src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEw_VdwCSTab_GeomW3P1_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "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_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_avx_256_single
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_avx_256_single
  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,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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              jnrE,jnrF,jnrG,jnrH;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  83     real             scratch[4*DIM];
  84     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     real *           vdwioffsetptr0;
  86     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     real *           vdwioffsetptr1;
  88     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     real *           vdwioffsetptr2;
  90     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  92     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  94     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  95     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  96     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     int              nvdwtype;
  99     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 100     int              *vdwtype;
 101     real             *vdwparam;
 102     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 103     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 104     __m256i          vfitab;
 105     __m128i          vfitab_lo,vfitab_hi;
 106     __m128i          ifour       = _mm_set1_epi32(4);
 107     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 108     real             *vftab;
 109     __m256i          ewitab;
 110     __m128i          ewitab_lo,ewitab_hi;
 111     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 112     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 113     real             *ewtab;
 114     __m256           dummy_mask,cutoff_mask;
 115     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 116     __m256           one     = _mm256_set1_ps(1.0);
 117     __m256           two     = _mm256_set1_ps(2.0);
 118     x                = xx[0];
 119     f                = ff[0];
 120
 121     nri              = nlist->nri;
 122     iinr             = nlist->iinr;
 123     jindex           = nlist->jindex;
 124     jjnr             = nlist->jjnr;
 125     shiftidx         = nlist->shift;
 126     gid              = nlist->gid;
 127     shiftvec         = fr->shift_vec[0];
 128     fshift           = fr->fshift[0];
 129     facel            = _mm256_set1_ps(fr->ic->epsfac);
 130     charge           = mdatoms->chargeA;
 131     nvdwtype         = fr->ntype;
 132     vdwparam         = fr->nbfp;
 133     vdwtype          = mdatoms->typeA;
 134
 135     vftab            = kernel_data->table_vdw->data;
 136     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 137
 138     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 139     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 140     beta2            = _mm256_mul_ps(beta,beta);
 141     beta3            = _mm256_mul_ps(beta,beta2);
 142
 143     ewtab            = fr->ic->tabq_coul_FDV0;
 144     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 145     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 146
 147     /* Setup water-specific parameters */
 148     inr              = nlist->iinr[0];
 149     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 150     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 151     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 152     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158     j_coord_offsetC = 0;
 159     j_coord_offsetD = 0;
 160     j_coord_offsetE = 0;
 161     j_coord_offsetF = 0;
 162     j_coord_offsetG = 0;
 163     j_coord_offsetH = 0;
 164
 165     outeriter        = 0;
 166     inneriter        = 0;
 167
 168     for(iidx=0;iidx<4*DIM;iidx++)
 169     {
 170         scratch[iidx] = 0.0;
 171     }
 172
 173     /* Start outer loop over neighborlists */
 174     for(iidx=0; iidx<nri; iidx++)
 175     {
 176         /* Load shift vector for this list */
 177         i_shift_offset   = DIM*shiftidx[iidx];
 178
 179         /* Load limits for loop over neighbors */
 180         j_index_start    = jindex[iidx];
 181         j_index_end      = jindex[iidx+1];
 182
 183         /* Get outer coordinate index */
 184         inr              = iinr[iidx];
 185         i_coord_offset   = DIM*inr;
 186
 187         /* Load i particle coords and add shift vector */
 188         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 189                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 190
 191         fix0             = _mm256_setzero_ps();
 192         fiy0             = _mm256_setzero_ps();
 193         fiz0             = _mm256_setzero_ps();
 194         fix1             = _mm256_setzero_ps();
 195         fiy1             = _mm256_setzero_ps();
 196         fiz1             = _mm256_setzero_ps();
 197         fix2             = _mm256_setzero_ps();
 198         fiy2             = _mm256_setzero_ps();
 199         fiz2             = _mm256_setzero_ps();
 200
 201         /* Reset potential sums */
 202         velecsum         = _mm256_setzero_ps();
 203         vvdwsum          = _mm256_setzero_ps();
 204
 205         /* Start inner kernel loop */
 206         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 207         {
 208
 209             /* Get j neighbor index, and coordinate index */
 210             jnrA             = jjnr[jidx];
 211             jnrB             = jjnr[jidx+1];
 212             jnrC             = jjnr[jidx+2];
 213             jnrD             = jjnr[jidx+3];
 214             jnrE             = jjnr[jidx+4];
 215             jnrF             = jjnr[jidx+5];
 216             jnrG             = jjnr[jidx+6];
 217             jnrH             = jjnr[jidx+7];
 218             j_coord_offsetA  = DIM*jnrA;
 219             j_coord_offsetB  = DIM*jnrB;
 220             j_coord_offsetC  = DIM*jnrC;
 221             j_coord_offsetD  = DIM*jnrD;
 222             j_coord_offsetE  = DIM*jnrE;
 223             j_coord_offsetF  = DIM*jnrF;
 224             j_coord_offsetG  = DIM*jnrG;
 225             j_coord_offsetH  = DIM*jnrH;
 226
 227             /* load j atom coordinates */
 228             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 229                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 230                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 231                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 232                                                  &jx0,&jy0,&jz0);
 233
 234             /* Calculate displacement vector */
 235             dx00             = _mm256_sub_ps(ix0,jx0);
 236             dy00             = _mm256_sub_ps(iy0,jy0);
 237             dz00             = _mm256_sub_ps(iz0,jz0);
 238             dx10             = _mm256_sub_ps(ix1,jx0);
 239             dy10             = _mm256_sub_ps(iy1,jy0);
 240             dz10             = _mm256_sub_ps(iz1,jz0);
 241             dx20             = _mm256_sub_ps(ix2,jx0);
 242             dy20             = _mm256_sub_ps(iy2,jy0);
 243             dz20             = _mm256_sub_ps(iz2,jz0);
 244
 245             /* Calculate squared distance and things based on it */
 246             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 247             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 248             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 249
 250             rinv00           = avx256_invsqrt_f(rsq00);
 251             rinv10           = avx256_invsqrt_f(rsq10);
 252             rinv20           = avx256_invsqrt_f(rsq20);
 253
 254             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 255             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 256             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 257
 258             /* Load parameters for j particles */
 259             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 260                                                                  charge+jnrC+0,charge+jnrD+0,
 261                                                                  charge+jnrE+0,charge+jnrF+0,
 262                                                                  charge+jnrG+0,charge+jnrH+0);
 263             vdwjidx0A        = 2*vdwtype[jnrA+0];
 264             vdwjidx0B        = 2*vdwtype[jnrB+0];
 265             vdwjidx0C        = 2*vdwtype[jnrC+0];
 266             vdwjidx0D        = 2*vdwtype[jnrD+0];
 267             vdwjidx0E        = 2*vdwtype[jnrE+0];
 268             vdwjidx0F        = 2*vdwtype[jnrF+0];
 269             vdwjidx0G        = 2*vdwtype[jnrG+0];
 270             vdwjidx0H        = 2*vdwtype[jnrH+0];
 271
 272             fjx0             = _mm256_setzero_ps();
 273             fjy0             = _mm256_setzero_ps();
 274             fjz0             = _mm256_setzero_ps();
 275
 276             /**************************
