src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_template_avx_256_double.pre

   1 /* #if 0 */
   2 /*
   3  * This file is part of the GROMACS molecular simulation package.
   4  *
   5  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   6  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   7  * and including many others, as listed in the AUTHORS file in the
   8  * top-level source directory and at http://www.gromacs.org.
   9  *
  10  * GROMACS is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU Lesser General Public License
  12  * as published by the Free Software Foundation; either version 2.1
  13  * of the License, or (at your option) any later version.
  14  *
  15  * GROMACS is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public
  21  * License along with GROMACS; if not, see
  22  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  23  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  24  *
  25  * If you want to redistribute modifications to GROMACS, please
  26  * consider that scientific software is very special. Version
  27  * control is crucial - bugs must be traceable. We will be happy to
  28  * consider code for inclusion in the official distribution, but
  29  * derived work must not be called official GROMACS. Details are found
  30  * in the README & COPYING files - if they are missing, get the
  31  * official version at http://www.gromacs.org.
  32  *
  33  * To help us fund GROMACS development, we humbly ask that you cite
  34  * the research papers on the package. Check out http://www.gromacs.org.
  35  */
  36 #error This file must be processed with the Gromacs pre-preprocessor
  37 /* #endif */
  38 /* #if INCLUDE_HEADER */
  39 #include "gmxpre.h"
  40
  41 #include "config.h"
  42
  43 #include <math.h>
  44
  45 #include "../nb_kernel.h"
  46 #include "gromacs/gmxlib/nrnb.h"
  47
  48 #include "kernelutil_x86_avx_256_double.h"
  49 /* #endif */
  50
  51 /* ## List of variables set by the generating script:                                    */
  52 /* ##                                                                                    */
  53 /* ## Setttings that apply to the entire kernel:                                         */
  54 /* ## KERNEL_ELEC:           String, choice for electrostatic interactions               */
  55 /* ## KERNEL_VDW:            String, choice for van der Waals interactions               */
  56 /* ## KERNEL_NAME:           String, name of this kernel                                 */
  57 /* ## KERNEL_VF:             String telling if we calculate potential, force, or both    */
  58 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
  59 /* ##                                                                                    */
  60 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops:             */
  61 /* ## PARTICLES_I[]/         Arrays with lists of i/j particles to use in kernel. It is  */
  62 /* ## PARTICLES_J[]:         just [0] for particle geometry, but can be longer for water */
  63 /* ## PARTICLES_ELEC_I[]/    Arrays with lists of i/j particle that have electrostatics  */
  64 /* ## PARTICLES_ELEC_J[]:    interactions that should be calculated in this kernel.      */
  65 /* ## PARTICLES_VDW_I[]/     Arrays with the list of i/j particle that have VdW          */
  66 /* ## PARTICLES_VDW_J[]:     interactions that should be calculated in this kernel.      */
  67 /* ##                                                                                    */
  68 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle)   */
  69 /* ## PAIRS_IJ[]:            Array with (i,j) tuples of pairs for which interactions     */
  70 /* ##                        should be calculated in this kernel. Zero-charge particles  */
  71 /* ##                        do not have interactions with particles without vdw, and    */
  72 /* ##                        Vdw-only interactions are not evaluated in a no-vdw-kernel. */
  73 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
  74 /* ##                        For each i-j pair, the element [I][J] is a list of strings  */
  75 /* ##                        defining properties/flags of this interaction. Examples     */
  76 /* ##                        include 'electrostatics'/'vdw' if that type of interaction  */
  77 /* ##                        should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values  */
  78 /* ##                        are needed, and 'exactcutoff' or 'shift','switch' to        */
  79 /* ##                        decide if the force/potential should be modified. This way  */
  80 /* ##                        we only calculate values absolutely needed for each case.   */
  81
  82 /* ## Calculate the size and offset for (merged/interleaved) table data */
  83
  84 /*
  85  * Gromacs nonbonded kernel:   {KERNEL_NAME}
  86  * Electrostatics interaction: {KERNEL_ELEC}
  87  * VdW interaction:            {KERNEL_VDW}
  88  * Geometry:                   {GEOMETRY_I}-{GEOMETRY_J}
  89  * Calculate force/pot:        {KERNEL_VF}
  90  */
  91 void
  92 {KERNEL_NAME}
  93                     (t_nblist                    * gmx_restrict       nlist,
  94                      rvec                        * gmx_restrict          xx,
  95                      rvec                        * gmx_restrict          ff,
  96                      struct t_forcerec           * gmx_restrict          fr,
  97                      t_mdatoms                   * gmx_restrict     mdatoms,
  98                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  99                      t_nrnb                      * gmx_restrict        nrnb)
 100 {
 101     /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
 102     /* ## so there is no point in going to extremes to exclude variables that are not needed. */
 103     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 104      * just 0 for non-waters.
 105      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 106      * jnr indices corresponding to data put in the four positions in the SIMD register.
