src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_VF_avx_256_single
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            LJEwald
  56  * Geometry:                   Water4-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_VF_avx_256_single
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrE,jnrF,jnrG,jnrH;
  78     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  79     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  80     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  81     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  86     real             scratch[4*DIM];
  87     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     real *           vdwioffsetptr0;
  89     real *           vdwgridioffsetptr0;
  90     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  91     real *           vdwioffsetptr1;
  92     real *           vdwgridioffsetptr1;
  93     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  94     real *           vdwioffsetptr2;
  95     real *           vdwgridioffsetptr2;
  96     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  97     real *           vdwioffsetptr3;
  98     real *           vdwgridioffsetptr3;
  99     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 100     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 101     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 102     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 104     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 106     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 113     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 114     __m256           c6grid_00;
 115     __m256           c6grid_10;
 116     __m256           c6grid_20;
 117     __m256           c6grid_30;
 118     real             *vdwgridparam;
 119     __m256           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 120     __m256           one_half  = _mm256_set1_ps(0.5);
 121     __m256           minus_one = _mm256_set1_ps(-1.0);
 122     __m256i          ewitab;
 123     __m128i          ewitab_lo,ewitab_hi;
 124     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 125     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 126     real             *ewtab;
 127     __m256           dummy_mask,cutoff_mask;
 128     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 129     __m256           one     = _mm256_set1_ps(1.0);
 130     __m256           two     = _mm256_set1_ps(2.0);
 131     x                = xx[0];
 132     f                = ff[0];
 133
 134     nri              = nlist->nri;
 135     iinr             = nlist->iinr;
 136     jindex           = nlist->jindex;
 137     jjnr             = nlist->jjnr;
 138     shiftidx         = nlist->shift;
 139     gid              = nlist->gid;
 140     shiftvec         = fr->shift_vec[0];
 141     fshift           = fr->fshift[0];
 142     facel            = _mm256_set1_ps(fr->epsfac);
 143     charge           = mdatoms->chargeA;
 144     nvdwtype         = fr->ntype;
 145     vdwparam         = fr->nbfp;
 146     vdwtype          = mdatoms->typeA;
 147     vdwgridparam     = fr->ljpme_c6grid;
 148     sh_lj_ewald      = _mm256_set1_ps(fr->ic->sh_lj_ewald);
 149     ewclj            = _mm256_set1_ps(fr->ewaldcoeff_lj);
 150     ewclj2           = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
 151
 152     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 153     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 154     beta2            = _mm256_mul_ps(beta,beta);
 155     beta3            = _mm256_mul_ps(beta,beta2);
 156
 157     ewtab            = fr->ic->tabq_coul_FDV0;
 158     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 159     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 160
 161     /* Setup water-specific parameters */
 162     inr              = nlist->iinr[0];
 163     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 164     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 165     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 166     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 167     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
 168
 169     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 170     rcutoff_scalar   = fr->rcoulomb;
 171     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 172     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 173
 174     sh_vdw_invrcut6  = _mm256_set1_ps(fr->ic->sh_invrc6);
 175     rvdw             = _mm256_set1_ps(fr->rvdw);
 176
 177     /* Avoid stupid compiler warnings */
 178     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 179     j_coord_offsetA = 0;
 180     j_coord_offsetB = 0;
 181     j_coord_offsetC = 0;
 182     j_coord_offsetD = 0;
 183     j_coord_offsetE = 0;
 184     j_coord_offsetF = 0;
 185     j_coord_offsetG = 0;
 186     j_coord_offsetH = 0;
 187
 188     outeriter        = 0;
 189     inneriter        = 0;
 190
 191     for(iidx=0;iidx<4*DIM;iidx++)
 192     {
 193         scratch[iidx] = 0.0;
 194     }
 195
 196     /* Start outer loop over neighborlists */
 197     for(iidx=0; iidx<nri; iidx++)
 198     {
 199         /* Load shift vector for this list */
 200         i_shift_offset   = DIM*shiftidx[iidx];
 201
 202         /* Load limits for loop over neighbors */
 203         j_index_start    = jindex[iidx];
 204         j_index_end      = jindex[iidx+1];
 205
 206         /* Get outer coordinate index */
 207         inr              = iinr[iidx];
 208         i_coord_offset   = DIM*inr;
 209
 210         /* Load i particle coords and add shift vector */
 211         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 212                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 213
 214         fix0             = _mm256_setzero_ps();
 215         fiy0             = _mm256_setzero_ps();
 216         fiz0             = _mm256_setzero_ps();
 217         fix1             = _mm256_setzero_ps();
 218         fiy1             = _mm256_setzero_ps();
 219         fiz1             = _mm256_setzero_ps();
 220         fix2             = _mm256_setzero_ps();
 221         fiy2             = _mm256_setzero_ps();
 222         fiz2             = _mm256_setzero_ps();
 223         fix3             = _mm256_setzero_ps();
 224         fiy3             = _mm256_setzero_ps();
 225         fiz3             = _mm256_setzero_ps();
 226
 227         /* Reset potential sums */
 228         velecsum         = _mm256_setzero_ps();
 229         vvdwsum          = _mm256_setzero_ps();
 230
 231         /* Start inner kernel loop */
 232         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 233         {
 234
 235             /* Get j neighbor index, and coordinate index */
 236             jnrA             = jjnr[jidx];
 237             jnrB             = jjnr[jidx+1];
 238             jnrC             = jjnr[jidx+2];
 239             jnrD             = jjnr[jidx+3];
 240             jnrE             = jjnr[jidx+4];
 241             jnrF             = jjnr[jidx+5];
 242             jnrG             = jjnr[jidx+6];
 243             jnrH             = jjnr[jidx+7];
 244             j_coord_offsetA  = DIM*jnrA;
 245             j_coord_offsetB  = DIM*jnrB;
 246             j_coord_offsetC  = DIM*jnrC;
 247             j_coord_offsetD  = DIM*jnrD;
 248             j_coord_offsetE  = DIM*jnrE;
 249             j_coord_offsetF  = DIM*jnrF;
 250             j_coord_offsetG  = DIM*jnrG;
 251             j_coord_offsetH  = DIM*jnrH;
 252
 253             /* load j atom coordinates */
 254             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 255                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 256                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 257                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 258                                                  &jx0,&jy0,&jz0);
 259
 260             /* Calculate displacement vector */
 261             dx00             = _mm256_sub_ps(ix0,jx0);
 262             dy00             = _mm256_sub_ps(iy0,jy0);
 263             dz00             = _mm256_sub_ps(iz0,jz0);
 264             dx10             = _mm256_sub_ps(ix1,jx0);
 265             dy10             = _mm256_sub_ps(iy1,jy0);
 266             dz10             = _mm256_sub_ps(iz1,jz0);
 267             dx20             = _mm256_sub_ps(ix2,jx0);
 268             dy20             = _mm256_sub_ps(iy2,jy0);
 269             dz20             = _mm256_sub_ps(iz2,jz0);
 270             dx30             = _mm256_sub_ps(ix3,jx0);
 271             dy30             = _mm256_sub_ps(iy3,jy0);
 272             dz30             = _mm256_sub_ps(iz3,jz0);
 273
 274             /* Calculate squared distance and things based on it */
 275             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 276             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 277             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 278             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
 279
 280             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 281             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 282             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 283             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
 284
 285             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 286             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 287             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 288             rinvsq30         = _mm256_mul_ps(rinv30,rinv30);
 289
 290             /* Load parameters for j particles */
 291             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 292                                                                  charge+jnrC+0,charge+jnrD+0,
 293                                                                  charge+jnrE+0,charge+jnrF+0,
 294                                                                  charge+jnrG+0,charge+jnrH+0);
 295             vdwjidx0A        = 2*vdwtype[jnrA+0];
 296             vdwjidx0B        = 2*vdwtype[jnrB+0];
 297             vdwjidx0C        = 2*vdwtype[jnrC+0];
 298             vdwjidx0D        = 2*vdwtype[jnrD+0];