 277              * CALCULATE INTERACTIONS *
 278              **************************/
 279
 280             r00              = _mm256_mul_ps(rsq00,rinv00);
 281
 282             /* Compute parameters for interactions between i and j atoms */
 283             qq00             = _mm256_mul_ps(iq0,jq0);
 284             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 285                                             vdwioffsetptr0+vdwjidx0B,
 286                                             vdwioffsetptr0+vdwjidx0C,
 287                                             vdwioffsetptr0+vdwjidx0D,
 288                                             vdwioffsetptr0+vdwjidx0E,
 289                                             vdwioffsetptr0+vdwjidx0F,
 290                                             vdwioffsetptr0+vdwjidx0G,
 291                                             vdwioffsetptr0+vdwjidx0H,
 292                                             &c6_00,&c12_00);
 293
 294             /* Calculate table index by multiplying r with table scale and truncate to integer */
 295             rt               = _mm256_mul_ps(r00,vftabscale);
 296             vfitab           = _mm256_cvttps_epi32(rt);
 297             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 298             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 299             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 300             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 301             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 302             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 303
 304             /* EWALD ELECTROSTATICS */
 305
 306             /* Analytical PME correction */
 307             zeta2            = _mm256_mul_ps(beta2,rsq00);
 308             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
 309             pmecorrF         = avx256_pmecorrF_f(zeta2);
 310             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 311             felec            = _mm256_mul_ps(qq00,felec);
 312             pmecorrV         = avx256_pmecorrV_f(zeta2);
 313             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 314             velec            = _mm256_sub_ps(rinv00,pmecorrV);
 315             velec            = _mm256_mul_ps(qq00,velec);
 316
 317             /* CUBIC SPLINE TABLE DISPERSION */
 318             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 319                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 320             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 321                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 322             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 323                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 324             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 325                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 326             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 327             Heps             = _mm256_mul_ps(vfeps,H);
 328             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 329             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 330             vvdw6            = _mm256_mul_ps(c6_00,VV);
 331             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 332             fvdw6            = _mm256_mul_ps(c6_00,FF);
 333
 334             /* CUBIC SPLINE TABLE REPULSION */
 335             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 336             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 337             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 338                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 339             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 340                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 341             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 342                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 343             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 344                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 345             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 346             Heps             = _mm256_mul_ps(vfeps,H);
 347             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 348             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 349             vvdw12           = _mm256_mul_ps(c12_00,VV);
 350             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 351             fvdw12           = _mm256_mul_ps(c12_00,FF);
 352             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 353             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velecsum         = _mm256_add_ps(velecsum,velec);
 357             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 358
 359             fscal            = _mm256_add_ps(felec,fvdw);
 360
 361             /* Calculate temporary vectorial force */
 362             tx               = _mm256_mul_ps(fscal,dx00);
 363             ty               = _mm256_mul_ps(fscal,dy00);
 364             tz               = _mm256_mul_ps(fscal,dz00);
 365
 366             /* Update vectorial force */
 367             fix0             = _mm256_add_ps(fix0,tx);
 368             fiy0             = _mm256_add_ps(fiy0,ty);
 369             fiz0             = _mm256_add_ps(fiz0,tz);
 370
 371             fjx0             = _mm256_add_ps(fjx0,tx);
 372             fjy0             = _mm256_add_ps(fjy0,ty);
 373             fjz0             = _mm256_add_ps(fjz0,tz);
 374
 375             /**************************
 376              * CALCULATE INTERACTIONS *
 377              **************************/
 378
 379             r10              = _mm256_mul_ps(rsq10,rinv10);
 380
 381             /* Compute parameters for interactions between i and j atoms */
 382             qq10             = _mm256_mul_ps(iq1,jq0);
 383
 384             /* EWALD ELECTROSTATICS */
 385
 386             /* Analytical PME correction */
 387             zeta2            = _mm256_mul_ps(beta2,rsq10);
 388             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 389             pmecorrF         = avx256_pmecorrF_f(zeta2);
 390             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 391             felec            = _mm256_mul_ps(qq10,felec);
 392             pmecorrV         = avx256_pmecorrV_f(zeta2);
 393             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 394             velec            = _mm256_sub_ps(rinv10,pmecorrV);
 395             velec            = _mm256_mul_ps(qq10,velec);
 396
 397             /* Update potential sum for this i atom from the interaction with this j atom. */
 398             velecsum         = _mm256_add_ps(velecsum,velec);
 399
 400             fscal            = felec;
 401
 402             /* Calculate temporary vectorial force */
 403             tx               = _mm256_mul_ps(fscal,dx10);
 404             ty               = _mm256_mul_ps(fscal,dy10);
 405             tz               = _mm256_mul_ps(fscal,dz10);
 406
 407             /* Update vectorial force */
 408             fix1             = _mm256_add_ps(fix1,tx);
 409             fiy1             = _mm256_add_ps(fiy1,ty);
 410             fiz1             = _mm256_add_ps(fiz1,tz);
 411
 412             fjx0             = _mm256_add_ps(fjx0,tx);
 413             fjy0             = _mm256_add_ps(fjy0,ty);
 414             fjz0             = _mm256_add_ps(fjz0,tz);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             r20              = _mm256_mul_ps(rsq20,rinv20);
 421
 422             /* Compute parameters for interactions between i and j atoms */
 423             qq20             = _mm256_mul_ps(iq2,jq0);
 424
 425             /* EWALD ELECTROSTATICS */
 426
 427             /* Analytical PME correction */
 428             zeta2            = _mm256_mul_ps(beta2,rsq20);
 429             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
 430             pmecorrF         = avx256_pmecorrF_f(zeta2);
 431             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 432             felec            = _mm256_mul_ps(qq20,felec);
 433             pmecorrV         = avx256_pmecorrV_f(zeta2);
 434             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 435             velec            = _mm256_sub_ps(rinv20,pmecorrV);
 436             velec            = _mm256_mul_ps(qq20,velec);
 437
 438             /* Update potential sum for this i atom from the interaction with this j atom. */
 439             velecsum         = _mm256_add_ps(velecsum,velec);
 440
 441             fscal            = felec;
 442
 443             /* Calculate temporary vectorial force */
 444             tx               = _mm256_mul_ps(fscal,dx20);
 445             ty               = _mm256_mul_ps(fscal,dy20);
 446             tz               = _mm256_mul_ps(fscal,dz20);
 447
 448             /* Update vectorial force */
 449             fix2             = _mm256_add_ps(fix2,tx);
 450             fiy2             = _mm256_add_ps(fiy2,ty);
 451             fiz2             = _mm256_add_ps(fiz2,tz);