 107      */
 108     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 109     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 110     int              jnrA,jnrB,jnrC,jnrD;
 111     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 112     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 113     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 114     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 115     real             rcutoff_scalar;
 116     real             *shiftvec,*fshift,*x,*f;
 117     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 118     real             scratch[4*DIM];
 119     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 120     /* #for I in PARTICLES_I */
 121     real *           vdwioffsetptr{I};
 122     /* #if 'LJEwald' in KERNEL_VDW */
 123     real *           vdwgridioffsetptr{I};
 124     /* #endif                      */
 125     __m256d          ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
 126     /* #endfor */
 127     /* #for J in PARTICLES_J */
 128     int              vdwjidx{J}A,vdwjidx{J}B,vdwjidx{J}C,vdwjidx{J}D;
 129     __m256d          jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
 130     /* #endfor */
 131     /* #for I,J in PAIRS_IJ */
 132     __m256d          dx{I}{J},dy{I}{J},dz{I}{J},rsq{I}{J},rinv{I}{J},rinvsq{I}{J},r{I}{J},qq{I}{J},c6_{I}{J},c12_{I}{J};
 133     /* #endfor */
 134     /* #if KERNEL_ELEC != 'None' */
 135     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 136     real             *charge;
 137     /* #endif */
 138     /* #if KERNEL_VDW != 'None' */
 139     int              nvdwtype;
 140     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 141     int              *vdwtype;
 142     real             *vdwparam;
 143     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 144     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 145     /* #endif */
 146     /* #if 'Table' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
 147     __m128i          vfitab;
 148     __m128i          ifour       = _mm_set1_epi32(4);
 149     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 150     real             *vftab;
 151     /* #endif */
 152     /* #if 'LJEwald' in KERNEL_VDW */
 153     /* #for I,J in PAIRS_IJ */
 154     __m256d           c6grid_{I}{J};
 155     /* #endfor */
 156     real             *vdwgridparam;
 157     __m256d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 158     __m256d           one_half  = _mm256_set1_pd(0.5);
 159     __m256d           minus_one = _mm256_set1_pd(-1.0);
 160     /* #endif */
 161     /* #if 'Ewald' in KERNEL_ELEC */
 162     __m128i          ewitab;
 163     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 164     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 165     real             *ewtab;
 166     /* #endif */
 167     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 168     __m256d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 169     real             rswitch_scalar,d_scalar;
 170     /* #endif */
 171     __m256d          dummy_mask,cutoff_mask;
 172     __m128           tmpmask0,tmpmask1;
 173     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 174     __m256d          one     = _mm256_set1_pd(1.0);
 175     __m256d          two     = _mm256_set1_pd(2.0);
 176     x                = xx[0];
 177     f                = ff[0];
 178
 179     nri              = nlist->nri;
 180     iinr             = nlist->iinr;
 181     jindex           = nlist->jindex;
 182     jjnr             = nlist->jjnr;
 183     shiftidx         = nlist->shift;
 184     gid              = nlist->gid;
 185     shiftvec         = fr->shift_vec[0];
 186     fshift           = fr->fshift[0];
 187     /* #if KERNEL_ELEC != 'None' */
 188     facel            = _mm256_set1_pd(fr->ic->epsfac);
 189     charge           = mdatoms->chargeA;
 190     /*     #if 'ReactionField' in KERNEL_ELEC */
 191     krf              = _mm256_set1_pd(fr->ic->k_rf);
 192     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
 193     crf              = _mm256_set1_pd(fr->ic->c_rf);
 194     /*     #endif */
 195     /* #endif */
 196     /* #if KERNEL_VDW != 'None' */
 197     nvdwtype         = fr->ntype;
 198     vdwparam         = fr->nbfp;
 199     vdwtype          = mdatoms->typeA;
 200     /* #endif */
 201     /* #if 'LJEwald' in KERNEL_VDW */
 202     vdwgridparam     = fr->ljpme_c6grid;
 203     sh_lj_ewald      = _mm256_set1_pd(fr->ic->sh_lj_ewald);
 204     ewclj            = _mm256_set1_pd(fr->ic->ewaldcoeff_lj);
 205     ewclj2           = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
 206     /* #endif */
 207
 208     /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 209     vftab            = kernel_data->table_elec_vdw->data;
 210     vftabscale       = _mm256_set1_pd(kernel_data->table_elec_vdw->scale);
 211     /* #elif 'Table' in KERNEL_ELEC */
 212     vftab            = kernel_data->table_elec->data;
 213     vftabscale       = _mm256_set1_pd(kernel_data->table_elec->scale);
 214     /* #elif 'Table' in KERNEL_VDW */
 215     vftab            = kernel_data->table_vdw->data;
 216     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 217     /* #endif */
 218
 219     /* #if 'Ewald' in KERNEL_ELEC */
 220     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 221     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 222     beta2            = _mm256_mul_pd(beta,beta);
 223     beta3            = _mm256_mul_pd(beta,beta2);
 224
 225     /*     #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
 226     ewtab            = fr->ic->tabq_coul_F;
 227     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 228     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 229     /*     #else */
 230     ewtab            = fr->ic->tabq_coul_FDV0;
 231     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 232     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 233      /*     #endif */
 234     /* #endif */
 235
 236     /* #if 'Water' in GEOMETRY_I */
 237     /* Setup water-specific parameters */
 238     inr              = nlist->iinr[0];
 239     /*     #for I in PARTICLES_ELEC_I */
 240     iq{I}              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+{I}]));
 241     /*     #endfor */
 242     /*     #for I in PARTICLES_VDW_I */
 243     vdwioffsetptr{I}   = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
 244     /*         #if 'LJEwald' in KERNEL_VDW */
 245     vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
 246     /*         #endif */
 247     /*     #endfor */
 248     /* #endif */
 249
 250     /* #if 'Water' in GEOMETRY_J */
 251     /*     #for J in PARTICLES_ELEC_J */
 252     jq{J}              = _mm256_set1_pd(charge[inr+{J}]);
 253     /*     #endfor */
 254     /*     #for J in PARTICLES_VDW_J */
 255     vdwjidx{J}A        = 2*vdwtype[inr+{J}];
 256     /*     #endfor */
 257     /*     #for I,J in PAIRS_IJ */
 258     /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 259     qq{I}{J}             = _mm256_mul_pd(iq{I},jq{J});