 299             vdwjidx0E        = 2*vdwtype[jnrE+0];
 300             vdwjidx0F        = 2*vdwtype[jnrF+0];
 301             vdwjidx0G        = 2*vdwtype[jnrG+0];
 302             vdwjidx0H        = 2*vdwtype[jnrH+0];
 303
 304             fjx0             = _mm256_setzero_ps();
 305             fjy0             = _mm256_setzero_ps();
 306             fjz0             = _mm256_setzero_ps();
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 313             {
 314
 315             r00              = _mm256_mul_ps(rsq00,rinv00);
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 319                                             vdwioffsetptr0+vdwjidx0B,
 320                                             vdwioffsetptr0+vdwjidx0C,
 321                                             vdwioffsetptr0+vdwjidx0D,
 322                                             vdwioffsetptr0+vdwjidx0E,
 323                                             vdwioffsetptr0+vdwjidx0F,
 324                                             vdwioffsetptr0+vdwjidx0G,
 325                                             vdwioffsetptr0+vdwjidx0H,
 326                                             &c6_00,&c12_00);
 327
 328             c6grid_00       = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
 329                                                                   vdwgridioffsetptr0+vdwjidx0B,
 330                                                                   vdwgridioffsetptr0+vdwjidx0C,
 331                                                                   vdwgridioffsetptr0+vdwjidx0D,
 332                                                                   vdwgridioffsetptr0+vdwjidx0E,
 333                                                                   vdwgridioffsetptr0+vdwjidx0F,
 334                                                                   vdwgridioffsetptr0+vdwjidx0G,
 335                                                                   vdwgridioffsetptr0+vdwjidx0H);
 336
 337             /* Analytical LJ-PME */
 338             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 339             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
 340             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
 341             exponent         = gmx_simd_exp_r(ewcljrsq);
 342             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 343             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
 344             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 345             vvdw6            = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
 346             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 347             vvdw             = _mm256_sub_ps(_mm256_mul_ps( _mm256_sub_ps(vvdw12 , _mm256_mul_ps(c12_00,_mm256_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 348                                           _mm256_mul_ps( _mm256_sub_ps(vvdw6,_mm256_add_ps(_mm256_mul_ps(c6_00,sh_vdw_invrcut6),_mm256_mul_ps(c6grid_00,sh_lj_ewald))),one_sixth));
 349             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 350             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
 351
 352             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 356             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 357
 358             fscal            = fvdw;
 359
 360             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 361
 362             /* Calculate temporary vectorial force */
 363             tx               = _mm256_mul_ps(fscal,dx00);
 364             ty               = _mm256_mul_ps(fscal,dy00);
 365             tz               = _mm256_mul_ps(fscal,dz00);
 366
 367             /* Update vectorial force */
 368             fix0             = _mm256_add_ps(fix0,tx);
 369             fiy0             = _mm256_add_ps(fiy0,ty);
 370             fiz0             = _mm256_add_ps(fiz0,tz);
 371
 372             fjx0             = _mm256_add_ps(fjx0,tx);
 373             fjy0             = _mm256_add_ps(fjy0,ty);
 374             fjz0             = _mm256_add_ps(fjz0,tz);
 375
 376             }
 377
 378             /**************************
 379              * CALCULATE INTERACTIONS *
 380              **************************/
 381
 382             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 383             {
 384
 385             r10              = _mm256_mul_ps(rsq10,rinv10);
 386
 387             /* Compute parameters for interactions between i and j atoms */
 388             qq10             = _mm256_mul_ps(iq1,jq0);
 389
 390             /* EWALD ELECTROSTATICS */
 391
 392             /* Analytical PME correction */
 393             zeta2            = _mm256_mul_ps(beta2,rsq10);
 394             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 395             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 396             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 397             felec            = _mm256_mul_ps(qq10,felec);
 398             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 399             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 400             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv10,sh_ewald),pmecorrV);
 401             velec            = _mm256_mul_ps(qq10,velec);
 402
 403             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 404
 405             /* Update potential sum for this i atom from the interaction with this j atom. */
 406             velec            = _mm256_and_ps(velec,cutoff_mask);
 407             velecsum         = _mm256_add_ps(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 412
 413             /* Calculate temporary vectorial force */
 414             tx               = _mm256_mul_ps(fscal,dx10);
 415             ty               = _mm256_mul_ps(fscal,dy10);
 416             tz               = _mm256_mul_ps(fscal,dz10);
 417
 418             /* Update vectorial force */
 419             fix1             = _mm256_add_ps(fix1,tx);
 420             fiy1             = _mm256_add_ps(fiy1,ty);
 421             fiz1             = _mm256_add_ps(fiz1,tz);
 422
 423             fjx0             = _mm256_add_ps(fjx0,tx);
 424             fjy0             = _mm256_add_ps(fjy0,ty);
 425             fjz0             = _mm256_add_ps(fjz0,tz);
 426
 427             }
 428
 429             /**************************
 430              * CALCULATE INTERACTIONS *
 431              **************************/
 432
 433             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 434             {
 435
 436             r20              = _mm256_mul_ps(rsq20,rinv20);
 437
 438             /* Compute parameters for interactions between i and j atoms */
 439             qq20             = _mm256_mul_ps(iq2,jq0);
 440
 441             /* EWALD ELECTROSTATICS */
 442
 443             /* Analytical PME correction */
 444             zeta2            = _mm256_mul_ps(beta2,rsq20);
 445             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
 446             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 447             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 448             felec            = _mm256_mul_ps(qq20,felec);
 449             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 450             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 451             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv20,sh_ewald),pmecorrV);
 452             velec            = _mm256_mul_ps(qq20,velec);
 453
 454             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             velec            = _mm256_and_ps(velec,cutoff_mask);
 458             velecsum         = _mm256_add_ps(velecsum,velec);
 459
 460             fscal            = felec;
 461
 462             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 463
 464             /* Calculate temporary vectorial force */
 465             tx               = _mm256_mul_ps(fscal,dx20);
 466             ty               = _mm256_mul_ps(fscal,dy20);
 467             tz               = _mm256_mul_ps(fscal,dz20);
 468
 469             /* Update vectorial force */
 470             fix2             = _mm256_add_ps(fix2,tx);
 471             fiy2             = _mm256_add_ps(fiy2,ty);
 472             fiz2             = _mm256_add_ps(fiz2,tz);
 473
 474             fjx0             = _mm256_add_ps(fjx0,tx);
 475             fjy0             = _mm256_add_ps(fjy0,ty);
 476             fjz0             = _mm256_add_ps(fjz0,tz);
 477
 478             }
 479
 480             /**************************
 481              * CALCULATE INTERACTIONS *
 482              **************************/
 483
 484             if (gmx_mm256_any_lt(rsq30,rcutoff2))
 485             {
 486
 487             r30              = _mm256_mul_ps(rsq30,rinv30);
 488
 489             /* Compute parameters for interactions between i and j atoms */
 490             qq30             = _mm256_mul_ps(iq3,jq0);
 491
 492             /* EWALD ELECTROSTATICS */
 493
 494             /* Analytical PME correction */
 495             zeta2            = _mm256_mul_ps(beta2,rsq30);
 496             rinv3            = _mm256_mul_ps(rinvsq30,rinv30);
 497             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 498             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 499             felec            = _mm256_mul_ps(qq30,felec);
 500             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 501             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 502             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv30,sh_ewald),pmecorrV);
 503             velec            = _mm256_mul_ps(qq30,velec);
 504
 505             cutoff_mask      = _mm256_cmp_ps(rsq30,rcutoff2,_CMP_LT_OQ);
 506
 507             /* Update potential sum for this i atom from the interaction with this j atom. */
 508             velec            = _mm256_and_ps(velec,cutoff_mask);
 509             velecsum         = _mm256_add_ps(velecsum,velec);
 510
 511             fscal            = felec;
 512
 513             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 514
 515             /* Calculate temporary vectorial force */
 516             tx               = _mm256_mul_ps(fscal,dx30);
 517             ty               = _mm256_mul_ps(fscal,dy30);
