src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_template_avx_256_single.pre

   1 /* #if 0 */
   2 /*
   3  * This file is part of the GROMACS molecular simulation package.
   4  *
   5  * Copyright (c) 2012,2013,2014, 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/legacyheaders/types/simple.h"
  47 #include "gromacs/math/vec.h"
  48 #include "gromacs/legacyheaders/nrnb.h"
  49
  50 #include "gromacs/simd/math_x86_avx_256_single.h"
  51 #include "kernelutil_x86_avx_256_single.h"
  52 /* #endif */
  53
  54 /* ## List of variables set by the generating script:                                    */
  55 /* ##                                                                                    */
  56 /* ## Setttings that apply to the entire kernel:                                         */
  57 /* ## KERNEL_ELEC:           String, choice for electrostatic interactions               */
  58 /* ## KERNEL_VDW:            String, choice for van der Waals interactions               */
  59 /* ## KERNEL_NAME:           String, name of this kernel                                 */
  60 /* ## KERNEL_VF:             String telling if we calculate potential, force, or both    */
  61 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
  62 /* ##                                                                                    */
  63 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops:             */
  64 /* ## PARTICLES_I[]/         Arrays with lists of i/j particles to use in kernel. It is  */
  65 /* ## PARTICLES_J[]:         just [0] for particle geometry, but can be longer for water */
  66 /* ## PARTICLES_ELEC_I[]/    Arrays with lists of i/j particle that have electrostatics  */
  67 /* ## PARTICLES_ELEC_J[]:    interactions that should be calculated in this kernel.      */
  68 /* ## PARTICLES_VDW_I[]/     Arrays with the list of i/j particle that have VdW          */
  69 /* ## PARTICLES_VDW_J[]:     interactions that should be calculated in this kernel.      */
  70 /* ##                                                                                    */
  71 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle)   */
  72 /* ## PAIRS_IJ[]:            Array with (i,j) tuples of pairs for which interactions     */
  73 /* ##                        should be calculated in this kernel. Zero-charge particles  */
  74 /* ##                        do not have interactions with particles without vdw, and    */
  75 /* ##                        Vdw-only interactions are not evaluated in a no-vdw-kernel. */
  76 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
  77 /* ##                        For each i-j pair, the element [I][J] is a list of strings  */
  78 /* ##                        defining properties/flags of this interaction. Examples     */
  79 /* ##                        include 'electrostatics'/'vdw' if that type of interaction  */
  80 /* ##                        should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values  */
  81 /* ##                        are needed, and 'exactcutoff' or 'shift','switch' to        */
  82 /* ##                        decide if the force/potential should be modified. This way  */
  83 /* ##                        we only calculate values absolutely needed for each case.   */
  84
  85 /* ## Calculate the size and offset for (merged/interleaved) table data */
  86
  87
  88
  89
  90 /*
  91  * Gromacs nonbonded kernel:   {KERNEL_NAME}
  92  * Electrostatics interaction: {KERNEL_ELEC}
  93  * VdW interaction:            {KERNEL_VDW}
  94  * Geometry:                   {GEOMETRY_I}-{GEOMETRY_J}
  95  * Calculate force/pot:        {KERNEL_VF}
  96  */
  97 void
  98 {KERNEL_NAME}
  99                     (t_nblist                    * gmx_restrict       nlist,
 100                      rvec                        * gmx_restrict          xx,
 101                      rvec                        * gmx_restrict          ff,
 102                      t_forcerec                  * gmx_restrict          fr,
 103                      t_mdatoms                   * gmx_restrict     mdatoms,
 104                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 105                      t_nrnb                      * gmx_restrict        nrnb)
 106 {
 107     /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
 108     /* ## so there is no point in going to extremes to exclude variables that are not needed. */
 109     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 110      * just 0 for non-waters.
 111      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 112      * jnr indices corresponding to data put in the four positions in the SIMD register.