 260     /*         #endif */
 261     /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 262     /*             #if 'LJEwald' in KERNEL_VDW */
 263     c6_{I}{J}            = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A]);
 264     c12_{I}{J}           = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A+1]);
 265     c6grid_{I}{J}        = _mm256_set1_pd(vdwgridioffsetptr{I}[vdwjidx{J}A]);
 266     /*             #else */
 267     c6_{I}{J}            = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A]);
 268     c12_{I}{J}           = _mm256_set1_pd(vdwioffsetptr{I}[vdwjidx{J}A+1]);
 269     /*             #endif */
 270     /*         #endif */
 271     /*     #endfor */
 272     /* #endif */
 273
 274     /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
 275     /*     #if KERNEL_ELEC!='None' */
 276     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 277     rcutoff_scalar   = fr->ic->rcoulomb;
 278     /*     #else */
 279     rcutoff_scalar   = fr->ic->rvdw;
 280     /*     #endif */
 281     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 282     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 283     /* #endif */
 284
 285     /* #if KERNEL_MOD_VDW=='PotentialShift' */
 286     sh_vdw_invrcut6  = _mm256_set1_pd(fr->ic->sh_invrc6);
 287     rvdw             = _mm256_set1_pd(fr->ic->rvdw);
 288     /* #endif */
 289
 290     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 291     /*     #if KERNEL_MOD_ELEC=='PotentialSwitch'  */
 292     rswitch_scalar   = fr->ic->rcoulomb_switch;
 293     rswitch          = _mm256_set1_pd(rswitch_scalar);
 294     /*     #else */
 295     rswitch_scalar   = fr->ic->rvdw_switch;
 296     rswitch          = _mm256_set1_pd(rswitch_scalar);
 297     /*     #endif */
 298     /* Setup switch parameters */
 299     d_scalar         = rcutoff_scalar-rswitch_scalar;
 300     d                = _mm256_set1_pd(d_scalar);
 301     swV3             = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 302     swV4             = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 303     swV5             = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 304     /*     #if 'Force' in KERNEL_VF */
 305     swF2             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 306     swF3             = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 307     swF4             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 308     /*     #endif */
 309     /* #endif */
 310
 311     /* Avoid stupid compiler warnings */
 312     jnrA = jnrB = jnrC = jnrD = 0;
 313     j_coord_offsetA = 0;
 314     j_coord_offsetB = 0;
 315     j_coord_offsetC = 0;
 316     j_coord_offsetD = 0;
 317
 318     /* ## Keep track of the floating point operations we issue for reporting! */
 319     /* #define OUTERFLOPS 0 */
 320     outeriter        = 0;
 321     inneriter        = 0;
 322
 323     for(iidx=0;iidx<4*DIM;iidx++)
 324     {
 325         scratch[iidx] = 0.0;
 326     }
 327
 328     /* Start outer loop over neighborlists */
 329     for(iidx=0; iidx<nri; iidx++)
 330     {
 331         /* Load shift vector for this list */
 332         i_shift_offset   = DIM*shiftidx[iidx];
 333
 334         /* Load limits for loop over neighbors */
 335         j_index_start    = jindex[iidx];
 336         j_index_end      = jindex[iidx+1];
 337
 338         /* Get outer coordinate index */
 339         inr              = iinr[iidx];
 340         i_coord_offset   = DIM*inr;
 341
 342         /* Load i particle coords and add shift vector */
 343         /* #if GEOMETRY_I == 'Particle' */
 344         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 345         /* #elif GEOMETRY_I == 'Water3' */
 346         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 347                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 348         /* #elif GEOMETRY_I == 'Water4' */
 349         /*     #if 0 in PARTICLES_I                 */
 350         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 351                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 352         /*     #else                                */
 353         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 354                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 355         /*     #endif                               */
 356         /* #endif                                   */
 357
 358         /* #if 'Force' in KERNEL_VF */
 359         /*     #for I in PARTICLES_I */
 360         fix{I}             = _mm256_setzero_pd();
 361         fiy{I}             = _mm256_setzero_pd();
 362         fiz{I}             = _mm256_setzero_pd();
 363         /*     #endfor */
 364         /* #endif */
 365
 366         /* ## For water we already preloaded parameters at the start of the kernel */
 367         /* #if not 'Water' in GEOMETRY_I */
 368         /* Load parameters for i particles */
 369         /*     #for I in PARTICLES_ELEC_I */
 370         iq{I}              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+{I}]));
 371         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 372         /*     #endfor */
 373         /*     #for I in PARTICLES_VDW_I */
 374         vdwioffsetptr{I}   = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
 375         /*         #if 'LJEwald' in KERNEL_VDW */
 376         vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
 377         /*         #endif */
 378         /*     #endfor */
 379         /* #endif */
 380
 381         /* #if 'Potential' in KERNEL_VF */
 382         /* Reset potential sums */
 383         /*     #if KERNEL_ELEC != 'None' */
 384         velecsum         = _mm256_setzero_pd();
 385         /*     #endif */
 386         /*     #if KERNEL_VDW != 'None' */
 387         vvdwsum          = _mm256_setzero_pd();
 388         /*     #endif */
 389         /* #endif */
 390
 391         /* #for ROUND in ['Loop','Epilogue'] */
 392
 393         /* #if ROUND =='Loop' */
 394         /* Start inner kernel loop */
 395         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 396         {
 397         /* ## First round is normal loop (next statement resets indentation) */
 398         /*     #if 0 */
 399         }
 400         /*     #endif */
 401         /* #else */
 402         if(jidx<j_index_end)
 403         {
 404         /* ## Second round is epilogue */
 405         /* #endif */
 406         /* #define INNERFLOPS 0 */
 407
 408             /* Get j neighbor index, and coordinate index */
 409             /* #if ROUND =='Loop' */
 410             jnrA             = jjnr[jidx];
 411             jnrB             = jjnr[jidx+1];
 412             jnrC             = jjnr[jidx+2];
 413             jnrD             = jjnr[jidx+3];
 414             /* #else */
 415             jnrlistA         = jjnr[jidx];
 416             jnrlistB         = jjnr[jidx+1];
 417             jnrlistC         = jjnr[jidx+2];
 418             jnrlistD         = jjnr[jidx+3];
 419             /* Sign of each element will be negative for non-real atoms.