 518             tz               = _mm256_mul_ps(fscal,dz30);
 519
 520             /* Update vectorial force */
 521             fix3             = _mm256_add_ps(fix3,tx);
 522             fiy3             = _mm256_add_ps(fiy3,ty);
 523             fiz3             = _mm256_add_ps(fiz3,tz);
 524
 525             fjx0             = _mm256_add_ps(fjx0,tx);
 526             fjy0             = _mm256_add_ps(fjy0,ty);
 527             fjz0             = _mm256_add_ps(fjz0,tz);
 528
 529             }
 530
 531             fjptrA             = f+j_coord_offsetA;
 532             fjptrB             = f+j_coord_offsetB;
 533             fjptrC             = f+j_coord_offsetC;
 534             fjptrD             = f+j_coord_offsetD;
 535             fjptrE             = f+j_coord_offsetE;
 536             fjptrF             = f+j_coord_offsetF;
 537             fjptrG             = f+j_coord_offsetG;
 538             fjptrH             = f+j_coord_offsetH;
 539
 540             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 541
 542             /* Inner loop uses 392 flops */
 543         }
 544
 545         if(jidx<j_index_end)
 546         {
 547
 548             /* Get j neighbor index, and coordinate index */
 549             jnrlistA         = jjnr[jidx];
 550             jnrlistB         = jjnr[jidx+1];
 551             jnrlistC         = jjnr[jidx+2];
 552             jnrlistD         = jjnr[jidx+3];
 553             jnrlistE         = jjnr[jidx+4];
 554             jnrlistF         = jjnr[jidx+5];
 555             jnrlistG         = jjnr[jidx+6];
 556             jnrlistH         = jjnr[jidx+7];
 557             /* Sign of each element will be negative for non-real atoms.
 558              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 559              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 560              */
 561             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 562                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 563
 564             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 565             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 566             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 567             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 568             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 569             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 570             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 571             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 572             j_coord_offsetA  = DIM*jnrA;
 573             j_coord_offsetB  = DIM*jnrB;
 574             j_coord_offsetC  = DIM*jnrC;
 575             j_coord_offsetD  = DIM*jnrD;
 576             j_coord_offsetE  = DIM*jnrE;
 577             j_coord_offsetF  = DIM*jnrF;
 578             j_coord_offsetG  = DIM*jnrG;
 579             j_coord_offsetH  = DIM*jnrH;
 580
 581             /* load j atom coordinates */
 582             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 583                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 584                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 585                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 586                                                  &jx0,&jy0,&jz0);
 587
 588             /* Calculate displacement vector */
 589             dx00             = _mm256_sub_ps(ix0,jx0);
 590             dy00             = _mm256_sub_ps(iy0,jy0);
 591             dz00             = _mm256_sub_ps(iz0,jz0);
 592             dx10             = _mm256_sub_ps(ix1,jx0);
 593             dy10             = _mm256_sub_ps(iy1,jy0);
 594             dz10             = _mm256_sub_ps(iz1,jz0);
 595             dx20             = _mm256_sub_ps(ix2,jx0);
 596             dy20             = _mm256_sub_ps(iy2,jy0);
 597             dz20             = _mm256_sub_ps(iz2,jz0);
 598             dx30             = _mm256_sub_ps(ix3,jx0);
 599             dy30             = _mm256_sub_ps(iy3,jy0);
 600             dz30             = _mm256_sub_ps(iz3,jz0);
 601
 602             /* Calculate squared distance and things based on it */
 603             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 604             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 605             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 606             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
 607
 608             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 609             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 610             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 611             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
 612
 613             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 614             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 615             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 616             rinvsq30         = _mm256_mul_ps(rinv30,rinv30);
 617
 618             /* Load parameters for j particles */
 619             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 620                                                                  charge+jnrC+0,charge+jnrD+0,
 621                                                                  charge+jnrE+0,charge+jnrF+0,
 622                                                                  charge+jnrG+0,charge+jnrH+0);
 623             vdwjidx0A        = 2*vdwtype[jnrA+0];
 624             vdwjidx0B        = 2*vdwtype[jnrB+0];
 625             vdwjidx0C        = 2*vdwtype[jnrC+0];
 626             vdwjidx0D        = 2*vdwtype[jnrD+0];
 627             vdwjidx0E        = 2*vdwtype[jnrE+0];
 628             vdwjidx0F        = 2*vdwtype[jnrF+0];
 629             vdwjidx0G        = 2*vdwtype[jnrG+0];
 630             vdwjidx0H        = 2*vdwtype[jnrH+0];
 631
 632             fjx0             = _mm256_setzero_ps();
 633             fjy0             = _mm256_setzero_ps();
 634             fjz0             = _mm256_setzero_ps();
 635
 636             /**************************
 637              * CALCULATE INTERACTIONS *
 638              **************************/
 639
 640             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 641             {
 642
 643             r00              = _mm256_mul_ps(rsq00,rinv00);
 644             r00              = _mm256_andnot_ps(dummy_mask,r00);
 645
 646             /* Compute parameters for interactions between i and j atoms */
 647             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 648                                             vdwioffsetptr0+vdwjidx0B,
 649                                             vdwioffsetptr0+vdwjidx0C,
 650                                             vdwioffsetptr0+vdwjidx0D,
 651                                             vdwioffsetptr0+vdwjidx0E,
 652                                             vdwioffsetptr0+vdwjidx0F,
 653                                             vdwioffsetptr0+vdwjidx0G,
 654                                             vdwioffsetptr0+vdwjidx0H,
 655                                             &c6_00,&c12_00);
 656
 657             c6grid_00       = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
 658                                                                   vdwgridioffsetptr0+vdwjidx0B,
 659                                                                   vdwgridioffsetptr0+vdwjidx0C,
 660                                                                   vdwgridioffsetptr0+vdwjidx0D,
 661                                                                   vdwgridioffsetptr0+vdwjidx0E,
 662                                                                   vdwgridioffsetptr0+vdwjidx0F,
 663                                                                   vdwgridioffsetptr0+vdwjidx0G,
 664                                                                   vdwgridioffsetptr0+vdwjidx0H);
 665
 666             /* Analytical LJ-PME */
 667             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 668             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
 669             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
 670             exponent         = gmx_simd_exp_r(ewcljrsq);
 671             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 672             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
 673             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 674             vvdw6            = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
 675             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 676             vvdw             = _mm256_sub_ps(_mm256_mul_ps( _mm256_sub_ps(vvdw12 , _mm256_mul_ps(c12_00,_mm256_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 677                                           _mm256_mul_ps( _mm256_sub_ps(vvdw6,_mm256_add_ps(_mm256_mul_ps(c6_00,sh_vdw_invrcut6),_mm256_mul_ps(c6grid_00,sh_lj_ewald))),one_sixth));
 678             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 679             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
 680
 681             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 682
 683             /* Update potential sum for this i atom from the interaction with this j atom. */
 684             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 685             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 686             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 687
 688             fscal            = fvdw;
 689
 690             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 691
 692             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 693
 694             /* Calculate temporary vectorial force */
 695             tx               = _mm256_mul_ps(fscal,dx00);
 696             ty               = _mm256_mul_ps(fscal,dy00);
 697             tz               = _mm256_mul_ps(fscal,dz00);
 698
 699             /* Update vectorial force */
 700             fix0             = _mm256_add_ps(fix0,tx);
 701             fiy0             = _mm256_add_ps(fiy0,ty);
 702             fiz0             = _mm256_add_ps(fiz0,tz);
 703
 704             fjx0             = _mm256_add_ps(fjx0,tx);