 113      */
 114     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 115     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 116     int              jnrA,jnrB,jnrC,jnrD;
 117     int              jnrE,jnrF,jnrG,jnrH;
 118     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 119     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 120     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 121     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 122     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 123     real             rcutoff_scalar;
 124     real             *shiftvec,*fshift,*x,*f;
 125     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 126     real             scratch[4*DIM];
 127     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 128     /* #for I in PARTICLES_I */
 129     real *           vdwioffsetptr{I};
 130     /* #if 'LJEwald' in KERNEL_VDW */
 131     real *           vdwgridioffsetptr{I};
 132     /* #endif                      */
 133     __m256           ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
 134     /* #endfor */
 135     /* #for J in PARTICLES_J */
 136     int              vdwjidx{J}A,vdwjidx{J}B,vdwjidx{J}C,vdwjidx{J}D,vdwjidx{J}E,vdwjidx{J}F,vdwjidx{J}G,vdwjidx{J}H;
 137     __m256           jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
 138     /* #endfor */
 139     /* #for I,J in PAIRS_IJ */
 140     __m256           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};
 141     /* #endfor */
 142     /* #if KERNEL_ELEC != 'None' */
 143     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 144     real             *charge;
 145     /* #endif */
 146     /* #if 'GeneralizedBorn' in KERNEL_ELEC */
 147     __m256i          gbitab;
 148     __m128i          gbitab_lo,gbitab_hi;
 149     __m256           vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
 150     __m256           minushalf = _mm256_set1_ps(-0.5);
 151     real             *invsqrta,*dvda,*gbtab;
 152     /* #endif */
 153     /* #if KERNEL_VDW != 'None' */
 154     int              nvdwtype;
 155     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 156     int              *vdwtype;
 157     real             *vdwparam;
 158     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 159     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 160     /* #endif */
 161     /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
 162     __m256i          vfitab;
 163     __m128i          vfitab_lo,vfitab_hi;
 164     __m128i          ifour       = _mm_set1_epi32(4);
 165     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 166     real             *vftab;
 167     /* #endif */
 168     /* #if 'LJEwald' in KERNEL_VDW */
 169     /* #for I,J in PAIRS_IJ */
 170     __m256           c6grid_{I}{J};
 171     /* #endfor */
 172     real             *vdwgridparam;
 173     __m256           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 174     __m256           one_half  = _mm256_set1_ps(0.5);
 175     __m256           minus_one = _mm256_set1_ps(-1.0);
 176     /* #endif */
 177     /* #if 'Ewald' in KERNEL_ELEC */
 178     __m256i          ewitab;
 179     __m128i          ewitab_lo,ewitab_hi;
 180     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 181     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 182     real             *ewtab;
 183     /* #endif */
 184     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 185     __m256           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 186     real             rswitch_scalar,d_scalar;
 187     /* #endif */
 188     __m256           dummy_mask,cutoff_mask;
 189     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 190     __m256           one     = _mm256_set1_ps(1.0);
 191     __m256           two     = _mm256_set1_ps(2.0);
 192     x                = xx[0];
 193     f                = ff[0];
 194
 195     nri              = nlist->nri;
 196     iinr             = nlist->iinr;
 197     jindex           = nlist->jindex;
 198     jjnr             = nlist->jjnr;
 199     shiftidx         = nlist->shift;
 200     gid              = nlist->gid;
 201     shiftvec         = fr->shift_vec[0];
 202     fshift           = fr->fshift[0];
 203     /* #if KERNEL_ELEC != 'None' */
 204     facel            = _mm256_set1_ps(fr->epsfac);
 205     charge           = mdatoms->chargeA;
 206     /*     #if 'ReactionField' in KERNEL_ELEC */
 207     krf              = _mm256_set1_ps(fr->ic->k_rf);
 208     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 209     crf              = _mm256_set1_ps(fr->ic->c_rf);
 210     /*     #endif */
 211     /* #endif */
 212     /* #if KERNEL_VDW != 'None' */
 213     nvdwtype         = fr->ntype;
 214     vdwparam         = fr->nbfp;
 215     vdwtype          = mdatoms->typeA;
 216     /* #endif */
 217     /* #if 'LJEwald' in KERNEL_VDW */
 218     vdwgridparam     = fr->ljpme_c6grid;
 219     sh_lj_ewald      = _mm256_set1_ps(fr->ic->sh_lj_ewald);
 220     ewclj            = _mm256_set1_ps(fr->ewaldcoeff_lj);
 221     ewclj2           = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
 222     /* #endif */
 223
 224     /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 225     vftab            = kernel_data->table_elec_vdw->data;