 420              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 421              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 422              */
 423             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 424
 425             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 426             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 427             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 428
 429             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 430             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 431             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 432             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 433             /* #endif */
 434             j_coord_offsetA  = DIM*jnrA;
 435             j_coord_offsetB  = DIM*jnrB;
 436             j_coord_offsetC  = DIM*jnrC;
 437             j_coord_offsetD  = DIM*jnrD;
 438
 439             /* load j atom coordinates */
 440             /* #if GEOMETRY_J == 'Particle'             */
 441             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 442                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 443                                                  &jx0,&jy0,&jz0);
 444             /* #elif GEOMETRY_J == 'Water3'             */
 445             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 446                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 447                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 448             /* #elif GEOMETRY_J == 'Water4'             */
 449             /*     #if 0 in PARTICLES_J                 */
 450             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 451                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 452                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 453                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 454             /*     #else                                */
 455             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 456                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 457                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 458             /*     #endif                               */
 459             /* #endif                                   */
 460
 461             /* Calculate displacement vector */
 462             /* #for I,J in PAIRS_IJ */
 463             dx{I}{J}             = _mm256_sub_pd(ix{I},jx{J});
 464             dy{I}{J}             = _mm256_sub_pd(iy{I},jy{J});
 465             dz{I}{J}             = _mm256_sub_pd(iz{I},jz{J});
 466             /*     #define INNERFLOPS INNERFLOPS+3 */
 467             /* #endfor */
 468
 469             /* Calculate squared distance and things based on it */
 470             /* #for I,J in PAIRS_IJ */
 471             rsq{I}{J}            = gmx_mm256_calc_rsq_pd(dx{I}{J},dy{I}{J},dz{I}{J});
 472             /*     #define INNERFLOPS INNERFLOPS+5 */
 473             /* #endfor */
 474
 475             /* #for I,J in PAIRS_IJ */
 476             /*     #if 'rinv' in INTERACTION_FLAGS[I][J] */
 477             rinv{I}{J}           = avx256_invsqrt_d(rsq{I}{J});
 478             /*         #define INNERFLOPS INNERFLOPS+5 */
 479             /*     #endif */
 480             /* #endfor */
 481
 482             /* #for I,J in PAIRS_IJ */
 483             /*     #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
 484             /*         # if 'rinv' not in INTERACTION_FLAGS[I][J] */
 485             rinvsq{I}{J}         = avx256_inv_d(rsq{I}{J});
 486             /*             #define INNERFLOPS INNERFLOPS+4 */
 487             /*         #else */
 488             rinvsq{I}{J}         = _mm256_mul_pd(rinv{I}{J},rinv{I}{J});
 489             /*             #define INNERFLOPS INNERFLOPS+1 */
 490             /*         #endif */
 491             /*     #endif */
 492             /* #endfor */
 493
 494             /* #if not 'Water' in GEOMETRY_J */
 495             /* Load parameters for j particles */
 496             /*     #for J in PARTICLES_ELEC_J */
 497             jq{J}              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+{J},charge+jnrB+{J},
 498                                                                  charge+jnrC+{J},charge+jnrD+{J});
 499             /*     #endfor */
 500             /*     #for J in PARTICLES_VDW_J */
 501             vdwjidx{J}A        = 2*vdwtype[jnrA+{J}];
 502             vdwjidx{J}B        = 2*vdwtype[jnrB+{J}];
 503             vdwjidx{J}C        = 2*vdwtype[jnrC+{J}];
 504             vdwjidx{J}D        = 2*vdwtype[jnrD+{J}];
 505             /*     #endfor */
 506             /* #endif */
 507
 508             /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
 509             /*     #for J in PARTICLES_J */
 510             fjx{J}             = _mm256_setzero_pd();
 511             fjy{J}             = _mm256_setzero_pd();
 512             fjz{J}             = _mm256_setzero_pd();
 513             /*     #endfor */
 514             /* #endif */
 515
 516             /* #for I,J in PAIRS_IJ */
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 523             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 524             /*         ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
 525             if (gmx_mm256_any_lt(rsq{I}{J},rcutoff2))
 526             {
 527                 /*     #if 0    ## this and the next two lines is a hack to maintain auto-indentation in template file */
 528             }
 529             /*         #endif */
 530             /*         #define INNERFLOPS INNERFLOPS+1 */
 531             /*     #endif */
 532
 533             /*     #if 'r' in INTERACTION_FLAGS[I][J] */
 534             r{I}{J}              = _mm256_mul_pd(rsq{I}{J},rinv{I}{J});
 535             /*         #if ROUND == 'Epilogue' */
 536             r{I}{J}              = _mm256_andnot_pd(dummy_mask,r{I}{J});
 537             /*             #define INNERFLOPS INNERFLOPS+1 */
 538             /*         #endif */
 539             /*         #define INNERFLOPS INNERFLOPS+1 */
 540             /*     #endif */
 541
 542             /*     ## For water geometries we already loaded parameters at the start of the kernel */
 543             /*     #if not 'Water' in GEOMETRY_J */
 544             /* Compute parameters for interactions between i and j atoms */
 545             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 546             qq{I}{J}             = _mm256_mul_pd(iq{I},jq{J});
 547             /*             #define INNERFLOPS INNERFLOPS+1 */
 548             /*         #endif */
 549             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 550             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr{I}+vdwjidx{J}A,
 551                                             vdwioffsetptr{I}+vdwjidx{J}B,
 552                                             vdwioffsetptr{I}+vdwjidx{J}C,
 553                                             vdwioffsetptr{I}+vdwjidx{J}D,
 554                                             &c6_{I}{J},&c12_{I}{J});
 555
 556             /*           #if 'LJEwald' in KERNEL_VDW */
 557             c6grid_{I}{J}       = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr{I}+vdwjidx{J}A,