 705             fjy0             = _mm256_add_ps(fjy0,ty);
 706             fjz0             = _mm256_add_ps(fjz0,tz);
 707
 708             }
 709
 710             /**************************
 711              * CALCULATE INTERACTIONS *
 712              **************************/
 713
 714             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 715             {
 716
 717             r10              = _mm256_mul_ps(rsq10,rinv10);
 718             r10              = _mm256_andnot_ps(dummy_mask,r10);
 719
 720             /* Compute parameters for interactions between i and j atoms */
 721             qq10             = _mm256_mul_ps(iq1,jq0);
 722
 723             /* EWALD ELECTROSTATICS */
 724
 725             /* Analytical PME correction */
 726             zeta2            = _mm256_mul_ps(beta2,rsq10);
 727             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 728             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 729             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 730             felec            = _mm256_mul_ps(qq10,felec);
 731             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 732             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 733             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv10,sh_ewald),pmecorrV);
 734             velec            = _mm256_mul_ps(qq10,velec);
 735
 736             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 737
 738             /* Update potential sum for this i atom from the interaction with this j atom. */
 739             velec            = _mm256_and_ps(velec,cutoff_mask);
 740             velec            = _mm256_andnot_ps(dummy_mask,velec);
 741             velecsum         = _mm256_add_ps(velecsum,velec);
 742
 743             fscal            = felec;
 744
 745             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 746
 747             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 748
 749             /* Calculate temporary vectorial force */
 750             tx               = _mm256_mul_ps(fscal,dx10);
 751             ty               = _mm256_mul_ps(fscal,dy10);
 752             tz               = _mm256_mul_ps(fscal,dz10);
 753
 754             /* Update vectorial force */
 755             fix1             = _mm256_add_ps(fix1,tx);
 756             fiy1             = _mm256_add_ps(fiy1,ty);
 757             fiz1             = _mm256_add_ps(fiz1,tz);
 758
 759             fjx0             = _mm256_add_ps(fjx0,tx);
 760             fjy0             = _mm256_add_ps(fjy0,ty);
 761             fjz0             = _mm256_add_ps(fjz0,tz);
 762
 763             }
 764
 765             /**************************
 766              * CALCULATE INTERACTIONS *
 767              **************************/
 768
 769             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 770             {
 771
 772             r20              = _mm256_mul_ps(rsq20,rinv20);
 773             r20              = _mm256_andnot_ps(dummy_mask,r20);
 774
 775             /* Compute parameters for interactions between i and j atoms */
 776             qq20             = _mm256_mul_ps(iq2,jq0);
 777
 778             /* EWALD ELECTROSTATICS */
 779
 780             /* Analytical PME correction */
 781             zeta2            = _mm256_mul_ps(beta2,rsq20);
 782             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
 783             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 784             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 785             felec            = _mm256_mul_ps(qq20,felec);
 786             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 787             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 788             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv20,sh_ewald),pmecorrV);
 789             velec            = _mm256_mul_ps(qq20,velec);
 790
 791             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 792
 793             /* Update potential sum for this i atom from the interaction with this j atom. */
 794             velec            = _mm256_and_ps(velec,cutoff_mask);
 795             velec            = _mm256_andnot_ps(dummy_mask,velec);
 796             velecsum         = _mm256_add_ps(velecsum,velec);
 797
 798             fscal            = felec;
 799
 800             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 801
 802             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 803
 804             /* Calculate temporary vectorial force */
 805             tx               = _mm256_mul_ps(fscal,dx20);
 806             ty               = _mm256_mul_ps(fscal,dy20);
 807             tz               = _mm256_mul_ps(fscal,dz20);
 808
 809             /* Update vectorial force */
 810             fix2             = _mm256_add_ps(fix2,tx);
 811             fiy2             = _mm256_add_ps(fiy2,ty);
 812             fiz2             = _mm256_add_ps(fiz2,tz);
 813
 814             fjx0             = _mm256_add_ps(fjx0,tx);
 815             fjy0             = _mm256_add_ps(fjy0,ty);
 816             fjz0             = _mm256_add_ps(fjz0,tz);
 817
 818             }
 819
 820             /**************************
 821              * CALCULATE INTERACTIONS *
 822              **************************/
 823
 824             if (gmx_mm256_any_lt(rsq30,rcutoff2))
 825             {
 826
 827             r30              = _mm256_mul_ps(rsq30,rinv30);
 828             r30              = _mm256_andnot_ps(dummy_mask,r30);
 829
 830             /* Compute parameters for interactions between i and j atoms */
 831             qq30             = _mm256_mul_ps(iq3,jq0);
 832
 833             /* EWALD ELECTROSTATICS */
 834
 835             /* Analytical PME correction */
 836             zeta2            = _mm256_mul_ps(beta2,rsq30);
 837             rinv3            = _mm256_mul_ps(rinvsq30,rinv30);
 838             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 839             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 840             felec            = _mm256_mul_ps(qq30,felec);
 841             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 842             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 843             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv30,sh_ewald),pmecorrV);
 844             velec            = _mm256_mul_ps(qq30,velec);
 845
 846             cutoff_mask      = _mm256_cmp_ps(rsq30,rcutoff2,_CMP_LT_OQ);
 847
 848             /* Update potential sum for this i atom from the interaction with this j atom. */
 849             velec            = _mm256_and_ps(velec,cutoff_mask);
 850             velec            = _mm256_andnot_ps(dummy_mask,velec);
 851             velecsum         = _mm256_add_ps(velecsum,velec);
 852
 853             fscal            = felec;
 854
 855             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 856
 857             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 858
 859             /* Calculate temporary vectorial force */
 860             tx               = _mm256_mul_ps(fscal,dx30);
 861             ty               = _mm256_mul_ps(fscal,dy30);
 862             tz               = _mm256_mul_ps(fscal,dz30);
 863
 864             /* Update vectorial force */
 865             fix3             = _mm256_add_ps(fix3,tx);
 866             fiy3             = _mm256_add_ps(fiy3,ty);
 867             fiz3             = _mm256_add_ps(fiz3,tz);
 868
 869             fjx0             = _mm256_add_ps(fjx0,tx);
 870             fjy0             = _mm256_add_ps(fjy0,ty);
 871             fjz0             = _mm256_add_ps(fjz0,tz);
 872
 873             }
 874
 875             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 876             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 877             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 878             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 879             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 880             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 881             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 882             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 883
 884             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 885
 886             /* Inner loop uses 396 flops */
 887         }
 888
 889         /* End of innermost loop */
 890
 891         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 892                                                  f+i_coord_offset,fshift+i_shift_offset);
 893
 894         ggid                        = gid[iidx];
 895         /* Update potential energies */
 896         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 897         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 898
 899         /* Increment number of inner iterations */
 900         inneriter                  += j_index_end - j_index_start;
 901
 902         /* Outer loop uses 26 flops */
 903     }
 904
 905     /* Increment number of outer iterations */
 906     outeriter        += nri;
 907
 908     /* Update outer/inner flops */
 909
 910     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*396);
 911 }
 912 /*
 913  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_avx_256_single
 914  * Electrostatics interaction: Ewald
 915  * VdW interaction:            LJEwald
 916  * Geometry:                   Water4-Particle
 917  * Calculate force/pot:        Force
 918  */
 919 void
 920 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_avx_256_single
 921                     (t_nblist                    * gmx_restrict       nlist,
 922                      rvec                        * gmx_restrict          xx,
 923                      rvec                        * gmx_restrict          ff,
 924                      t_forcerec                  * gmx_restrict          fr,
 925                      t_mdatoms                   * gmx_restrict     mdatoms,
 926                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 927                      t_nrnb                      * gmx_restrict        nrnb)
 928 {
 929     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 930      * just 0 for non-waters.