 226     vftabscale       = _mm256_set1_ps(kernel_data->table_elec_vdw->scale);
 227     /* #elif 'Table' in KERNEL_ELEC */
 228     vftab            = kernel_data->table_elec->data;
 229     vftabscale       = _mm256_set1_ps(kernel_data->table_elec->scale);
 230     /* #elif 'Table' in KERNEL_VDW */
 231     vftab            = kernel_data->table_vdw->data;
 232     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 233     /* #endif */
 234
 235     /* #if 'Ewald' in KERNEL_ELEC */
 236     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 237     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
 238     beta2            = _mm256_mul_ps(beta,beta);
 239     beta3            = _mm256_mul_ps(beta,beta2);
 240
 241     /*     #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
 242     ewtab            = fr->ic->tabq_coul_F;
 243     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 244     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 245     /*     #else */
 246     ewtab            = fr->ic->tabq_coul_FDV0;
 247     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 248     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 249      /*     #endif */
 250     /* #endif */
 251
 252     /* #if KERNEL_ELEC=='GeneralizedBorn' */
 253     invsqrta         = fr->invsqrta;
 254     dvda             = fr->dvda;
 255     gbtabscale       = _mm256_set1_ps(fr->gbtab.scale);
 256     gbtab            = fr->gbtab.data;
 257     gbinvepsdiff     = _mm256_set1_ps((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 258     /* #endif */
 259
 260     /* #if 'Water' in GEOMETRY_I */
 261     /* Setup water-specific parameters */
 262     inr              = nlist->iinr[0];
 263     /*     #for I in PARTICLES_ELEC_I */
 264     iq{I}              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+{I}]));
 265     /*     #endfor */
 266     /*     #for I in PARTICLES_VDW_I */
 267     vdwioffsetptr{I}   = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
 268     /*         #if 'LJEwald' in KERNEL_VDW */
 269     vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
 270     /*         #endif */
 271     /*     #endfor */
 272     /* #endif */
 273
 274     /* #if 'Water' in GEOMETRY_J */
 275     /*     #for J in PARTICLES_ELEC_J */
 276     jq{J}              = _mm256_set1_ps(charge[inr+{J}]);
 277     /*     #endfor */
 278     /*     #for J in PARTICLES_VDW_J */
 279     vdwjidx{J}A        = 2*vdwtype[inr+{J}];
 280     /*     #endfor */
 281     /*     #for I,J in PAIRS_IJ */
 282     /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 283     qq{I}{J}             = _mm256_mul_ps(iq{I},jq{J});
 284     /*         #endif */
 285     /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 286         /*         #if 'LJEwald' in KERNEL_VDW */
 287     c6_{I}{J}            = _mm256_set1_ps(vdwioffsetptr{I}[vdwjidx{J}A]);
 288     c12_{I}{J}           = _mm256_set1_ps(vdwioffsetptr{I}[vdwjidx{J}A+1]);
 289     c6grid_{I}{J}        = _mm256_set1_ps(vdwgridioffsetptr{I}[vdwjidx{J}A]);
 290         /*         #else */
 291     c6_{I}{J}            = _mm256_set1_ps(vdwioffsetptr{I}[vdwjidx{J}A]);
 292     c12_{I}{J}           = _mm256_set1_ps(vdwioffsetptr{I}[vdwjidx{J}A+1]);
 293         /*         #endif */
 294     /*         #endif */
 295     /*     #endfor */
 296     /* #endif */
 297
 298     /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
 299     /*     #if KERNEL_ELEC!='None' */
 300     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 301     rcutoff_scalar   = fr->rcoulomb;
 302     /*     #else */
 303     rcutoff_scalar   = fr->rvdw;
 304     /*     #endif */
 305     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 306     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 307     /* #endif */
 308
 309     /* #if KERNEL_MOD_VDW=='PotentialShift' */
 310     sh_vdw_invrcut6  = _mm256_set1_ps(fr->ic->sh_invrc6);
 311     rvdw             = _mm256_set1_ps(fr->rvdw);
 312     /* #endif */
 313
 314     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 315     /*     #if KERNEL_MOD_ELEC=='PotentialSwitch'  */
 316     rswitch_scalar   = fr->rcoulomb_switch;
 317     rswitch          = _mm256_set1_ps(rswitch_scalar);
 318     /*     #else */
 319     rswitch_scalar   = fr->rvdw_switch;
 320     rswitch          = _mm256_set1_ps(rswitch_scalar);
 321     /*     #endif */
 322     /* Setup switch parameters */
 323     d_scalar         = rcutoff_scalar-rswitch_scalar;
 324     d                = _mm256_set1_ps(d_scalar);
 325     swV3             = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 326     swV4             = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 327     swV5             = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 328     /*     #if 'Force' in KERNEL_VF */
 329     swF2             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 330     swF3             = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 331     swF4             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 332     /*     #endif */
 333     /* #endif */
 334
 335     /* Avoid stupid compiler warnings */
 336     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 337     j_coord_offsetA = 0;
 338     j_coord_offsetB = 0;
 339     j_coord_offsetC = 0;
 340     j_coord_offsetD = 0;
 341     j_coord_offsetE = 0;