 558                                                                   vdwgridioffsetptr{I}+vdwjidx{J}B,
 559                                                                   vdwgridioffsetptr{I}+vdwjidx{J}C,
 560                                                                   vdwgridioffsetptr{I}+vdwjidx{J}D);
 561             /*           #endif */
 562             /*         #endif */
 563             /*     #endif */
 564
 565             /*     #if 'table' in INTERACTION_FLAGS[I][J] */
 566             /* Calculate table index by multiplying r with table scale and truncate to integer */
 567             rt               = _mm256_mul_pd(r{I}{J},vftabscale);
 568             vfitab           = _mm256_cvttpd_epi32(rt);
 569             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 570             /*         #define INNERFLOPS INNERFLOPS+4                          */
 571             /*         #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW     */
 572             /*             ## 3 tables, 4 bytes per point: multiply index by 12 */
 573             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 574             /*         #elif 'Table' in KERNEL_ELEC                             */
 575             /*             ## 1 table, 4 bytes per point: multiply index by 4   */
 576             vfitab           = _mm_slli_epi32(vfitab,2);
 577             /*         #elif 'Table' in KERNEL_VDW                              */
 578             /*             ## 2 tables, 4 bytes per point: multiply index by 8  */
 579             vfitab           = _mm_slli_epi32(vfitab,3);
 580             /*         #endif                                                   */
 581             /*     #endif */
 582
 583             /*     ## ELECTROSTATIC INTERACTIONS */
 584             /*     #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 585
 586             /*         #if KERNEL_ELEC=='Coulomb' */
 587
 588             /* COULOMB ELECTROSTATICS */
 589             velec            = _mm256_mul_pd(qq{I}{J},rinv{I}{J});
 590             /*             #define INNERFLOPS INNERFLOPS+1 */
 591             /*             #if 'Force' in KERNEL_VF */
 592             felec            = _mm256_mul_pd(velec,rinvsq{I}{J});
 593             /*                 #define INNERFLOPS INNERFLOPS+1 */
 594             /*             #endif */
 595
 596             /*         #elif KERNEL_ELEC=='ReactionField' */
 597
 598             /* REACTION-FIELD ELECTROSTATICS */
 599             /*             #if 'Potential' in KERNEL_VF */
 600             velec            = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_add_pd(rinv{I}{J},_mm256_mul_pd(krf,rsq{I}{J})),crf));
 601             /*                 #define INNERFLOPS INNERFLOPS+4 */
 602             /*             #endif */
 603             /*             #if 'Force' in KERNEL_VF */
 604             felec            = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_mul_pd(rinv{I}{J},rinvsq{I}{J}),krf2));
 605             /*                 #define INNERFLOPS INNERFLOPS+3 */
 606             /*             #endif */
 607
 608             /*         #elif KERNEL_ELEC=='Ewald' */
 609             /* EWALD ELECTROSTATICS */
 610
 611             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 612             ewrt             = _mm256_mul_pd(r{I}{J},ewtabscale);
 613             ewitab           = _mm256_cvttpd_epi32(ewrt);
 614             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 615             /*             #define INNERFLOPS INNERFLOPS+4 */
 616             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
 617             ewitab           = _mm_slli_epi32(ewitab,2);
 618             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 619             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 620             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 621             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 622             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 623             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 624             /*                 #define INNERFLOPS INNERFLOPS+2 */
 625             /*                 #if KERNEL_MOD_ELEC=='PotentialShift' */
 626             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 627             velec            = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(_mm256_sub_pd(rinv{I}{J},sh_ewald),velec));
 628             /*                     #define INNERFLOPS INNERFLOPS+7 */
 629             /*                 #else */
 630             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 631             velec            = _mm256_mul_pd(qq{I}{J},_mm256_sub_pd(rinv{I}{J},velec));
 632             /*                     #define INNERFLOPS INNERFLOPS+6 */
 633             /*                 #endif */
 634             /*                 #if 'Force' in KERNEL_VF */
 635             felec            = _mm256_mul_pd(_mm256_mul_pd(qq{I}{J},rinv{I}{J}),_mm256_sub_pd(rinvsq{I}{J},felec));
 636             /*                      #define INNERFLOPS INNERFLOPS+3 */
 637             /*                 #endif */
 638             /*             #elif KERNEL_VF=='Force' */
 639             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
 640                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
 641                                             &ewtabF,&ewtabFn);
 642             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
 643             felec            = _mm256_mul_pd(_mm256_mul_pd(qq{I}{J},rinv{I}{J}),_mm256_sub_pd(rinvsq{I}{J},felec));
 644             /*                 #define INNERFLOPS INNERFLOPS+7 */
 645             /*             #endif */
 646
 647             /*         #elif KERNEL_ELEC=='CubicSplineTable' */
 648
 649             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 650             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 651             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 652             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 653             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 654             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 655             Heps             = _mm256_mul_pd(vfeps,H);
 656             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 657             /*             #define INNERFLOPS INNERFLOPS+4 */
 658             /*             #if 'Potential' in KERNEL_VF */
 659             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 660             velec            = _mm256_mul_pd(qq{I}{J},VV);
 661             /*                 #define INNERFLOPS INNERFLOPS+3 */
 662             /*             #endif */
 663             /*             #if 'Force' in KERNEL_VF */
 664             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 665             felec            = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_mul_pd(qq{I}{J},FF),_mm256_mul_pd(vftabscale,rinv{I}{J})));
 666             /*                 #define INNERFLOPS INNERFLOPS+7 */
 667             /*             #endif */
 668             /*         #endif */
 669             /*         ## End of check for electrostatics interaction forms */
 670             /*     #endif */
 671             /*     ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
 672