 931      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 932      * jnr indices corresponding to data put in the four positions in the SIMD register.
 933      */
 934     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 935     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 936     int              jnrA,jnrB,jnrC,jnrD;
 937     int              jnrE,jnrF,jnrG,jnrH;
 938     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 939     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 940     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 941     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 942     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 943     real             rcutoff_scalar;
 944     real             *shiftvec,*fshift,*x,*f;
 945     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 946     real             scratch[4*DIM];
 947     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 948     real *           vdwioffsetptr0;
 949     real *           vdwgridioffsetptr0;
 950     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 951     real *           vdwioffsetptr1;
 952     real *           vdwgridioffsetptr1;
 953     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 954     real *           vdwioffsetptr2;
 955     real *           vdwgridioffsetptr2;
 956     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 957     real *           vdwioffsetptr3;
 958     real *           vdwgridioffsetptr3;
 959     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 960     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 961     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 962     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 963     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 964     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 965     __m256           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 966     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 967     real             *charge;
 968     int              nvdwtype;
 969     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 970     int              *vdwtype;
 971     real             *vdwparam;
 972     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 973     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 974     __m256           c6grid_00;
 975     __m256           c6grid_10;
 976     __m256           c6grid_20;
 977     __m256           c6grid_30;
 978     real             *vdwgridparam;
 979     __m256           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 980     __m256           one_half  = _mm256_set1_ps(0.5);
 981     __m256           minus_one = _mm256_set1_ps(-1.0);
 982     __m256i          ewitab;
 983     __m128i          ewitab_lo,ewitab_hi;
 984     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 985     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 986     real             *ewtab;
 987     __m256           dummy_mask,cutoff_mask;
 988     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 989     __m256           one     = _mm256_set1_ps(1.0);
 990     __m256           two     = _mm256_set1_ps(2.0);
 991     x                = xx[0];
 992     f                = ff[0];
 993
 994     nri              = nlist->nri;
 995     iinr             = nlist->iinr;
 996     jindex           = nlist->jindex;
 997     jjnr             = nlist->jjnr;
 998     shiftidx         = nlist->shift;
 999     gid              = nlist->gid;
1000     shiftvec         = fr->shift_vec[0];
1001     fshift           = fr->fshift[0];
1002     facel            = _mm256_set1_ps(fr->epsfac);
1003     charge           = mdatoms->chargeA;
1004     nvdwtype         = fr->ntype;
1005     vdwparam         = fr->nbfp;
1006     vdwtype          = mdatoms->typeA;
1007     vdwgridparam     = fr->ljpme_c6grid;
1008     sh_lj_ewald      = _mm256_set1_ps(fr->ic->sh_lj_ewald);
1009     ewclj            = _mm256_set1_ps(fr->ewaldcoeff_lj);
1010     ewclj2           = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
1011
1012     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
1013     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1014     beta2            = _mm256_mul_ps(beta,beta);
1015     beta3            = _mm256_mul_ps(beta,beta2);
1016
1017     ewtab            = fr->ic->tabq_coul_F;
1018     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
1019     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1020
1021     /* Setup water-specific parameters */
1022     inr              = nlist->iinr[0];
1023     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1024     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1025     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1026     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1027     vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1028
1029     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1030     rcutoff_scalar   = fr->rcoulomb;
1031     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
1032     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
1033
1034     sh_vdw_invrcut6  = _mm256_set1_ps(fr->ic->sh_invrc6);
1035     rvdw             = _mm256_set1_ps(fr->rvdw);
1036
1037     /* Avoid stupid compiler warnings */
1038     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1039     j_coord_offsetA = 0;
1040     j_coord_offsetB = 0;
1041     j_coord_offsetC = 0;
1042     j_coord_offsetD = 0;
1043     j_coord_offsetE = 0;
1044     j_coord_offsetF = 0;
1045     j_coord_offsetG = 0;
1046     j_coord_offsetH = 0;
1047
1048     outeriter        = 0;
1049     inneriter        = 0;
1050
1051     for(iidx=0;iidx<4*DIM;iidx++)
1052     {
1053         scratch[iidx] = 0.0;
1054     }
1055
1056     /* Start outer loop over neighborlists */
1057     for(iidx=0; iidx<nri; iidx++)
1058     {
1059         /* Load shift vector for this list */
1060         i_shift_offset   = DIM*shiftidx[iidx];
1061
1062         /* Load limits for loop over neighbors */
1063         j_index_start    = jindex[iidx];
1064         j_index_end      = jindex[iidx+1];
1065
1066         /* Get outer coordinate index */
1067         inr              = iinr[iidx];
1068         i_coord_offset   = DIM*inr;
1069
1070         /* Load i particle coords and add shift vector */
1071         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1072                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1073
1074         fix0             = _mm256_setzero_ps();
1075         fiy0             = _mm256_setzero_ps();
1076         fiz0             = _mm256_setzero_ps();
1077         fix1             = _mm256_setzero_ps();
1078         fiy1             = _mm256_setzero_ps();
1079         fiz1             = _mm256_setzero_ps();
1080         fix2             = _mm256_setzero_ps();
1081         fiy2             = _mm256_setzero_ps();
1082         fiz2             = _mm256_setzero_ps();
1083         fix3             = _mm256_setzero_ps();
1084         fiy3             = _mm256_setzero_ps();
1085         fiz3             = _mm256_setzero_ps();
1086
1087         /* Start inner kernel loop */
1088         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1089         {
1090
1091             /* Get j neighbor index, and coordinate index */
1092             jnrA             = jjnr[jidx];
1093             jnrB             = jjnr[jidx+1];
1094             jnrC             = jjnr[jidx+2];
1095             jnrD             = jjnr[jidx+3];
1096             jnrE             = jjnr[jidx+4];
1097             jnrF             = jjnr[jidx+5];
1098             jnrG             = jjnr[jidx+6];
1099             jnrH             = jjnr[jidx+7];
1100             j_coord_offsetA  = DIM*jnrA;
1101             j_coord_offsetB  = DIM*jnrB;
1102             j_coord_offsetC  = DIM*jnrC;
1103             j_coord_offsetD  = DIM*jnrD;
1104             j_coord_offsetE  = DIM*jnrE;
1105             j_coord_offsetF  = DIM*jnrF;
1106             j_coord_offsetG  = DIM*jnrG;
1107             j_coord_offsetH  = DIM*jnrH;
1108
1109             /* load j atom coordinates */
1110             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1111                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1112                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1113                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1114                                                  &jx0,&jy0,&jz0);
1115
1116             /* Calculate displacement vector */
1117             dx00             = _mm256_sub_ps(ix0,jx0);
1118             dy00             = _mm256_sub_ps(iy0,jy0);
1119             dz00             = _mm256_sub_ps(iz0,jz0);
1120             dx10             = _mm256_sub_ps(ix1,jx0);
1121             dy10             = _mm256_sub_ps(iy1,jy0);
1122             dz10             = _mm256_sub_ps(iz1,jz0);
1123             dx20             = _mm256_sub_ps(ix2,jx0);
1124             dy20             = _mm256_sub_ps(iy2,jy0);
1125             dz20             = _mm256_sub_ps(iz2,jz0);
1126             dx30             = _mm256_sub_ps(ix3,jx0);