 342     j_coord_offsetF = 0;
 343     j_coord_offsetG = 0;
 344     j_coord_offsetH = 0;
 345
 346     /* ## Keep track of the floating point operations we issue for reporting! */
 347     /* #define OUTERFLOPS 0 */
 348     outeriter        = 0;
 349     inneriter        = 0;
 350
 351     for(iidx=0;iidx<4*DIM;iidx++)
 352     {
 353         scratch[iidx] = 0.0;
 354     }
 355
 356     /* Start outer loop over neighborlists */
 357     for(iidx=0; iidx<nri; iidx++)
 358     {
 359         /* Load shift vector for this list */
 360         i_shift_offset   = DIM*shiftidx[iidx];
 361
 362         /* Load limits for loop over neighbors */
 363         j_index_start    = jindex[iidx];
 364         j_index_end      = jindex[iidx+1];
 365
 366         /* Get outer coordinate index */
 367         inr              = iinr[iidx];
 368         i_coord_offset   = DIM*inr;
 369
 370         /* Load i particle coords and add shift vector */
 371         /* #if GEOMETRY_I == 'Particle' */
 372         gmx_mm256_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 373         /* #elif GEOMETRY_I == 'Water3' */
 374         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 375                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 376         /* #elif GEOMETRY_I == 'Water4' */
 377         /*     #if 0 in PARTICLES_I                 */
 378         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 379                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 380         /*     #else                                */
 381         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 382                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 383         /*     #endif                               */
 384         /* #endif                                   */
 385
 386         /* #if 'Force' in KERNEL_VF */
 387         /*     #for I in PARTICLES_I */
 388         fix{I}             = _mm256_setzero_ps();
 389         fiy{I}             = _mm256_setzero_ps();
 390         fiz{I}             = _mm256_setzero_ps();
 391         /*     #endfor */
 392         /* #endif */
 393
 394         /* ## For water we already preloaded parameters at the start of the kernel */
 395         /* #if not 'Water' in GEOMETRY_I */
 396         /* Load parameters for i particles */
 397         /*     #for I in PARTICLES_ELEC_I */
 398         iq{I}              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+{I}]));
 399         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 400         /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 401         isai{I}            = _mm256_set1_ps(invsqrta[inr+{I}]);
 402         /*         #endif */
 403         /*     #endfor */
 404         /*     #for I in PARTICLES_VDW_I */
 405         vdwioffsetptr{I}   = vdwparam+2*nvdwtype*vdwtype[inr+{I}];
 406         /*         #if 'LJEwald' in KERNEL_VDW */
 407         vdwgridioffsetptr{I} = vdwgridparam+2*nvdwtype*vdwtype[inr+{I}];
 408         /*         #endif */
 409         /*     #endfor */
 410         /* #endif */
 411
 412         /* #if 'Potential' in KERNEL_VF */
 413         /* Reset potential sums */
 414         /*     #if KERNEL_ELEC != 'None' */
 415         velecsum         = _mm256_setzero_ps();
 416         /*     #endif */
 417         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
 418         vgbsum           = _mm256_setzero_ps();
 419         /*     #endif */
 420         /*     #if KERNEL_VDW != 'None' */
 421         vvdwsum          = _mm256_setzero_ps();
 422         /*     #endif */
 423         /* #endif */
 424         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
 425         dvdasum          = _mm256_setzero_ps();
 426         /*     #endif */
 427
 428         /* #for ROUND in ['Loop','Epilogue'] */
 429
 430         /* #if ROUND =='Loop' */
 431         /* Start inner kernel loop */
 432         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 433         {
 434         /* ## First round is normal loop (next statement resets indentation) */
 435         /*     #if 0 */
 436         }
 437         /*     #endif */
 438         /* #else */
 439         if(jidx<j_index_end)
 440         {
 441         /* ## Second round is epilogue */
 442         /* #endif */
 443         /* #define INNERFLOPS 0 */
 444
 445             /* Get j neighbor index, and coordinate index */
 446             /* #if ROUND =='Loop' */
 447             jnrA             = jjnr[jidx];
 448             jnrB             = jjnr[jidx+1];
 449             jnrC             = jjnr[jidx+2];
 450             jnrD             = jjnr[jidx+3];
 451             jnrE             = jjnr[jidx+4];
 452             jnrF             = jjnr[jidx+5];
 453             jnrG             = jjnr[jidx+6];
 454             jnrH             = jjnr[jidx+7];
 455             /* #else */
 456             jnrlistA         = jjnr[jidx];
 457             jnrlistB         = jjnr[jidx+1];
 458             jnrlistC         = jjnr[jidx+2];
 459             jnrlistD         = jjnr[jidx+3];
 460             jnrlistE         = jjnr[jidx+4];
 461             jnrlistF         = jjnr[jidx+5];
 462             jnrlistG         = jjnr[jidx+6];
 463             jnrlistH         = jjnr[jidx+7];
 464             /* Sign of each element will be negative for non-real atoms.