 452
 453             fjx0             = _mm256_add_ps(fjx0,tx);
 454             fjy0             = _mm256_add_ps(fjy0,ty);
 455             fjz0             = _mm256_add_ps(fjz0,tz);
 456
 457             fjptrA             = f+j_coord_offsetA;
 458             fjptrB             = f+j_coord_offsetB;
 459             fjptrC             = f+j_coord_offsetC;
 460             fjptrD             = f+j_coord_offsetD;
 461             fjptrE             = f+j_coord_offsetE;
 462             fjptrF             = f+j_coord_offsetF;
 463             fjptrG             = f+j_coord_offsetG;
 464             fjptrH             = f+j_coord_offsetH;
 465
 466             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 467
 468             /* Inner loop uses 289 flops */
 469         }
 470
 471         if(jidx<j_index_end)
 472         {
 473
 474             /* Get j neighbor index, and coordinate index */
 475             jnrlistA         = jjnr[jidx];
 476             jnrlistB         = jjnr[jidx+1];
 477             jnrlistC         = jjnr[jidx+2];
 478             jnrlistD         = jjnr[jidx+3];
 479             jnrlistE         = jjnr[jidx+4];
 480             jnrlistF         = jjnr[jidx+5];
 481             jnrlistG         = jjnr[jidx+6];
 482             jnrlistH         = jjnr[jidx+7];
 483             /* Sign of each element will be negative for non-real atoms.
 484              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 485              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 486              */
 487             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 488                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 489
 490             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 491             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 492             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 493             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 494             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 495             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 496             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 497             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 498             j_coord_offsetA  = DIM*jnrA;
 499             j_coord_offsetB  = DIM*jnrB;
 500             j_coord_offsetC  = DIM*jnrC;
 501             j_coord_offsetD  = DIM*jnrD;
 502             j_coord_offsetE  = DIM*jnrE;
 503             j_coord_offsetF  = DIM*jnrF;
 504             j_coord_offsetG  = DIM*jnrG;
 505             j_coord_offsetH  = DIM*jnrH;
 506
 507             /* load j atom coordinates */
 508             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 509                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 510                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 511                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 512                                                  &jx0,&jy0,&jz0);
 513
 514             /* Calculate displacement vector */
 515             dx00             = _mm256_sub_ps(ix0,jx0);
 516             dy00             = _mm256_sub_ps(iy0,jy0);
 517             dz00             = _mm256_sub_ps(iz0,jz0);
 518             dx10             = _mm256_sub_ps(ix1,jx0);
 519             dy10             = _mm256_sub_ps(iy1,jy0);
 520             dz10             = _mm256_sub_ps(iz1,jz0);
 521             dx20             = _mm256_sub_ps(ix2,jx0);
 522             dy20             = _mm256_sub_ps(iy2,jy0);
 523             dz20             = _mm256_sub_ps(iz2,jz0);
 524
 525             /* Calculate squared distance and things based on it */
 526             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 527             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 528             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 529
 530             rinv00           = avx256_invsqrt_f(rsq00);
 531             rinv10           = avx256_invsqrt_f(rsq10);
 532             rinv20           = avx256_invsqrt_f(rsq20);
 533
 534             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 535             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 536             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 537
 538             /* Load parameters for j particles */
 539             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 540                                                                  charge+jnrC+0,charge+jnrD+0,
 541                                                                  charge+jnrE+0,charge+jnrF+0,
 542                                                                  charge+jnrG+0,charge+jnrH+0);
 543             vdwjidx0A        = 2*vdwtype[jnrA+0];
 544             vdwjidx0B        = 2*vdwtype[jnrB+0];
 545             vdwjidx0C        = 2*vdwtype[jnrC+0];
 546             vdwjidx0D        = 2*vdwtype[jnrD+0];
 547             vdwjidx0E        = 2*vdwtype[jnrE+0];
 548             vdwjidx0F        = 2*vdwtype[jnrF+0];
 549             vdwjidx0G        = 2*vdwtype[jnrG+0];
 550             vdwjidx0H        = 2*vdwtype[jnrH+0];
 551
 552             fjx0             = _mm256_setzero_ps();
 553             fjy0             = _mm256_setzero_ps();
 554             fjz0             = _mm256_setzero_ps();
 555
 556             /**************************
 557              * CALCULATE INTERACTIONS *
 558              **************************/
 559
 560             r00              = _mm256_mul_ps(rsq00,rinv00);
 561             r00              = _mm256_andnot_ps(dummy_mask,r00);
 562
 563             /* Compute parameters for interactions between i and j atoms */
 564             qq00             = _mm256_mul_ps(iq0,jq0);
 565             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 566                                             vdwioffsetptr0+vdwjidx0B,
 567                                             vdwioffsetptr0+vdwjidx0C,
 568                                             vdwioffsetptr0+vdwjidx0D,
 569                                             vdwioffsetptr0+vdwjidx0E,
 570                                             vdwioffsetptr0+vdwjidx0F,
 571                                             vdwioffsetptr0+vdwjidx0G,
 572                                             vdwioffsetptr0+vdwjidx0H,
 573                                             &c6_00,&c12_00);
 574
 575             /* Calculate table index by multiplying r with table scale and truncate to integer */
 576             rt               = _mm256_mul_ps(r00,vftabscale);
 577             vfitab           = _mm256_cvttps_epi32(rt);
 578             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 579             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 580             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 581             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 582             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 583             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 584
 585             /* EWALD ELECTROSTATICS */
 586
 587             /* Analytical PME correction */
 588             zeta2            = _mm256_mul_ps(beta2,rsq00);
 589             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
 590             pmecorrF         = avx256_pmecorrF_f(zeta2);
 591             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 592             felec            = _mm256_mul_ps(qq00,felec);
 593             pmecorrV         = avx256_pmecorrV_f(zeta2);
 594             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 595             velec            = _mm256_sub_ps(rinv00,pmecorrV);
 596             velec            = _mm256_mul_ps(qq00,velec);
 597
 598             /* CUBIC SPLINE TABLE DISPERSION */
 599             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 600                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 601             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 602                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 603             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 604                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 605             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 606                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 607             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 608             Heps             = _mm256_mul_ps(vfeps,H);
 609             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 610             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 611             vvdw6            = _mm256_mul_ps(c6_00,VV);
 612             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 613             fvdw6            = _mm256_mul_ps(c6_00,FF);
 614
 615             /* CUBIC SPLINE TABLE REPULSION */