 673             /*     #if 'vdw' in INTERACTION_FLAGS[I][J] */
 674
 675             /*         #if KERNEL_VDW=='LennardJones' */
 676
 677             /* LENNARD-JONES DISPERSION/REPULSION */
 678
 679             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 680             /*             #define INNERFLOPS INNERFLOPS+2 */
 681             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 682             vvdw6            = _mm256_mul_pd(c6_{I}{J},rinvsix);
 683             vvdw12           = _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(rinvsix,rinvsix));
 684             /*                 #define INNERFLOPS INNERFLOPS+3 */
 685             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 686             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 687                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_{I}{J},sh_vdw_invrcut6)),one_sixth));
 688             /*                     #define INNERFLOPS INNERFLOPS+8 */
 689             /*                 #else */
 690             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 691             /*                     #define INNERFLOPS INNERFLOPS+3 */
 692             /*                 #endif */
 693             /*                 ## Check for force inside potential check, i.e. this means we already did the potential part */
 694             /*                 #if 'Force' in KERNEL_VF */
 695             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq{I}{J});
 696             /*                     #define INNERFLOPS INNERFLOPS+2 */
 697             /*                 #endif */
 698             /*             #elif KERNEL_VF=='Force' */
 699             /*                 ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
 700             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_{I}{J},rinvsix),c6_{I}{J}),_mm256_mul_pd(rinvsix,rinvsq{I}{J}));
 701             /*                 #define INNERFLOPS INNERFLOPS+4 */
 702             /*             #endif */
 703
 704             /*         #elif KERNEL_VDW=='CubicSplineTable' */
 705
 706             /* CUBIC SPLINE TABLE DISPERSION */
 707             /*             #if 'Table' in KERNEL_ELEC */
 708             vfitab           = _mm_add_epi32(vfitab,ifour);
 709             /*             #endif                     */
 710             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 711             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 712             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 713             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 714             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 715             Heps             = _mm256_mul_pd(vfeps,H);
 716             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 717             /*             #define INNERFLOPS INNERFLOPS+4 */
 718             /*             #if 'Potential' in KERNEL_VF */
 719             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 720             vvdw6            = _mm256_mul_pd(c6_{I}{J},VV);
 721             /*                 #define INNERFLOPS INNERFLOPS+3 */
 722             /*             #endif */
 723             /*             #if 'Force' in KERNEL_VF */
 724             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 725             fvdw6            = _mm256_mul_pd(c6_{I}{J},FF);
 726             /*                 #define INNERFLOPS INNERFLOPS+4 */
 727             /*             #endif */
 728
 729             /* CUBIC SPLINE TABLE REPULSION */
 730             vfitab           = _mm_add_epi32(vfitab,ifour);
 731             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 732             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 733             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 734             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 735             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 736             Heps             = _mm256_mul_pd(vfeps,H);
 737             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 738             /*             #define INNERFLOPS INNERFLOPS+4 */
 739             /*             #if 'Potential' in KERNEL_VF */
 740             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 741             vvdw12           = _mm256_mul_pd(c12_{I}{J},VV);
 742             /*                 #define INNERFLOPS INNERFLOPS+3 */
 743             /*             #endif */
 744             /*             #if 'Force' in KERNEL_VF */
 745             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 746             fvdw12           = _mm256_mul_pd(c12_{I}{J},FF);
 747             /*                 #define INNERFLOPS INNERFLOPS+5 */
 748             /*             #endif */
 749             /*             #if 'Potential' in KERNEL_VF */
 750             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 751             /*                 #define INNERFLOPS INNERFLOPS+1 */
 752             /*             #endif */
 753             /*             #if 'Force' in KERNEL_VF */
 754             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv{I}{J})));
 755             /*                 #define INNERFLOPS INNERFLOPS+4 */
 756             /*             #endif */
 757
 758             /*         #elif KERNEL_VDW=='LJEwald' */
 759
 760             /* Analytical LJ-PME */
 761             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 762             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq{I}{J});
 763             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 764             exponent         = avx256_exp_d(ewcljrsq);
 765             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 766             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 767             /*                 #define INNERFLOPS INNERFLOPS+11 */
 768             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 769             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 770             vvdw6            = _mm256_mul_pd(_mm256_sub_pd(c6_{I}{J},_mm256_mul_pd(c6grid_{I}{J},_mm256_sub_pd(one,poly))),rinvsix);
 771             vvdw12           = _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(rinvsix,rinvsix));
 772             /*                     #define INNERFLOPS INNERFLOPS+6 */
 773             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 774             vvdw             = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_{I}{J},_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 775                                           _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_add_pd(_mm256_mul_pd(c6_{I}{J},sh_vdw_invrcut6),_mm256_mul_pd(c6grid_{I}{J},sh_lj_ewald))),one_sixth));
 776             /*                     #define INNERFLOPS INNERFLOPS+10 */
 777             /*                 #else */
 778             vvdw             = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
 779             /*                 #define INNERFLOPS INNERFLOPS+6 */
 780             /*                 #endif */
 781             /*                  ## Check for force inside potential check, i.e. this means we already did the potential part */
 782             /*                  #if 'Force' in KERNEL_VF */
 783             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 784             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_{I}{J},one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq{I}{J});