 465              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 466              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 467              */
 468             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 469                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 470
 471             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 472             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 473             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 474             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 475             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 476             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 477             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 478             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 479             /* #endif */
 480             j_coord_offsetA  = DIM*jnrA;
 481             j_coord_offsetB  = DIM*jnrB;
 482             j_coord_offsetC  = DIM*jnrC;
 483             j_coord_offsetD  = DIM*jnrD;
 484             j_coord_offsetE  = DIM*jnrE;
 485             j_coord_offsetF  = DIM*jnrF;
 486             j_coord_offsetG  = DIM*jnrG;
 487             j_coord_offsetH  = DIM*jnrH;
 488
 489             /* load j atom coordinates */
 490             /* #if GEOMETRY_J == 'Particle'             */
 491             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 492                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 493                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 494                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 495                                                  &jx0,&jy0,&jz0);
 496             /* #elif GEOMETRY_J == 'Water3'             */
 497             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 498                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 499                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 500                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 501                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 502             /* #elif GEOMETRY_J == 'Water4'             */
 503             /*     #if 0 in PARTICLES_J                 */
 504             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 505                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 506                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 507                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 508                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 509                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 510             /*     #else                                */
 511             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 512                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 513                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 514                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 515                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 516             /*     #endif                               */
 517             /* #endif                                   */
 518
 519             /* Calculate displacement vector */
 520             /* #for I,J in PAIRS_IJ */
 521             dx{I}{J}             = _mm256_sub_ps(ix{I},jx{J});
 522             dy{I}{J}             = _mm256_sub_ps(iy{I},jy{J});
 523             dz{I}{J}             = _mm256_sub_ps(iz{I},jz{J});
 524             /*     #define INNERFLOPS INNERFLOPS+3 */
 525             /* #endfor */
 526
 527             /* Calculate squared distance and things based on it */
 528             /* #for I,J in PAIRS_IJ */
 529             rsq{I}{J}            = gmx_mm256_calc_rsq_ps(dx{I}{J},dy{I}{J},dz{I}{J});
 530             /*     #define INNERFLOPS INNERFLOPS+5 */
 531             /* #endfor */
 532
 533             /* #for I,J in PAIRS_IJ */
 534             /*     #if 'rinv' in INTERACTION_FLAGS[I][J] */
 535             rinv{I}{J}           = gmx_mm256_invsqrt_ps(rsq{I}{J});
 536             /*         #define INNERFLOPS INNERFLOPS+5 */
 537             /*     #endif */
 538             /* #endfor */
 539
 540             /* #for I,J in PAIRS_IJ */
 541             /*     #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
 542             /*         # if 'rinv' not in INTERACTION_FLAGS[I][J] */
 543             rinvsq{I}{J}         = gmx_mm256_inv_ps(rsq{I}{J});
 544             /*             #define INNERFLOPS INNERFLOPS+4 */
 545             /*         #else */
 546             rinvsq{I}{J}         = _mm256_mul_ps(rinv{I}{J},rinv{I}{J});
 547             /*             #define INNERFLOPS INNERFLOPS+1 */
 548             /*         #endif */
 549             /*     #endif */
 550             /* #endfor */
 551
 552             /* #if not 'Water' in GEOMETRY_J */
 553             /* Load parameters for j particles */
 554             /*     #for J in PARTICLES_ELEC_J */
 555             jq{J}              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+{J},charge+jnrB+{J},
 556                                                                  charge+jnrC+{J},charge+jnrD+{J},
 557                                                                  charge+jnrE+{J},charge+jnrF+{J},
 558                                                                  charge+jnrG+{J},charge+jnrH+{J});
 559             /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 560             isaj{J}            = gmx_mm256_load_8real_swizzle_ps(invsqrta+jnrA+{J},invsqrta+jnrB+{J},
 561                                                                  invsqrta+jnrC+{J},invsqrta+jnrD+{J},
 562                                                                  invsqrta+jnrE+{J},invsqrta+jnrF+{J},
 563                                                                  invsqrta+jnrG+{J},invsqrta+jnrH+{J});
 564             /*         #endif */
 565             /*     #endfor */
 566             /*     #for J in PARTICLES_VDW_J */
 567             vdwjidx{J}A        = 2*vdwtype[jnrA+{J}];
 568             vdwjidx{J}B        = 2*vdwtype[jnrB+{J}];
 569             vdwjidx{J}C        = 2*vdwtype[jnrC+{J}];
 570             vdwjidx{J}D        = 2*vdwtype[jnrD+{J}];
 571             vdwjidx{J}E        = 2*vdwtype[jnrE+{J}];
 572             vdwjidx{J}F        = 2*vdwtype[jnrF+{J}];
 573             vdwjidx{J}G        = 2*vdwtype[jnrG+{J}];
 574             vdwjidx{J}H        = 2*vdwtype[jnrH+{J}];
 575             /*     #endfor */
 576             /* #endif */
 577
 578             /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
 579             /*     #for J in PARTICLES_J */
 580             fjx{J}             = _mm256_setzero_ps();
 581             fjy{J}             = _mm256_setzero_ps();
 582             fjz{J}             = _mm256_setzero_ps();
 583             /*     #endfor */
 584             /* #endif */
 585
 586             /* #for I,J in PAIRS_IJ */
 587
 588             /**************************
 589              * CALCULATE INTERACTIONS *
 590              **************************/