1127             dy30             = _mm256_sub_ps(iy3,jy0);
1128             dz30             = _mm256_sub_ps(iz3,jz0);
1129
1130             /* Calculate squared distance and things based on it */
1131             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1132             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1133             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1134             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
1135
1136             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1137             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1138             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1139             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
1140
1141             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1142             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1143             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1144             rinvsq30         = _mm256_mul_ps(rinv30,rinv30);
1145
1146             /* Load parameters for j particles */
1147             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1148                                                                  charge+jnrC+0,charge+jnrD+0,
1149                                                                  charge+jnrE+0,charge+jnrF+0,
1150                                                                  charge+jnrG+0,charge+jnrH+0);
1151             vdwjidx0A        = 2*vdwtype[jnrA+0];
1152             vdwjidx0B        = 2*vdwtype[jnrB+0];
1153             vdwjidx0C        = 2*vdwtype[jnrC+0];
1154             vdwjidx0D        = 2*vdwtype[jnrD+0];
1155             vdwjidx0E        = 2*vdwtype[jnrE+0];
1156             vdwjidx0F        = 2*vdwtype[jnrF+0];
1157             vdwjidx0G        = 2*vdwtype[jnrG+0];
1158             vdwjidx0H        = 2*vdwtype[jnrH+0];
1159
1160             fjx0             = _mm256_setzero_ps();
1161             fjy0             = _mm256_setzero_ps();
1162             fjz0             = _mm256_setzero_ps();
1163
1164             /**************************
1165              * CALCULATE INTERACTIONS *
1166              **************************/
1167
1168             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1169             {
1170
1171             r00              = _mm256_mul_ps(rsq00,rinv00);
1172
1173             /* Compute parameters for interactions between i and j atoms */
1174             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1175                                             vdwioffsetptr0+vdwjidx0B,
1176                                             vdwioffsetptr0+vdwjidx0C,
1177                                             vdwioffsetptr0+vdwjidx0D,
1178                                             vdwioffsetptr0+vdwjidx0E,
1179                                             vdwioffsetptr0+vdwjidx0F,
1180                                             vdwioffsetptr0+vdwjidx0G,
1181                                             vdwioffsetptr0+vdwjidx0H,
1182                                             &c6_00,&c12_00);
1183
1184             c6grid_00       = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
1185                                                                   vdwgridioffsetptr0+vdwjidx0B,
1186                                                                   vdwgridioffsetptr0+vdwjidx0C,
1187                                                                   vdwgridioffsetptr0+vdwjidx0D,
1188                                                                   vdwgridioffsetptr0+vdwjidx0E,
1189                                                                   vdwgridioffsetptr0+vdwjidx0F,
1190                                                                   vdwgridioffsetptr0+vdwjidx0G,
1191                                                                   vdwgridioffsetptr0+vdwjidx0H);
1192
1193             /* Analytical LJ-PME */
1194             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1195             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
1196             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1197             exponent         = gmx_simd_exp_r(ewcljrsq);
1198             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1199             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1200             /* f6A = 6 * C6grid * (1 - poly) */
1201             f6A              = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1202             /* f6B = C6grid * exponent * beta^6 */
1203             f6B              = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
1204             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1205             fvdw              = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1206
1207             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1208
1209             fscal            = fvdw;
1210
1211             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1212
1213             /* Calculate temporary vectorial force */
1214             tx               = _mm256_mul_ps(fscal,dx00);
1215             ty               = _mm256_mul_ps(fscal,dy00);
1216             tz               = _mm256_mul_ps(fscal,dz00);
1217
1218             /* Update vectorial force */
1219             fix0             = _mm256_add_ps(fix0,tx);
1220             fiy0             = _mm256_add_ps(fiy0,ty);
1221             fiz0             = _mm256_add_ps(fiz0,tz);
1222
1223             fjx0             = _mm256_add_ps(fjx0,tx);
1224             fjy0             = _mm256_add_ps(fjy0,ty);
1225             fjz0             = _mm256_add_ps(fjz0,tz);
1226
1227             }
1228
1229             /**************************
1230              * CALCULATE INTERACTIONS *
1231              **************************/
1232
1233             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1234             {
1235
1236             r10              = _mm256_mul_ps(rsq10,rinv10);
1237
1238             /* Compute parameters for interactions between i and j atoms */
1239             qq10             = _mm256_mul_ps(iq1,jq0);
1240
1241             /* EWALD ELECTROSTATICS */
1242
1243             /* Analytical PME correction */
1244             zeta2            = _mm256_mul_ps(beta2,rsq10);
1245             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
1246             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1247             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1248             felec            = _mm256_mul_ps(qq10,felec);
1249
1250             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1251
1252             fscal            = felec;
1253
1254             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1255
1256             /* Calculate temporary vectorial force */
1257             tx               = _mm256_mul_ps(fscal,dx10);
1258             ty               = _mm256_mul_ps(fscal,dy10);
1259             tz               = _mm256_mul_ps(fscal,dz10);
1260
1261             /* Update vectorial force */
1262             fix1             = _mm256_add_ps(fix1,tx);
1263             fiy1             = _mm256_add_ps(fiy1,ty);
1264             fiz1             = _mm256_add_ps(fiz1,tz);
1265
1266             fjx0             = _mm256_add_ps(fjx0,tx);
1267             fjy0             = _mm256_add_ps(fjy0,ty);
1268             fjz0             = _mm256_add_ps(fjz0,tz);
1269
1270             }
1271
1272             /**************************
1273              * CALCULATE INTERACTIONS *
1274              **************************/
1275
1276             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1277             {
1278
1279             r20              = _mm256_mul_ps(rsq20,rinv20);
1280
1281             /* Compute parameters for interactions between i and j atoms */
1282             qq20             = _mm256_mul_ps(iq2,jq0);
1283
1284             /* EWALD ELECTROSTATICS */
1285
1286             /* Analytical PME correction */
1287             zeta2            = _mm256_mul_ps(beta2,rsq20);
1288             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1289             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1290             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1291             felec            = _mm256_mul_ps(qq20,felec);
1292
1293             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1294
1295             fscal            = felec;
1296
1297             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1298
1299             /* Calculate temporary vectorial force */
1300             tx               = _mm256_mul_ps(fscal,dx20);
1301             ty               = _mm256_mul_ps(fscal,dy20);
1302             tz               = _mm256_mul_ps(fscal,dz20);
1303
1304             /* Update vectorial force */
1305             fix2             = _mm256_add_ps(fix2,tx);
1306             fiy2             = _mm256_add_ps(fiy2,ty);
1307             fiz2             = _mm256_add_ps(fiz2,tz);
1308
1309             fjx0             = _mm256_add_ps(fjx0,tx);
1310             fjy0             = _mm256_add_ps(fjy0,ty);