 785             /*                 #define INNERFLOPS INNERFLOPS+6 */
 786             /*                  #endif */
 787             /*              #elif KERNEL_VF=='Force' */
 788             /* f6A = 6 * C6grid * (1 - poly) */
 789             f6A              = _mm256_mul_pd(c6grid_{I}{J},_mm256_sub_pd(one,poly));
 790             /* f6B = C6grid * exponent * beta^6 */
 791             f6B              = _mm256_mul_pd(_mm256_mul_pd(c6grid_{I}{J},one_sixth),_mm256_mul_pd(exponent,ewclj6));
 792             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 793             fvdw              = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_{I}{J},rinvsix),_mm256_sub_pd(c6_{I}{J},f6A)),rinvsix),f6B),rinvsq{I}{J});
 794             /*                 #define INNERFLOPS INNERFLOPS+11 */
 795             /*              #endif */
 796             /*         #endif */
 797             /*         ## End of check for vdw interaction forms */
 798             /*     #endif */
 799             /*     ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
 800
 801             /*     #if 'switch' in INTERACTION_FLAGS[I][J] */
 802             d                = _mm256_sub_pd(r{I}{J},rswitch);
 803             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 804             d2               = _mm256_mul_pd(d,d);
 805             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 806             /*         #define INNERFLOPS INNERFLOPS+10 */
 807
 808             /*         #if 'Force' in KERNEL_VF */
 809             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 810             /*             #define INNERFLOPS INNERFLOPS+5 */
 811             /*         #endif */
 812
 813             /* Evaluate switch function */
 814             /*         #if 'Force' in KERNEL_VF */
 815             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 816             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 817             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv{I}{J},_mm256_mul_pd(velec,dsw)) );
 818             /*                 #define INNERFLOPS INNERFLOPS+4 */
 819             /*             #endif */
 820             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 821             fvdw             = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv{I}{J},_mm256_mul_pd(vvdw,dsw)) );
 822             /*                 #define INNERFLOPS INNERFLOPS+4 */
 823             /*             #endif */
 824             /*         #endif */
 825             /*         #if 'Potential' in KERNEL_VF */
 826             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 827             velec            = _mm256_mul_pd(velec,sw);
 828             /*                 #define INNERFLOPS INNERFLOPS+1 */
 829             /*             #endif */
 830             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 831             vvdw             = _mm256_mul_pd(vvdw,sw);
 832             /*                 #define INNERFLOPS INNERFLOPS+1 */
 833             /*             #endif */
 834             /*         #endif */
 835             /*     #endif */
 836             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 837             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 838             cutoff_mask      = _mm256_cmp_pd(rsq{I}{J},rcutoff2,_CMP_LT_OQ);
 839             /*         #define INNERFLOPS INNERFLOPS+1 */
 840             /*     #endif */
 841
 842             /*     #if 'Potential' in KERNEL_VF */
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 845             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 846             velec            = _mm256_and_pd(velec,cutoff_mask);
 847             /*                 #define INNERFLOPS INNERFLOPS+1 */
 848             /*             #endif                                       */
 849             /*             #if ROUND == 'Epilogue' */
 850             velec            = _mm256_andnot_pd(dummy_mask,velec);
 851             /*             #endif */
 852             velecsum         = _mm256_add_pd(velecsum,velec);
 853             /*             #define INNERFLOPS INNERFLOPS+1 */
 854             /*         #endif */
 855             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 856             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 857             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 858             vvdw             = _mm256_and_pd(vvdw,cutoff_mask);
 859             /*                 #define INNERFLOPS INNERFLOPS+1 */
 860             /*             #endif                                       */
 861             /*             #if ROUND == 'Epilogue' */
 862             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 863             /*             #endif */
 864             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 865             /*             #define INNERFLOPS INNERFLOPS+1 */
 866             /*         #endif */
 867             /*     #endif */
 868
 869             /*     #if 'Force' in KERNEL_VF */
 870
 871             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
 872             fscal            = _mm256_add_pd(felec,fvdw);
 873             /*             #define INNERFLOPS INNERFLOPS+1 */
 874             /*         #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
 875             fscal            = felec;
 876             /*         #elif 'vdw' in INTERACTION_FLAGS[I][J] */
 877             fscal            = fvdw;
 878             /*        #endif */
 879
 880             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 881             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 882             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 883             /*                 #define INNERFLOPS INNERFLOPS+1 */
 884             /*             #endif                                       */
 885
 886             /*             #if ROUND == 'Epilogue' */
 887             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 888             /*             #endif */
 889
 890             /* Calculate temporary vectorial force */
 891             tx               = _mm256_mul_pd(fscal,dx{I}{J});
 892             ty               = _mm256_mul_pd(fscal,dy{I}{J});
 893             tz               = _mm256_mul_pd(fscal,dz{I}{J});
 894
 895             /* Update vectorial force */
 896             fix{I}             = _mm256_add_pd(fix{I},tx);
 897             fiy{I}             = _mm256_add_pd(fiy{I},ty);
 898             fiz{I}             = _mm256_add_pd(fiz{I},tz);
 899             /*             #define INNERFLOPS INNERFLOPS+6 */
 900
 901             /* #if GEOMETRY_I == 'Particle'             */
 902             /*     #if ROUND == 'Loop' */
 903             fjptrA             = f+j_coord_offsetA;
 904             fjptrB             = f+j_coord_offsetB;
 905             fjptrC             = f+j_coord_offsetC;
 906             fjptrD             = f+j_coord_offsetD;
 907             /*     #else */
 908             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 909             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 910             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 911             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 912             /*     #endif */