 591
 592             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 593             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 594             /*         ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
 595             if (gmx_mm256_any_lt(rsq{I}{J},rcutoff2))
 596             {
 597                 /*     #if 0    ## this and the next two lines is a hack to maintain auto-indentation in template file */
 598             }
 599             /*         #endif */
 600             /*         #define INNERFLOPS INNERFLOPS+1 */
 601             /*     #endif */
 602
 603             /*     #if 'r' in INTERACTION_FLAGS[I][J] */
 604             r{I}{J}              = _mm256_mul_ps(rsq{I}{J},rinv{I}{J});
 605             /*         #if ROUND == 'Epilogue' */
 606             r{I}{J}              = _mm256_andnot_ps(dummy_mask,r{I}{J});
 607             /*             #define INNERFLOPS INNERFLOPS+1 */
 608             /*         #endif */
 609             /*         #define INNERFLOPS INNERFLOPS+1 */
 610             /*     #endif */
 611
 612             /*     ## For water geometries we already loaded parameters at the start of the kernel */
 613             /*     #if not 'Water' in GEOMETRY_J */
 614             /* Compute parameters for interactions between i and j atoms */
 615             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 616             qq{I}{J}             = _mm256_mul_ps(iq{I},jq{J});
 617             /*             #define INNERFLOPS INNERFLOPS+1 */
 618             /*         #endif */
 619             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 620             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr{I}+vdwjidx{J}A,
 621                                             vdwioffsetptr{I}+vdwjidx{J}B,
 622                                             vdwioffsetptr{I}+vdwjidx{J}C,
 623                                             vdwioffsetptr{I}+vdwjidx{J}D,
 624                                             vdwioffsetptr{I}+vdwjidx{J}E,
 625                                             vdwioffsetptr{I}+vdwjidx{J}F,
 626                                             vdwioffsetptr{I}+vdwjidx{J}G,
 627                                             vdwioffsetptr{I}+vdwjidx{J}H,
 628                                             &c6_{I}{J},&c12_{I}{J});
 629
 630             /*         #if 'LJEwald' in KERNEL_VDW */
 631             c6grid_{I}{J}       = gmx_mm256_load_8real_swizzle_ps(vdwgridioffsetptr{I}+vdwjidx{J}A,
 632                                                                   vdwgridioffsetptr{I}+vdwjidx{J}B,
 633                                                                   vdwgridioffsetptr{I}+vdwjidx{J}C,
 634                                                                   vdwgridioffsetptr{I}+vdwjidx{J}D,
 635                                                                   vdwgridioffsetptr{I}+vdwjidx{J}E,
 636                                                                   vdwgridioffsetptr{I}+vdwjidx{J}F,
 637                                                                   vdwgridioffsetptr{I}+vdwjidx{J}G,
 638                                                                   vdwgridioffsetptr{I}+vdwjidx{J}H);
 639             /*          #endif */
 640             /*         #endif */
 641             /*     #endif */
 642
 643             /*     #if 'table' in INTERACTION_FLAGS[I][J] */
 644             /* Calculate table index by multiplying r with table scale and truncate to integer */
 645             rt               = _mm256_mul_ps(r{I}{J},vftabscale);
 646             vfitab           = _mm256_cvttps_epi32(rt);
 647             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 648             /*         #define INNERFLOPS INNERFLOPS+4                          */
 649             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 650             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 651             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 652             /*         #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 653             /*             ## 3 tables, 4 bytes per point: multiply index by 12 */
 654             vfitab_lo        = _mm_slli_epi32(_mm_add_epi32(vfitab_lo,_mm_slli_epi32(vfitab_lo,1)),2);
 655             vfitab_hi        = _mm_slli_epi32(_mm_add_epi32(vfitab_hi,_mm_slli_epi32(vfitab_hi,1)),2);
 656             /*         #elif 'Table' in KERNEL_ELEC                             */
 657             /*             ## 1 table, 4 bytes per point: multiply index by 4   */
 658             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 659             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 660             /*         #elif 'Table' in KERNEL_VDW */
 661             /*             ## 2 tables, 4 bytes per point: multiply index by 8  */
 662             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 663             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 664             /*         #endif                                                   */
 665             /*     #endif */
 666
 667             /*     ## ELECTROSTATIC INTERACTIONS */
 668             /*     #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 669
 670             /*         #if KERNEL_ELEC=='Coulomb' */
 671
 672             /* COULOMB ELECTROSTATICS */
 673             velec            = _mm256_mul_ps(qq{I}{J},rinv{I}{J});
 674             /*             #define INNERFLOPS INNERFLOPS+1 */
 675             /*             #if 'Force' in KERNEL_VF */
 676             felec            = _mm256_mul_ps(velec,rinvsq{I}{J});
 677             /*                 #define INNERFLOPS INNERFLOPS+1 */
 678             /*             #endif */
 679
 680             /*         #elif KERNEL_ELEC=='ReactionField' */
 681
 682             /* REACTION-FIELD ELECTROSTATICS */
 683             /*             #if 'Potential' in KERNEL_VF */
 684             velec            = _mm256_mul_ps(qq{I}{J},_mm256_sub_ps(_mm256_add_ps(rinv{I}{J},_mm256_mul_ps(krf,rsq{I}{J})),crf));
 685             /*                 #define INNERFLOPS INNERFLOPS+4 */
 686             /*             #endif */
 687             /*             #if 'Force' in KERNEL_VF */
 688             felec            = _mm256_mul_ps(qq{I}{J},_mm256_sub_ps(_mm256_mul_ps(rinv{I}{J},rinvsq{I}{J}),krf2));
 689             /*                 #define INNERFLOPS INNERFLOPS+3 */
 690             /*             #endif */
 691
 692             /*         #elif KERNEL_ELEC=='GeneralizedBorn' */
 693
 694             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 695             isaprod          = _mm256_mul_ps(isai{I},isaj{J});
 696             gbqqfactor       = _mm256_xor_ps(signbit,_mm256_mul_ps(qq{I}{J},_mm256_mul_ps(isaprod,gbinvepsdiff)));
 697             gbscale          = _mm256_mul_ps(isaprod,gbtabscale);
 698             /*             #define INNERFLOPS INNERFLOPS+5 */
 699
 700             /* Calculate generalized born table index - this is a separate table from the normal one,
 701              * but we use the same procedure by multiplying r with scale and truncating to integer.