 616             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 617             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 618             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 619                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 620             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 621                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 622             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 623                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 624             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 625                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 626             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 627             Heps             = _mm256_mul_ps(vfeps,H);
 628             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 629             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 630             vvdw12           = _mm256_mul_ps(c12_00,VV);
 631             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 632             fvdw12           = _mm256_mul_ps(c12_00,FF);
 633             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 634             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 635
 636             /* Update potential sum for this i atom from the interaction with this j atom. */
 637             velec            = _mm256_andnot_ps(dummy_mask,velec);
 638             velecsum         = _mm256_add_ps(velecsum,velec);
 639             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 640             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 641
 642             fscal            = _mm256_add_ps(felec,fvdw);
 643
 644             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 645
 646             /* Calculate temporary vectorial force */
 647             tx               = _mm256_mul_ps(fscal,dx00);
 648             ty               = _mm256_mul_ps(fscal,dy00);
 649             tz               = _mm256_mul_ps(fscal,dz00);
 650
 651             /* Update vectorial force */
 652             fix0             = _mm256_add_ps(fix0,tx);
 653             fiy0             = _mm256_add_ps(fiy0,ty);
 654             fiz0             = _mm256_add_ps(fiz0,tz);
 655
 656             fjx0             = _mm256_add_ps(fjx0,tx);
 657             fjy0             = _mm256_add_ps(fjy0,ty);
 658             fjz0             = _mm256_add_ps(fjz0,tz);
 659
 660             /**************************
 661              * CALCULATE INTERACTIONS *
 662              **************************/
 663
 664             r10              = _mm256_mul_ps(rsq10,rinv10);
 665             r10              = _mm256_andnot_ps(dummy_mask,r10);
 666
 667             /* Compute parameters for interactions between i and j atoms */
 668             qq10             = _mm256_mul_ps(iq1,jq0);
 669
 670             /* EWALD ELECTROSTATICS */
 671
 672             /* Analytical PME correction */
 673             zeta2            = _mm256_mul_ps(beta2,rsq10);
 674             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 675             pmecorrF         = avx256_pmecorrF_f(zeta2);
 676             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 677             felec            = _mm256_mul_ps(qq10,felec);
 678             pmecorrV         = avx256_pmecorrV_f(zeta2);
 679             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 680             velec            = _mm256_sub_ps(rinv10,pmecorrV);
 681             velec            = _mm256_mul_ps(qq10,velec);
 682
 683             /* Update potential sum for this i atom from the interaction with this j atom. */
 684             velec            = _mm256_andnot_ps(dummy_mask,velec);
 685             velecsum         = _mm256_add_ps(velecsum,velec);
 686
 687             fscal            = felec;
 688
 689             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 690
 691             /* Calculate temporary vectorial force */
 692             tx               = _mm256_mul_ps(fscal,dx10);
 693             ty               = _mm256_mul_ps(fscal,dy10);
 694             tz               = _mm256_mul_ps(fscal,dz10);
 695
 696             /* Update vectorial force */
 697             fix1             = _mm256_add_ps(fix1,tx);
 698             fiy1             = _mm256_add_ps(fiy1,ty);
 699             fiz1             = _mm256_add_ps(fiz1,tz);
 700
 701             fjx0             = _mm256_add_ps(fjx0,tx);
 702             fjy0             = _mm256_add_ps(fjy0,ty);
 703             fjz0             = _mm256_add_ps(fjz0,tz);
 704
 705             /**************************
 706              * CALCULATE INTERACTIONS *
 707              **************************/
 708
 709             r20              = _mm256_mul_ps(rsq20,rinv20);
 710             r20              = _mm256_andnot_ps(dummy_mask,r20);
 711
 712             /* Compute parameters for interactions between i and j atoms */
 713             qq20             = _mm256_mul_ps(iq2,jq0);
 714
 715             /* EWALD ELECTROSTATICS */
 716
 717             /* Analytical PME correction */
 718             zeta2            = _mm256_mul_ps(beta2,rsq20);
 719             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
 720             pmecorrF         = avx256_pmecorrF_f(zeta2);
 721             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 722             felec            = _mm256_mul_ps(qq20,felec);
 723             pmecorrV         = avx256_pmecorrV_f(zeta2);
 724             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 725             velec            = _mm256_sub_ps(rinv20,pmecorrV);
 726             velec            = _mm256_mul_ps(qq20,velec);
 727
 728             /* Update potential sum for this i atom from the interaction with this j atom. */
 729             velec            = _mm256_andnot_ps(dummy_mask,velec);
 730             velecsum         = _mm256_add_ps(velecsum,velec);
 731
 732             fscal            = felec;
 733
 734             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 735
 736             /* Calculate temporary vectorial force */
 737             tx               = _mm256_mul_ps(fscal,dx20);
 738             ty               = _mm256_mul_ps(fscal,dy20);
 739             tz               = _mm256_mul_ps(fscal,dz20);
 740
 741             /* Update vectorial force */
 742             fix2             = _mm256_add_ps(fix2,tx);
 743             fiy2             = _mm256_add_ps(fiy2,ty);
 744             fiz2             = _mm256_add_ps(fiz2,tz);
 745
 746             fjx0             = _mm256_add_ps(fjx0,tx);
 747             fjy0             = _mm256_add_ps(fjy0,ty);
 748             fjz0             = _mm256_add_ps(fjz0,tz);
 749
 750             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 751             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 752             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 753             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 754             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 755             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 756             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 757             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 758
 759             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 760
 761             /* Inner loop uses 292 flops */
 762         }
 763
 764         /* End of innermost loop */
 765
 766         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 767                                                  f+i_coord_offset,fshift+i_shift_offset);
 768
 769         ggid                        = gid[iidx];
 770         /* Update potential energies */
 771         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 772         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 773
 774         /* Increment number of inner iterations */
 775         inneriter                  += j_index_end - j_index_start;
 776
 777         /* Outer loop uses 20 flops */
 778     }
 779
 780     /* Increment number of outer iterations */
 781     outeriter        += nri;
 782
 783     /* Update outer/inner flops */
 784
 785     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*292);
 786 }
 787 /*
 788  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_avx_256_single
 789  * Electrostatics interaction: Ewald
 790  * VdW interaction:            CubicSplineTable
 791  * Geometry:                   Water3-Particle
 792  * Calculate force/pot:        Force
 793  */
 794 void
 795 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_avx_256_single
 796                     (t_nblist                    * gmx_restrict       nlist,
 797                      rvec                        * gmx_restrict          xx,
 798                      rvec                        * gmx_restrict          ff,
 799                      struct t_forcerec           * gmx_restrict          fr,
 800                      t_mdatoms                   * gmx_restrict     mdatoms,
 801                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 802                      t_nrnb                      * gmx_restrict        nrnb)
 803 {
 804     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 805      * just 0 for non-waters.