 913             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 914             /*     #define INNERFLOPS INNERFLOPS+3      */
 915             /* #else                                    */
 916             fjx{J}             = _mm256_add_pd(fjx{J},tx);
 917             fjy{J}             = _mm256_add_pd(fjy{J},ty);
 918             fjz{J}             = _mm256_add_pd(fjz{J},tz);
 919             /*     #define INNERFLOPS INNERFLOPS+3      */
 920             /* #endif                                   */
 921
 922             /*     #endif */
 923
 924             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 925             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 926             /*         #if 0    ## This and next two lines is a hack to maintain indentation in template file */
 927             {
 928                 /*     #endif */
 929             }
 930             /*     #endif */
 931             /*    ## End of check for the interaction being outside the cutoff */
 932
 933             /* #endfor */
 934             /* ## End of loop over i-j interaction pairs */
 935
 936             /* #if GEOMETRY_I != 'Particle' */
 937             /*     #if ROUND == 'Loop' */
 938             fjptrA             = f+j_coord_offsetA;
 939             fjptrB             = f+j_coord_offsetB;
 940             fjptrC             = f+j_coord_offsetC;
 941             fjptrD             = f+j_coord_offsetD;
 942             /*     #else */
 943             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 944             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 945             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 946             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 947             /*     #endif */
 948             /* #endif */
 949
 950             /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
 951             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 952             /*     #define INNERFLOPS INNERFLOPS+3      */
 953             /* #elif GEOMETRY_J == 'Water3'             */
 954             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 955                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 956             /*     #define INNERFLOPS INNERFLOPS+9      */
 957             /* #elif GEOMETRY_J == 'Water4'             */
 958             /*     #if 0 in PARTICLES_J                 */
 959             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 960                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 961                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 962             /*     #define INNERFLOPS INNERFLOPS+12     */
 963             /*     #else                                */
 964             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 965                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 966             /*     #define INNERFLOPS INNERFLOPS+9      */
 967             /*     #endif                               */
 968             /* #endif                                   */
 969
 970             /* Inner loop uses {INNERFLOPS} flops */
 971         }
 972
 973         /* #endfor */
 974
 975         /* End of innermost loop */
 976
 977         /* #if 'Force' in KERNEL_VF */
 978         /*     #if GEOMETRY_I == 'Particle'            */
 979         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 980                                                  f+i_coord_offset,fshift+i_shift_offset);
 981         /*         #define OUTERFLOPS OUTERFLOPS+6     */
 982         /*     #elif GEOMETRY_I == 'Water3'            */
 983         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 984                                                  f+i_coord_offset,fshift+i_shift_offset);
 985         /*         #define OUTERFLOPS OUTERFLOPS+18    */
 986         /*     #elif GEOMETRY_I == 'Water4'            */
 987         /*         #if 0 in PARTICLES_I                */
 988         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 989                                                  f+i_coord_offset,fshift+i_shift_offset);
 990         /*             #define OUTERFLOPS OUTERFLOPS+24    */
 991         /*         #else                               */
 992         gmx_mm256_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 993                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 994         /*             #define OUTERFLOPS OUTERFLOPS+18    */
 995         /*         #endif                              */
 996         /*     #endif                                  */
 997         /* #endif                                      */
 998
 999         /* #if 'Potential' in KERNEL_VF */
1000         ggid                        = gid[iidx];
1001         /* Update potential energies */
1002         /*     #if KERNEL_ELEC != 'None' */
1003         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1004         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1005         /*     #endif */
1006         /*     #if KERNEL_VDW != 'None' */
1007         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1008         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1009         /*     #endif */
1010         /* #endif */
1011
1012         /* Increment number of inner iterations */
1013         inneriter                  += j_index_end - j_index_start;
1014
1015         /* Outer loop uses {OUTERFLOPS} flops */
1016     }
1017
1018     /* Increment number of outer iterations */
1019     outeriter        += nri;
1020
1021     /* Update outer/inner flops */
1022     /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1023     /* ## primitive and replaces aggressively even in strings inside these directives, we need to      */
1024     /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source.      */
1025     /* #if GEOMETRY_I == 'Water3'            */
1026     /*     #define ISUFFIX '_W3'             */
1027     /* #elif GEOMETRY_I == 'Water4'          */
1028     /*     #define ISUFFIX '_W4'             */
1029     /* #else                                 */
1030     /*     #define ISUFFIX ''                */
1031     /* #endif                                */
1032     /* #if GEOMETRY_J == 'Water3'            */
1033     /*     #define JSUFFIX 'W3'              */
1034     /* #elif GEOMETRY_J == 'Water4'          */
1035     /*     #define JSUFFIX 'W4'              */
1036     /* #else                                 */
1037     /*     #define JSUFFIX ''                */
1038     /* #endif                                */
1039     /* #if 'PotentialAndForce' in KERNEL_VF  */
1040     /*     #define VFSUFFIX  '_VF'           */
1041     /* #elif 'Potential' in KERNEL_VF        */
1042     /*     #define VFSUFFIX '_V'             */
1043     /* #else                                 */
1044     /*     #define VFSUFFIX '_F'             */
1045     /* #endif                                */
1046
1047     /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1048     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1049     /* #elif KERNEL_ELEC != 'None' */
1050     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1051     /* #else */
1052     inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1053     /* #endif  */
1054 }