 702              */
 703             rt               = _mm256_mul_ps(r{I}{J},gbscale);
 704             gbitab           = _mm256_cvttps_epi32(rt);
 705             gbeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 706             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 707             gbitab_lo        = _mm256_extractf128_si256(gbitab,0x0);
 708             gbitab_hi        = _mm256_extractf128_si256(gbitab,0x1);
 709             gbitab_lo        = _mm_slli_epi32(gbitab_lo,2);
 710             gbitab_hi        = _mm_slli_epi32(gbitab_hi,2);
 711             Y                = gmx_mm256_set_m128(_mm_load_ps(gbtab + _mm_extract_epi32(gbitab_hi,0)),
 712                                                   _mm_load_ps(gbtab + _mm_extract_epi32(gbitab_lo,0)));
 713             F                = gmx_mm256_set_m128(_mm_load_ps(gbtab + _mm_extract_epi32(gbitab_hi,1)),
 714                                                   _mm_load_ps(gbtab + _mm_extract_epi32(gbitab_lo,1)));
 715             G                = gmx_mm256_set_m128(_mm_load_ps(gbtab + _mm_extract_epi32(gbitab_hi,2)),
 716                                                   _mm_load_ps(gbtab + _mm_extract_epi32(gbitab_lo,2)));
 717             H                = gmx_mm256_set_m128(_mm_load_ps(gbtab + _mm_extract_epi32(gbitab_hi,3)),
 718                                                   _mm_load_ps(gbtab + _mm_extract_epi32(gbitab_lo,3)));
 719             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 720             Heps             = _mm256_mul_ps(gbeps,H);
 721             Fp               = _mm256_add_ps(F,_mm256_mul_ps(gbeps,_mm256_add_ps(G,Heps)));
 722             VV               = _mm256_add_ps(Y,_mm256_mul_ps(gbeps,Fp));
 723             vgb              = _mm256_mul_ps(gbqqfactor,VV);
 724             /*             #define INNERFLOPS INNERFLOPS+10 */
 725
 726             /*             #if 'Force' in KERNEL_VF */
 727             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(gbeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 728             fgb              = _mm256_mul_ps(gbqqfactor,_mm256_mul_ps(FF,gbscale));
 729             dvdatmp          = _mm256_mul_ps(minushalf,_mm256_add_ps(vgb,_mm256_mul_ps(fgb,r{I}{J})));
 730             /*                 #if ROUND == 'Epilogue' */
 731             dvdatmp          = _mm256_andnot_ps(dummy_mask,dvdatmp);
 732             /*                 #endif */
 733             dvdasum          = _mm256_add_ps(dvdasum,dvdatmp);
 734             /*                 #if ROUND == 'Loop' */
 735             fjptrA           = dvda+jnrA;
 736             fjptrB           = dvda+jnrB;
 737             fjptrC           = dvda+jnrC;
 738             fjptrD           = dvda+jnrD;
 739             fjptrE           = dvda+jnrE;
 740             fjptrF           = dvda+jnrF;
 741             fjptrG           = dvda+jnrG;
 742             fjptrH           = dvda+jnrH;
 743             /*                 #else */
 744             /* The pointers to scratch make sure that this code with compilers that take gmx_restrict seriously (e.g. icc 13) really can't screw things up. */
 745             fjptrA             = (jnrlistA>=0) ? dvda+jnrA : scratch;
 746             fjptrB             = (jnrlistB>=0) ? dvda+jnrB : scratch;
 747             fjptrC             = (jnrlistC>=0) ? dvda+jnrC : scratch;
 748             fjptrD             = (jnrlistD>=0) ? dvda+jnrD : scratch;
 749             fjptrE             = (jnrlistE>=0) ? dvda+jnrE : scratch;
 750             fjptrF             = (jnrlistF>=0) ? dvda+jnrF : scratch;
 751             fjptrG             = (jnrlistG>=0) ? dvda+jnrG : scratch;
 752             fjptrH             = (jnrlistH>=0) ? dvda+jnrH : scratch;
 753             /*                 #endif */
 754             gmx_mm256_increment_8real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 755                                                  _mm256_mul_ps(dvdatmp,_mm256_mul_ps(isaj{J},isaj{J})));
 756             /*                 #define INNERFLOPS INNERFLOPS+12 */
 757             /*             #endif */
 758             velec            = _mm256_mul_ps(qq{I}{J},rinv{I}{J});
 759             /*                 #define INNERFLOPS INNERFLOPS+1 */
 760             /*             #if 'Force' in KERNEL_VF */
 761             felec            = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(velec,rinv{I}{J}),fgb),rinv{I}{J});
 762             /*                 #define INNERFLOPS INNERFLOPS+3 */
 763             /*             #endif */
 764
 765             /*         #elif KERNEL_ELEC=='Ewald' */
 766             /* EWALD ELECTROSTATICS */
 767
 768             /* Analytical PME correction */
 769             zeta2            = _mm256_mul_ps(beta2,rsq{I}{J});
 770             /*             #if 'Force' in KERNEL_VF */
 771             rinv3            = _mm256_mul_ps(rinvsq{I}{J},rinv{I}{J});
 772             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 773             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 774             felec            = _mm256_mul_ps(qq{I}{J},felec);
 775             /*                 #define INNERFLOPS INNERFLOPS+31 */
 776             /*             #endif */
 777             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
 778             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 779             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 780             /*                 #define INNERFLOPS INNERFLOPS+27 */
 781             /*                 #if KERNEL_MOD_ELEC=='PotentialShift' */