 806      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 807      * jnr indices corresponding to data put in the four positions in the SIMD register.
 808      */
 809     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 810     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 811     int              jnrA,jnrB,jnrC,jnrD;
 812     int              jnrE,jnrF,jnrG,jnrH;
 813     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 814     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 815     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 816     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 817     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 818     real             rcutoff_scalar;
 819     real             *shiftvec,*fshift,*x,*f;
 820     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 821     real             scratch[4*DIM];
 822     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 823     real *           vdwioffsetptr0;
 824     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 825     real *           vdwioffsetptr1;
 826     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 827     real *           vdwioffsetptr2;
 828     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 829     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 830     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 831     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 832     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 833     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 834     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 835     real             *charge;
 836     int              nvdwtype;
 837     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 838     int              *vdwtype;
 839     real             *vdwparam;
 840     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 841     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 842     __m256i          vfitab;
 843     __m128i          vfitab_lo,vfitab_hi;
 844     __m128i          ifour       = _mm_set1_epi32(4);
 845     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 846     real             *vftab;
 847     __m256i          ewitab;
 848     __m128i          ewitab_lo,ewitab_hi;
 849     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 850     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 851     real             *ewtab;
 852     __m256           dummy_mask,cutoff_mask;
 853     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 854     __m256           one     = _mm256_set1_ps(1.0);
 855     __m256           two     = _mm256_set1_ps(2.0);
 856     x                = xx[0];
 857     f                = ff[0];
 858
 859     nri              = nlist->nri;
 860     iinr             = nlist->iinr;
 861     jindex           = nlist->jindex;
 862     jjnr             = nlist->jjnr;
 863     shiftidx         = nlist->shift;
 864     gid              = nlist->gid;
 865     shiftvec         = fr->shift_vec[0];
 866     fshift           = fr->fshift[0];
 867     facel            = _mm256_set1_ps(fr->ic->epsfac);
 868     charge           = mdatoms->chargeA;
 869     nvdwtype         = fr->ntype;
 870     vdwparam         = fr->nbfp;
 871     vdwtype          = mdatoms->typeA;
 872
 873     vftab            = kernel_data->table_vdw->data;
 874     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 875
 876     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 877     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 878     beta2            = _mm256_mul_ps(beta,beta);
 879     beta3            = _mm256_mul_ps(beta,beta2);
 880
 881     ewtab            = fr->ic->tabq_coul_F;
 882     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 883     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 884
 885     /* Setup water-specific parameters */
 886     inr              = nlist->iinr[0];
 887     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 888     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 889     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 890     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 891
 892     /* Avoid stupid compiler warnings */
 893     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 894     j_coord_offsetA = 0;
 895     j_coord_offsetB = 0;
 896     j_coord_offsetC = 0;
 897     j_coord_offsetD = 0;
 898     j_coord_offsetE = 0;
 899     j_coord_offsetF = 0;
 900     j_coord_offsetG = 0;
 901     j_coord_offsetH = 0;
 902
 903     outeriter        = 0;
 904     inneriter        = 0;
 905
 906     for(iidx=0;iidx<4*DIM;iidx++)
 907     {
 908         scratch[iidx] = 0.0;
 909     }
 910
 911     /* Start outer loop over neighborlists */
 912     for(iidx=0; iidx<nri; iidx++)
 913     {
 914         /* Load shift vector for this list */
 915         i_shift_offset   = DIM*shiftidx[iidx];
 916
 917         /* Load limits for loop over neighbors */
 918         j_index_start    = jindex[iidx];
 919         j_index_end      = jindex[iidx+1];
 920
 921         /* Get outer coordinate index */
 922         inr              = iinr[iidx];
 923         i_coord_offset   = DIM*inr;
 924
 925         /* Load i particle coords and add shift vector */
 926         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 927                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 928
 929         fix0             = _mm256_setzero_ps();
 930         fiy0             = _mm256_setzero_ps();
 931         fiz0             = _mm256_setzero_ps();
 932         fix1             = _mm256_setzero_ps();
 933         fiy1             = _mm256_setzero_ps();
 934         fiz1             = _mm256_setzero_ps();
 935         fix2             = _mm256_setzero_ps();
 936         fiy2             = _mm256_setzero_ps();
 937         fiz2             = _mm256_setzero_ps();
 938
 939         /* Start inner kernel loop */
 940         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 941         {
 942
 943             /* Get j neighbor index, and coordinate index */
 944             jnrA             = jjnr[jidx];
 945             jnrB             = jjnr[jidx+1];
 946             jnrC             = jjnr[jidx+2];
 947             jnrD             = jjnr[jidx+3];
 948             jnrE             = jjnr[jidx+4];
 949             jnrF             = jjnr[jidx+5];
 950             jnrG             = jjnr[jidx+6];
 951             jnrH             = jjnr[jidx+7];
 952             j_coord_offsetA  = DIM*jnrA;
 953             j_coord_offsetB  = DIM*jnrB;
 954             j_coord_offsetC  = DIM*jnrC;
 955             j_coord_offsetD  = DIM*jnrD;
 956             j_coord_offsetE  = DIM*jnrE;
 957             j_coord_offsetF  = DIM*jnrF;
 958             j_coord_offsetG  = DIM*jnrG;
 959             j_coord_offsetH  = DIM*jnrH;
 960
 961             /* load j atom coordinates */
 962             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 963                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 964                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 965                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 966                                                  &jx0,&jy0,&jz0);
 967
 968             /* Calculate displacement vector */
 969             dx00             = _mm256_sub_ps(ix0,jx0);
 970             dy00             = _mm256_sub_ps(iy0,jy0);
 971             dz00             = _mm256_sub_ps(iz0,jz0);
 972             dx10             = _mm256_sub_ps(ix1,jx0);
 973             dy10             = _mm256_sub_ps(iy1,jy0);
 974             dz10             = _mm256_sub_ps(iz1,jz0);
 975             dx20             = _mm256_sub_ps(ix2,jx0);
 976             dy20             = _mm256_sub_ps(iy2,jy0);
 977             dz20             = _mm256_sub_ps(iz2,jz0);
 978
 979             /* Calculate squared distance and things based on it */
 980             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 981             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 982             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 983