1311             fjz0             = _mm256_add_ps(fjz0,tz);
1312
1313             }
1314
1315             /**************************
1316              * CALCULATE INTERACTIONS *
1317              **************************/
1318
1319             if (gmx_mm256_any_lt(rsq30,rcutoff2))
1320             {
1321
1322             r30              = _mm256_mul_ps(rsq30,rinv30);
1323
1324             /* Compute parameters for interactions between i and j atoms */
1325             qq30             = _mm256_mul_ps(iq3,jq0);
1326
1327             /* EWALD ELECTROSTATICS */
1328
1329             /* Analytical PME correction */
1330             zeta2            = _mm256_mul_ps(beta2,rsq30);
1331             rinv3            = _mm256_mul_ps(rinvsq30,rinv30);
1332             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1333             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1334             felec            = _mm256_mul_ps(qq30,felec);
1335
1336             cutoff_mask      = _mm256_cmp_ps(rsq30,rcutoff2,_CMP_LT_OQ);
1337
1338             fscal            = felec;
1339
1340             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1341
1342             /* Calculate temporary vectorial force */
1343             tx               = _mm256_mul_ps(fscal,dx30);
1344             ty               = _mm256_mul_ps(fscal,dy30);
1345             tz               = _mm256_mul_ps(fscal,dz30);
1346
1347             /* Update vectorial force */
1348             fix3             = _mm256_add_ps(fix3,tx);
1349             fiy3             = _mm256_add_ps(fiy3,ty);
1350             fiz3             = _mm256_add_ps(fiz3,tz);
1351
1352             fjx0             = _mm256_add_ps(fjx0,tx);
1353             fjy0             = _mm256_add_ps(fjy0,ty);
1354             fjz0             = _mm256_add_ps(fjz0,tz);
1355
1356             }
1357
1358             fjptrA             = f+j_coord_offsetA;
1359             fjptrB             = f+j_coord_offsetB;
1360             fjptrC             = f+j_coord_offsetC;
1361             fjptrD             = f+j_coord_offsetD;
1362             fjptrE             = f+j_coord_offsetE;
1363             fjptrF             = f+j_coord_offsetF;
1364             fjptrG             = f+j_coord_offsetG;
1365             fjptrH             = f+j_coord_offsetH;
1366
1367             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1368
1369             /* Inner loop uses 229 flops */
1370         }
1371
1372         if(jidx<j_index_end)
1373         {
1374
1375             /* Get j neighbor index, and coordinate index */
1376             jnrlistA         = jjnr[jidx];
1377             jnrlistB         = jjnr[jidx+1];
1378             jnrlistC         = jjnr[jidx+2];
1379             jnrlistD         = jjnr[jidx+3];
1380             jnrlistE         = jjnr[jidx+4];
1381             jnrlistF         = jjnr[jidx+5];
1382             jnrlistG         = jjnr[jidx+6];
1383             jnrlistH         = jjnr[jidx+7];
1384             /* Sign of each element will be negative for non-real atoms.
1385              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1386              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1387              */
1388             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1389                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1390
1391             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1392             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1393             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1394             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1395             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1396             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1397             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1398             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1399             j_coord_offsetA  = DIM*jnrA;
1400             j_coord_offsetB  = DIM*jnrB;
1401             j_coord_offsetC  = DIM*jnrC;
1402             j_coord_offsetD  = DIM*jnrD;
1403             j_coord_offsetE  = DIM*jnrE;
1404             j_coord_offsetF  = DIM*jnrF;
1405             j_coord_offsetG  = DIM*jnrG;
1406             j_coord_offsetH  = DIM*jnrH;
1407
1408             /* load j atom coordinates */
1409             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1410                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1411                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1412                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1413                                                  &jx0,&jy0,&jz0);
1414
1415             /* Calculate displacement vector */
1416             dx00             = _mm256_sub_ps(ix0,jx0);
1417             dy00             = _mm256_sub_ps(iy0,jy0);
1418             dz00             = _mm256_sub_ps(iz0,jz0);
1419             dx10             = _mm256_sub_ps(ix1,jx0);
1420             dy10             = _mm256_sub_ps(iy1,jy0);
1421             dz10             = _mm256_sub_ps(iz1,jz0);
1422             dx20             = _mm256_sub_ps(ix2,jx0);
1423             dy20             = _mm256_sub_ps(iy2,jy0);
1424             dz20             = _mm256_sub_ps(iz2,jz0);
1425             dx30             = _mm256_sub_ps(ix3,jx0);
1426             dy30             = _mm256_sub_ps(iy3,jy0);
1427             dz30             = _mm256_sub_ps(iz3,jz0);
1428
1429             /* Calculate squared distance and things based on it */
1430             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1431             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1432             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1433             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
1434
1435             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1436             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1437             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1438             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
1439
1440             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1441             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1442             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1443             rinvsq30         = _mm256_mul_ps(rinv30,rinv30);
1444
1445             /* Load parameters for j particles */
1446             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1447                                                                  charge+jnrC+0,charge+jnrD+0,
1448                                                                  charge+jnrE+0,charge+jnrF+0,
1449                                                                  charge+jnrG+0,charge+jnrH+0);
1450             vdwjidx0A        = 2*vdwtype[jnrA+0];
1451             vdwjidx0B        = 2*vdwtype[jnrB+0];
1452             vdwjidx0C        = 2*vdwtype[jnrC+0];
1453             vdwjidx0D        = 2*vdwtype[jnrD+0];
1454             vdwjidx0E        = 2*vdwtype[jnrE+0];
1455             vdwjidx0F        = 2*vdwtype[jnrF+0];
1456             vdwjidx0G        = 2*vdwtype[jnrG+0];
1457             vdwjidx0H        = 2*vdwtype[jnrH+0];
1458
1459             fjx0             = _mm256_setzero_ps();
1460             fjy0             = _mm256_setzero_ps();
1461             fjz0             = _mm256_setzero_ps();
1462
1463             /**************************
1464              * CALCULATE INTERACTIONS *
1465              **************************/
1466
1467             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1468             {
1469
1470             r00              = _mm256_mul_ps(rsq00,rinv00);
1471             r00              = _mm256_andnot_ps(dummy_mask,r00);
1472
1473             /* Compute parameters for interactions between i and j atoms */
1474             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1475                                             vdwioffsetptr0+vdwjidx0B,
1476                                             vdwioffsetptr0+vdwjidx0C,
1477                                             vdwioffsetptr0+vdwjidx0D,
1478                                             vdwioffsetptr0+vdwjidx0E,
1479                                             vdwioffsetptr0+vdwjidx0F,
1480                                             vdwioffsetptr0+vdwjidx0G,
1481                                             vdwioffsetptr0+vdwjidx0H,
1482                                             &c6_00,&c12_00);
1483
1484             c6grid_00       = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr0+vdwjidx0A,
1485                                                                   vdwgridioffsetptr0+vdwjidx0B,
1486                                                                   vdwgridioffsetptr0+vdwjidx0C,
1487                                                                   vdwgridioffsetptr0+vdwjidx0D,
1488                                                                   vdwgridioffsetptr0+vdwjidx0E,
1489                                                                   vdwgridioffsetptr0+vdwjidx0F,
