty);
2527             fjz1             = _mm256_add_pd(fjz1,tz);
2528
2529             }
2530
2531             /**************************
2532              * CALCULATE INTERACTIONS *
2533              **************************/
2534
2535             if (gmx_mm256_any_lt(rsq12,rcutoff2))
2536             {
2537
2538             r12              = _mm256_mul_pd(rsq12,rinv12);
2539             r12              = _mm256_andnot_pd(dummy_mask,r12);
2540
2541             /* EWALD ELECTROSTATICS */
2542
2543             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2544             ewrt             = _mm256_mul_pd(r12,ewtabscale);
2545             ewitab           = _mm256_cvttpd_epi32(ewrt);
2546             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2547             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2548                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2549                                             &ewtabF,&ewtabFn);
2550             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2551             felec            = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2552
2553             cutoff_mask      = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
2554
2555             fscal            = felec;
2556
2557             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2558
2559             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2560
2561             /* Calculate temporary vectorial force */
2562             tx               = _mm256_mul_pd(fscal,dx12);
2563             ty               = _mm256_mul_pd(fscal,dy12);
2564             tz               = _mm256_mul_pd(fscal,dz12);
2565
2566             /* Update vectorial force */
2567             fix1             = _mm256_add_pd(fix1,tx);
2568             fiy1             = _mm256_add_pd(fiy1,ty);
2569             fiz1             = _mm256_add_pd(fiz1,tz);
2570
2571             fjx2             = _mm256_add_pd(fjx2,tx);
2572             fjy2             = _mm256_add_pd(fjy2,ty);
2573             fjz2             = _mm256_add_pd(fjz2,tz);
2574
2575             }
2576
2577             /**************************
2578              * CALCULATE INTERACTIONS *
2579              **************************/
2580
2581             if (gmx_mm256_any_lt(rsq20,rcutoff2))
2582             {
2583
2584             r20              = _mm256_mul_pd(rsq20,rinv20);
2585             r20              = _mm256_andnot_pd(dummy_mask,r20);
2586
2587             /* EWALD ELECTROSTATICS */
2588
2589             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2590             ewrt             = _mm256_mul_pd(r20,ewtabscale);
2591             ewitab           = _mm256_cvttpd_epi32(ewrt);
2592             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2593             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2594                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2595                                             &ewtabF,&ewtabFn);
2596             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2597             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2598
2599             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
2600
2601             fscal            = felec;
2602
2603             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2604
2605             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2606
2607             /* Calculate temporary vectorial force */
2608             tx               = _mm256_mul_pd(fscal,dx20);
2609             ty               = _mm256_mul_pd(fscal,dy20);
2610             tz               = _mm256_mul_pd(fscal,dz20);
2611
2612             /* Update vectorial force */
2613             fix2             = _mm256_add_pd(fix2,tx);
2614             fiy2             = _mm256_add_pd(fiy2,ty);
2615             fiz2             = _mm256_add_pd(fiz2,tz);
2616
2617             fjx0             = _mm256_add_pd(fjx0,tx);
2618             fjy0             = _mm256_add_pd(fjy0,ty);
2619             fjz0             = _mm256_add_pd(fjz0,tz);
2620
2621             }
2622
2623             /**************************
2624              * CALCULATE INTERACTIONS *
2625              **************************/
2626
2627             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2628             {
2629
2630             r21              = _mm256_mul_pd(rsq21,rinv21);
2631             r21              = _mm256_andnot_pd(dummy_mask,r21);
2632
2633             /* EWALD ELECTROSTATICS */
2634
2635             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2636             ewrt             = _mm256_mul_pd(r21,ewtabscale);
2637             ewitab           = _mm256_cvttpd_epi32(ewrt);
2638             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2639             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2640                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2641                                             &ewtabF,&ewtabFn);
2642             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2643             felec            = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2644
2645             cutoff_mask      = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2646
2647             fscal            = felec;
2648
2649             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2650
2651             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2652
2653             /* Calculate temporary vectorial force */
2654             tx               = _mm256_mul_pd(fscal,dx21);
2655             ty               = _mm256_mul_pd(fscal,dy21);
2656             tz               = _mm256_mul_pd(fscal,dz21);
2657
2658             /* Update vectorial force */
2659             fix2             = _mm256_add_pd(fix2,tx);
2660             fiy2             = _mm256_add_pd(fiy2,ty);
2661             fiz2             = _mm256_add_pd(fiz2,tz);
2662
2663             fjx1             = _mm256_add_pd(fjx1,tx);
2664             fjy1             = _mm256_add_pd(fjy1,ty);
2665             fjz1             = _mm256_add_pd(fjz1,tz);
2666
2667             }
2668
2669             /**************************
2670              * CALCULATE INTERACTIONS *
2671              **************************/
2672
2673             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2674             {
2675
2676             r22              = _mm256_mul_pd(rsq22,rinv22);
2677             r22              = _mm256_andnot_pd(dummy_mask,r22);
2678
2679             /* EWALD ELECTROSTATICS */
2680
2681             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2682             ewrt             = _mm256_mul_pd(r22,ewtabscale);
2683             ewitab           = _mm256_cvttpd_epi32(ewrt);
2684             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2685             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2686                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2687                                             &ewtabF,&ewtabFn);
2688             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2689             felec            = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2690
2691             cutoff_mask      = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2692
2693             fscal            = felec;
2694
2695             fscal            = _mm256_and_pd(fscal,cutoff_mask);
2696
2697             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2698
2699             /* Calculate temporary vectorial force */
2700             tx               = _mm256_mul_pd(fscal,dx22);
2701             ty               = _mm256_mul_pd(fscal,dy22);
2702             tz               = _mm256_mul_pd(fscal,dz22);
2703
2704             /* Update vectorial force */
2705             fix2             = _mm256_add_pd(fix2,tx);
2706             fiy2             = _mm256_add_pd(fiy2,ty);
2707             fiz2             = _mm256_add_pd(fiz2,tz);
2708
2709             fjx2             = _mm256_add_pd(fjx2,tx);
2710             fjy2             = _mm256_add_pd(fjy2,ty);
2711             fjz2             = _mm256_add_pd(fjz2,tz);
2712
2713             }
2714
2715             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2716             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2717             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2718             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2719
2720             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2721                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2722
2723             /* Inner loop uses 383 flops */
2724         }
2725
2726         /* End of innermost loop */
2727
2728         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2729                                                  f+i_coord_offset,fshift+i_shift_offset);
2730
2731         /* Increment number of inner iterations */
2732         inneriter                  += j_index_end - j_index_start;
2733
2734         /* Outer loop uses 18 flops */
2735     }
2736
2737     /* Increment number of outer iterations */
2738     outeriter        += nri;
2739
2740     /* Update outer/inner flops */
2741
2742     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*383);
2743 }