 984             rinv00           = avx256_invsqrt_f(rsq00);
 985             rinv10           = avx256_invsqrt_f(rsq10);
 986             rinv20           = avx256_invsqrt_f(rsq20);
 987
 988             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 989             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 990             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 991
 992             /* Load parameters for j particles */
 993             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 994                                                                  charge+jnrC+0,charge+jnrD+0,
 995                                                                  charge+jnrE+0,charge+jnrF+0,
 996                                                                  charge+jnrG+0,charge+jnrH+0);
 997             vdwjidx0A        = 2*vdwtype[jnrA+0];
 998             vdwjidx0B        = 2*vdwtype[jnrB+0];
 999             vdwjidx0C        = 2*vdwtype[jnrC+0];
1000             vdwjidx0D        = 2*vdwtype[jnrD+0];
1001             vdwjidx0E        = 2*vdwtype[jnrE+0];
1002             vdwjidx0F        = 2*vdwtype[jnrF+0];
1003             vdwjidx0G        = 2*vdwtype[jnrG+0];
1004             vdwjidx0H        = 2*vdwtype[jnrH+0];
1005
1006             fjx0             = _mm256_setzero_ps();
1007             fjy0             = _mm256_setzero_ps();
1008             fjz0             = _mm256_setzero_ps();
1009
1010             /**************************
1011              * CALCULATE INTERACTIONS *
1012              **************************/
1013
1014             r00              = _mm256_mul_ps(rsq00,rinv00);
1015
1016             /* Compute parameters for interactions between i and j atoms */
1017             qq00             = _mm256_mul_ps(iq0,jq0);
1018             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1019                                             vdwioffsetptr0+vdwjidx0B,
1020                                             vdwioffsetptr0+vdwjidx0C,
1021                                             vdwioffsetptr0+vdwjidx0D,
1022                                             vdwioffsetptr0+vdwjidx0E,
1023                                             vdwioffsetptr0+vdwjidx0F,
1024                                             vdwioffsetptr0+vdwjidx0G,
1025                                             vdwioffsetptr0+vdwjidx0H,
1026                                             &c6_00,&c12_00);
1027
1028             /* Calculate table index by multiplying r with table scale and truncate to integer */
1029             rt               = _mm256_mul_ps(r00,vftabscale);
1030             vfitab           = _mm256_cvttps_epi32(rt);
1031             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1032             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1033             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1034             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1035             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1036             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1037
1038             /* EWALD ELECTROSTATICS */
1039
1040             /* Analytical PME correction */
1041             zeta2            = _mm256_mul_ps(beta2,rsq00);
1042             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
1043             pmecorrF         = avx256_pmecorrF_f(zeta2);
1044             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1045             felec            = _mm256_mul_ps(qq00,felec);
1046
1047             /* CUBIC SPLINE TABLE DISPERSION */
1048             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1049                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1050             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1051                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1052             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1053                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1054             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1055                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1056             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1057             Heps             = _mm256_mul_ps(vfeps,H);
1058             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1059             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1060             fvdw6            = _mm256_mul_ps(c6_00,FF);
1061
1062             /* CUBIC SPLINE TABLE REPULSION */
1063             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1064             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1065             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1066                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1067             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1068                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1069             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1070                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1071             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1072                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1073             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1074             Heps             = _mm256_mul_ps(vfeps,H);
1075             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1076             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1077             fvdw12           = _mm256_mul_ps(c12_00,FF);
1078             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1079
1080             fscal            = _mm256_add_ps(felec,fvdw);
1081
1082             /* Calculate temporary vectorial force */
1083             tx               = _mm256_mul_ps(fscal,dx00);
1084             ty               = _mm256_mul_ps(fscal,dy00);
1085             tz               = _mm256_mul_ps(fscal,dz00);
1086
1087             /* Update vectorial force */
1088             fix0             = _mm256_add_ps(fix0,tx);
1089             fiy0             = _mm256_add_ps(fiy0,ty);
1090             fiz0             = _mm256_add_ps(fiz0,tz);
1091
1092             fjx0             = _mm256_add_ps(fjx0,tx);
1093             fjy0             = _mm256_add_ps(fjy0,ty);
1094             fjz0             = _mm256_add_ps(fjz0,tz);
1095
1096             /**************************
1097              * CALCULATE INTERACTIONS *
1098              **************************/
1099
1100             r10              = _mm256_mul_ps(rsq10,rinv10);
1101
1102             /* Compute parameters for interactions between i and j atoms */
1103             qq10             = _mm256_mul_ps(iq1,jq0);
1104
1105             /* EWALD ELECTROSTATICS */
1106
1107             /* Analytical PME correction */
1108             zeta2            = _mm256_mul_ps(beta2,rsq10);
1109             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
1110             pmecorrF         = avx256_pmecorrF_f(zeta2);
1111             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1112             felec            = _mm256_mul_ps(qq10,felec);
1113
1114             fscal            = felec;
1115
1116             /* Calculate temporary vectorial force */
1117             tx               = _mm256_mul_ps(fscal,dx10);
1118             ty               = _mm256_mul_ps(fscal,dy10);
1119             tz               = _mm256_mul_ps(fscal,dz10);
1120
1121             /* Update vectorial force */
1122             fix1             = _mm256_add_ps(fix1,tx);
1123             fiy1             = _mm256_add_ps(fiy1,ty);
1124             fiz1             = _mm256_add_ps(fiz1,tz);
1125
1126             fjx0             = _mm256_add_ps(fjx0,tx);
1127             fjy0             = _mm256_add_ps(fjy0,ty);
1128             fjz0             = _mm256_add_ps(fjz0,tz);
1129
1130             /**************************
1131              * CALCULATE INTERACTIONS *
1132              **************************/
1133