1490                                                                   vdwgridioffsetptr0+vdwjidx0G,
1491                                                                   vdwgridioffsetptr0+vdwjidx0H);
1492
1493             /* Analytical LJ-PME */
1494             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1495             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq00);
1496             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1497             exponent         = gmx_simd_exp_r(ewcljrsq);
1498             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1499             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1500             /* f6A = 6 * C6grid * (1 - poly) */
1501             f6A              = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1502             /* f6B = C6grid * exponent * beta^6 */
1503             f6B              = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
1504             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1505             fvdw              = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1506
1507             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1508
1509             fscal            = fvdw;
1510
1511             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1512
1513             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1514
1515             /* Calculate temporary vectorial force */
1516             tx               = _mm256_mul_ps(fscal,dx00);
1517             ty               = _mm256_mul_ps(fscal,dy00);
1518             tz               = _mm256_mul_ps(fscal,dz00);
1519
1520             /* Update vectorial force */
1521             fix0             = _mm256_add_ps(fix0,tx);
1522             fiy0             = _mm256_add_ps(fiy0,ty);
1523             fiz0             = _mm256_add_ps(fiz0,tz);
1524
1525             fjx0             = _mm256_add_ps(fjx0,tx);
1526             fjy0             = _mm256_add_ps(fjy0,ty);
1527             fjz0             = _mm256_add_ps(fjz0,tz);
1528
1529             }
1530
1531             /**************************
1532              * CALCULATE INTERACTIONS *
1533              **************************/
1534
1535             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1536             {
1537
1538             r10              = _mm256_mul_ps(rsq10,rinv10);
1539             r10              = _mm256_andnot_ps(dummy_mask,r10);
1540
1541             /* Compute parameters for interactions between i and j atoms */
1542             qq10             = _mm256_mul_ps(iq1,jq0);
1543
1544             /* EWALD ELECTROSTATICS */
1545
1546             /* Analytical PME correction */
1547             zeta2            = _mm256_mul_ps(beta2,rsq10);
1548             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
1549             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1550             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1551             felec            = _mm256_mul_ps(qq10,felec);
1552
1553             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1554
1555             fscal            = felec;
1556
1557             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1558
1559             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1560
1561             /* Calculate temporary vectorial force */
1562             tx               = _mm256_mul_ps(fscal,dx10);
1563             ty               = _mm256_mul_ps(fscal,dy10);
1564             tz               = _mm256_mul_ps(fscal,dz10);
1565
1566             /* Update vectorial force */
1567             fix1             = _mm256_add_ps(fix1,tx);
1568             fiy1             = _mm256_add_ps(fiy1,ty);
1569             fiz1             = _mm256_add_ps(fiz1,tz);
1570
1571             fjx0             = _mm256_add_ps(fjx0,tx);
1572             fjy0             = _mm256_add_ps(fjy0,ty);
1573             fjz0             = _mm256_add_ps(fjz0,tz);
1574
1575             }
1576
1577             /**************************
1578              * CALCULATE INTERACTIONS *
1579              **************************/
1580
1581             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1582             {
1583
1584             r20              = _mm256_mul_ps(rsq20,rinv20);
1585             r20              = _mm256_andnot_ps(dummy_mask,r20);
1586
1587             /* Compute parameters for interactions between i and j atoms */
1588             qq20             = _mm256_mul_ps(iq2,jq0);
1589
1590             /* EWALD ELECTROSTATICS */
1591
1592             /* Analytical PME correction */
1593             zeta2            = _mm256_mul_ps(beta2,rsq20);
1594             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1595             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1596             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1597             felec            = _mm256_mul_ps(qq20,felec);
1598
1599             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1600
1601             fscal            = felec;
1602
1603             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1604
1605             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1606
1607             /* Calculate temporary vectorial force */
1608             tx               = _mm256_mul_ps(fscal,dx20);
1609             ty               = _mm256_mul_ps(fscal,dy20);
1610             tz               = _mm256_mul_ps(fscal,dz20);
1611
1612             /* Update vectorial force */
1613             fix2             = _mm256_add_ps(fix2,tx);
1614             fiy2             = _mm256_add_ps(fiy2,ty);
1615             fiz2             = _mm256_add_ps(fiz2,tz);
1616
1617             fjx0             = _mm256_add_ps(fjx0,tx);
1618             fjy0             = _mm256_add_ps(fjy0,ty);
1619             fjz0             = _mm256_add_ps(fjz0,tz);
1620
1621             }
1622
1623             /**************************
1624              * CALCULATE INTERACTIONS *
1625              **************************/
1626
1627             if (gmx_mm256_any_lt(rsq30,rcutoff2))
1628             {
1629
1630             r30              = _mm256_mul_ps(rsq30,rinv30);
1631             r30              = _mm256_andnot_ps(dummy_mask,r30);
1632
1633             /* Compute parameters for interactions between i and j atoms */
1634             qq30             = _mm256_mul_ps(iq3,jq0);
1635
1636             /* EWALD ELECTROSTATICS */
1637
1638             /* Analytical PME correction */
1639             zeta2            = _mm256_mul_ps(beta2,rsq30);
1640             rinv3            = _mm256_mul_ps(rinvsq30,rinv30);
1641             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1642             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1643             felec            = _mm256_mul_ps(qq30,felec);
1644
1645             cutoff_mask      = _mm256_cmp_ps(rsq30,rcutoff2,_CMP_LT_OQ);
1646
1647             fscal            = felec;
1648
1649             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1650
1651             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1652
1653             /* Calculate temporary vectorial force */
1654             tx               = _mm256_mul_ps(fscal,dx30);
1655             ty               = _mm256_mul_ps(fscal,dy30);
1656             tz               = _mm256_mul_ps(fscal,dz30);
1657
1658             /* Update vectorial force */
1659             fix3             = _mm256_add_ps(fix3,tx);
1660             fiy3             = _mm256_add_ps(fiy3,ty);
1661             fiz3             = _mm256_add_ps(fiz3,tz);
1662
1663             fjx0             = _mm256_add_ps(fjx0,tx);
1664             fjy0             = _mm256_add_ps(fjy0,ty);
1665             fjz0             = _mm256_add_ps(fjz0,tz);
1666
1667             }
1668
1669             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1670             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1671             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1672             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1673             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1674             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1675             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1676             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1677
1678             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1679
1680             /* Inner loop uses 233 flops */
1681         }
1682
1683         /* End of innermost loop */
1684
1685         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1686                                                  f+i_coord_offset,fshift+i_shift_offset);
1687
1688         /* Increment number of inner iterations */
1689         inneriter                  += j_index_end - j_index_start;
1690
1691         /* Outer loop uses 24 flops */
1692     }
1693
1694     /* Increment number of outer iterations */
1695     outeriter        += nri;
1696
1697     /* Update outer/inner flops */
1698
1699     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*233);
1700 }