1134             r20              = _mm256_mul_ps(rsq20,rinv20);
1135
1136             /* Compute parameters for interactions between i and j atoms */
1137             qq20             = _mm256_mul_ps(iq2,jq0);
1138
1139             /* EWALD ELECTROSTATICS */
1140
1141             /* Analytical PME correction */
1142             zeta2            = _mm256_mul_ps(beta2,rsq20);
1143             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1144             pmecorrF         = avx256_pmecorrF_f(zeta2);
1145             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1146             felec            = _mm256_mul_ps(qq20,felec);
1147
1148             fscal            = felec;
1149
1150             /* Calculate temporary vectorial force */
1151             tx               = _mm256_mul_ps(fscal,dx20);
1152             ty               = _mm256_mul_ps(fscal,dy20);
1153             tz               = _mm256_mul_ps(fscal,dz20);
1154
1155             /* Update vectorial force */
1156             fix2             = _mm256_add_ps(fix2,tx);
1157             fiy2             = _mm256_add_ps(fiy2,ty);
1158             fiz2             = _mm256_add_ps(fiz2,tz);
1159
1160             fjx0             = _mm256_add_ps(fjx0,tx);
1161             fjy0             = _mm256_add_ps(fjy0,ty);
1162             fjz0             = _mm256_add_ps(fjz0,tz);
1163
1164             fjptrA             = f+j_coord_offsetA;
1165             fjptrB             = f+j_coord_offsetB;
1166             fjptrC             = f+j_coord_offsetC;
1167             fjptrD             = f+j_coord_offsetD;
1168             fjptrE             = f+j_coord_offsetE;
1169             fjptrF             = f+j_coord_offsetF;
1170             fjptrG             = f+j_coord_offsetG;
1171             fjptrH             = f+j_coord_offsetH;
1172
1173             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1174
1175             /* Inner loop uses 197 flops */
1176         }
1177
1178         if(jidx<j_index_end)
1179         {
1180
1181             /* Get j neighbor index, and coordinate index */
1182             jnrlistA         = jjnr[jidx];
1183             jnrlistB         = jjnr[jidx+1];
1184             jnrlistC         = jjnr[jidx+2];
1185             jnrlistD         = jjnr[jidx+3];
1186             jnrlistE         = jjnr[jidx+4];
1187             jnrlistF         = jjnr[jidx+5];
1188             jnrlistG         = jjnr[jidx+6];
1189             jnrlistH         = jjnr[jidx+7];
1190             /* Sign of each element will be negative for non-real atoms.
1191              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1192              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1193              */
1194             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1195                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1196
1197             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1198             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1199             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1200             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1201             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1202             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1203             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1204             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1205             j_coord_offsetA  = DIM*jnrA;
1206             j_coord_offsetB  = DIM*jnrB;
1207             j_coord_offsetC  = DIM*jnrC;
1208             j_coord_offsetD  = DIM*jnrD;
1209             j_coord_offsetE  = DIM*jnrE;
1210             j_coord_offsetF  = DIM*jnrF;
1211             j_coord_offsetG  = DIM*jnrG;
1212             j_coord_offsetH  = DIM*jnrH;
1213
1214             /* load j atom coordinates */
1215             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1216                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1217                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1218                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1219                                                  &jx0,&jy0,&jz0);
1220
1221             /* Calculate displacement vector */
1222             dx00             = _mm256_sub_ps(ix0,jx0);
1223             dy00             = _mm256_sub_ps(iy0,jy0);
1224             dz00             = _mm256_sub_ps(iz0,jz0);
1225             dx10             = _mm256_sub_ps(ix1,jx0);
1226             dy10             = _mm256_sub_ps(iy1,jy0);
1227             dz10             = _mm256_sub_ps(iz1,jz0);
1228             dx20             = _mm256_sub_ps(ix2,jx0);
1229             dy20             = _mm256_sub_ps(iy2,jy0);
1230             dz20             = _mm256_sub_ps(iz2,jz0);
1231
1232             /* Calculate squared distance and things based on it */
1233             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1234             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1235             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1236
1237             rinv00           = avx256_invsqrt_f(rsq00);
1238             rinv10           = avx256_invsqrt_f(rsq10);
1239             rinv20           = avx256_invsqrt_f(rsq20);
1240
1241             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1242             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1243             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1244
1245             /* Load parameters for j particles */
1246             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1247                                                                  charge+jnrC+0,charge+jnrD+0,
1248                                                                  charge+jnrE+0,charge+jnrF+0,
1249                                                                  charge+jnrG+0,charge+jnrH+0);
1250             vdwjidx0A        = 2*vdwtype[jnrA+0];
1251             vdwjidx0B        = 2*vdwtype[jnrB+0];
1252             vdwjidx0C        = 2*vdwtype[jnrC+0];
1253             vdwjidx0D        = 2*vdwtype[jnrD+0];
1254             vdwjidx0E        = 2*vdwtype[jnrE+0];
1255             vdwjidx0F        = 2*vdwtype[jnrF+0];
1256             vdwjidx0G        = 2*vdwtype[jnrG+0];
1257             vdwjidx0H        = 2*vdwtype[jnrH+0];
1258
1259             fjx0             = _mm256_setzero_ps();
1260             fjy0             = _mm256_setzero_ps();
1261             fjz0             = _mm256_setzero_ps();
1262
1263             /**************************
1264              * CALCULATE INTERACTIONS *
1265              **************************/
1266
1267             r00              = _mm256_mul_ps(rsq00,rinv00);
1268             r00              = _mm256_andnot_ps(dummy_mask,r00);
1269
1270             /* Compute parameters for interactions between i and j atoms */
1271             qq00             = _mm256_mul_ps(iq0,jq0);
1272             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1273                                             vdwioffsetptr0+vdwjidx0B,
1274                                             vdwioffsetptr0+vdwjidx0C,
1275                                             vdwioffsetptr0+vdwjidx0D,
1276                                             vdwioffsetptr0+vdwjidx0E,
1277                                             vdwioffsetptr0+vdwjidx0F,
1278                                             vdwioffsetptr0+vdwjidx0G,
1279                                             vdwioffsetptr0+vdwjidx0H,
1280                                             &c6_00,&c12_00);
1281
1282             /* Calculate table index by multiplying r with table scale and truncate to integer */
1283             rt               = _mm256_mul_ps(r00,vftabscale);
1284             vfitab           = _mm256_cvttps_epi32(rt);
1285             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1286             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1287             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1288             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1289             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1290             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1291
1292             /* EWALD ELECTROSTATICS */
1293
1294             /* Analytical PME correction */
1295             zeta2            = _mm256_mul_ps(beta2,rsq00);
1296             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
1297             pmecorrF         = avx256_pmecorrF_f(zeta2);
1298             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1299             felec            = _mm256_mul_ps(qq00,felec);
1300
1301             /* CUBIC SPLINE TABLE DISPERSION */
1302             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1303                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1304             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1305                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1306             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1307                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1308             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1309                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1310             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1311             Heps             = _mm256_mul_ps(vfeps,H);
1312             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1313             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1314             fvdw6            = _mm256_mul_ps(c6_00,FF);
1315
1316             /* CUBIC SPLINE TABLE REPULSION */
1317             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1318             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1319             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1320                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1321             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1322                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1323             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1324                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1325             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1326                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1327             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1328             Heps             = _mm256_mul_ps(vfeps,H);
1329             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1330             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1331             fvdw12           = _mm256_mul_ps(c12_00,FF);
1332             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1333
1334             fscal            = _mm256_add_ps(felec,fvdw);
1335
1336             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1337
1338             /* Calculate temporary vectorial force */
1339             tx               = _mm256_mul_ps(fscal,dx00);
1340             ty               = _mm256_mul_ps(fscal,dy00);
1341             tz               = _mm256_mul_ps(fscal,dz00);
1342
1343             /* Update vectorial force */
1344             fix0             = _mm256_add_ps(fix0,tx);
1345             fiy0             = _mm256_add_ps(fiy0,ty);
1346             fiz0             = _mm256_add_ps(fiz0,tz);
1347
1348             fjx0             = _mm256_add_ps(fjx0,tx);
1349             fjy0             = _mm256_add_ps(fjy0,ty);
1350             fjz0             = _mm256_add_ps(fjz0,tz);
1351
1352             /**************************
1353              * CALCULATE INTERACTIONS *
1354              **************************/
1355
1356             r10              = _mm256_mul_ps(rsq10,rinv10);
1357             r10              = _mm256_andnot_ps(dummy_mask,r10);
1358
1359             /* Compute parameters for interactions between i and j atoms */
1360             qq10             = _mm256_mul_ps(iq1,jq0);
1361
1362             /* EWALD ELECTROSTATICS */
1363
1364             /* Analytical PME correction */
1365             zeta2            = _mm256_mul_ps(beta2,rsq10);
1366             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
1367             pmecorrF         = avx256_pmecorrF_f(zeta2);
1368             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1369             felec            = _mm256_mul_ps(qq10,felec);
1370
1371             fscal            = felec;
1372
1373             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1374
1375             /* Calculate temporary vectorial force */
1376             tx               = _mm256_mul_ps(fscal,dx10);
1377             ty               = _mm256_mul_ps(fscal,dy10);
1378             tz               = _mm256_mul_ps(fscal,dz10);
1379
1380             /* Update vectorial force */
1381             fix1             = _mm256_add_ps(fix1,tx);
1382             fiy1             = _mm256_add_ps(fiy1,ty);
1383             fiz1             = _mm256_add_ps(fiz1,tz);
1384
1385             fjx0             = _mm256_add_ps(fjx0,tx);
1386             fjy0             = _mm256_add_ps(fjy0,ty);
1387             fjz0             = _mm256_add_ps(fjz0,tz);
1388
1389             /**************************
1390              * CALCULATE INTERACTIONS *
1391              **************************/
1392
1393             r20              = _mm256_mul_ps(rsq20,rinv20);
1394             r20              = _mm256_andnot_ps(dummy_mask,r20);
1395
1396             /* Compute parameters for interactions between i and j atoms */
1397             qq20             = _mm256_mul_ps(iq2,jq0);
1398
1399             /* EWALD ELECTROSTATICS */
1400
1401             /* Analytical PME correction */
1402             zeta2            = _mm256_mul_ps(beta2,rsq20);
1403             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1404             pmecorrF         = avx256_pmecorrF_f(zeta2);
1405             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1406             felec            = _mm256_mul_ps(qq20,felec);
1407
1408             fscal            = felec;
1409
1410             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1411
1412             /* Calculate temporary vectorial force */
1413             tx               = _mm256_mul_ps(fscal,dx20);
1414             ty               = _mm256_mul_ps(fscal,dy20);
1415             tz               = _mm256_mul_ps(fscal,dz20);
1416
1417             /* Update vectorial force */
1418             fix2             = _mm256_add_ps(fix2,tx);
1419             fiy2             = _mm256_add_ps(fiy2,ty);
1420             fiz2             = _mm256_add_ps(fiz2,tz);
1421
1422             fjx0             = _mm256_add_ps(fjx0,tx);
1423             fjy0             = _mm256_add_ps(fjy0,ty);
1424             fjz0             = _mm256_add_ps(fjz0,tz);
1425
1426             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1427             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1428             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1429             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1430             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1431             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1432             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1433             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1434
1435             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1436
1437             /* Inner loop uses 200 flops */
1438         }
1439
1440         /* End of innermost loop */
1441
1442         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1443                                                  f+i_coord_offset,fshift+i_shift_offset);
1444
1445         /* Increment number of inner iterations */
1446         inneriter                  += j_index_end - j_index_start;
1447
1448         /* Outer loop uses 18 flops */
1449     }
1450
1451     /* Increment number of outer iterations */
1452     outeriter        += nri;
1453
1454     /* Update outer/inner flops */
1455
1456     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*200);
1457 }