 782             velec            = _mm256_sub_ps(_mm256_sub_ps(rinv{I}{J},sh_ewald),pmecorrV);
 783             /*                     #define INNERFLOPS INNERFLOPS+21 */
 784             /*                 #else */
 785             velec            = _mm256_sub_ps(rinv{I}{J},pmecorrV);
 786             /*                 #endif */
 787             velec            = _mm256_mul_ps(qq{I}{J},velec);
 788             /*             #endif */
 789
 790             /*         #elif KERNEL_ELEC=='CubicSplineTable' */
 791
 792             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 793             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 794                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 795             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 796                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 797             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 798                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 799             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 800                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 801             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 802             Heps             = _mm256_mul_ps(vfeps,H);
 803             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 804             /*             #define INNERFLOPS INNERFLOPS+4 */
 805             /*             #if 'Potential' in KERNEL_VF */
 806             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 807             velec            = _mm256_mul_ps(qq{I}{J},VV);
 808             /*                 #define INNERFLOPS INNERFLOPS+3 */
 809             /*             #endif */
 810             /*             #if 'Force' in KERNEL_VF */
 811             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 812             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq{I}{J},FF),_mm256_mul_ps(vftabscale,rinv{I}{J})));
 813             /*                 #define INNERFLOPS INNERFLOPS+7 */
 814             /*             #endif */
 815             /*         #endif */
 816             /*         ## End of check for electrostatics interaction forms */
 817             /*     #endif */
 818             /*     ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
 819
 820             /*     #if 'vdw' in INTERACTION_FLAGS[I][J] */
 821
 822             /*         #if KERNEL_VDW=='LennardJones' */
 823
 824             /* LENNARD-JONES DISPERSION/REPULSION */
 825
 826             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 827             /*             #define INNERFLOPS INNERFLOPS+2 */
 828             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 829             vvdw6            = _mm256_mul_ps(c6_{I}{J},rinvsix);
 830             vvdw12           = _mm256_mul_ps(c12_{I}{J},_mm256_mul_ps(rinvsix,rinvsix));
 831             /*                 #define INNERFLOPS INNERFLOPS+3 */
 832             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 833             vvdw             = _mm256_sub_ps(_mm256_mul_ps( _mm256_sub_ps(vvdw12 , _mm256_mul_ps(c12_{I}{J},_mm256_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 834                                           _mm256_mul_ps( _mm256_sub_ps(vvdw6,_mm256_mul_ps(c6_{I}{J},sh_vdw_invrcut6)),one_sixth));
 835             /*                     #define INNERFLOPS INNERFLOPS+8 */
 836             /*                 #else */
 837             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 838             /*                     #define INNERFLOPS INNERFLOPS+3 */
 839             /*                 #endif */
 840             /*                 ## Check for force inside potential check, i.e. this means we already did the potential part */
 841             /*                 #if 'Force' in KERNEL_VF */
 842             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq{I}{J});
 843             /*                     #define INNERFLOPS INNERFLOPS+2 */
 844             /*                 #endif */
 845             /*             #elif KERNEL_VF=='Force' */
 846             /*                 ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
 847             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_{I}{J},rinvsix),c6_{I}{J}),_mm256_mul_ps(rinvsix,rinvsq{I}{J}));
 848             /*                 #define INNERFLOPS INNERFLOPS+4 */
 849             /*             #endif */
 850
 851             /*         #elif KERNEL_VDW=='CubicSplineTable' */
 852
 853             /* CUBIC SPLINE TABLE DISPERSION */
 854             /*             #if 'Table' in KERNEL_ELEC */
 855             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 856             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 857             /*             #endif                     */
 858             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 859                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 860             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 861                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 862             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 863                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 864             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 865                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 866             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 867             Heps             = _mm256_mul_ps(vfeps,H);
 868             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 869             /*             #define INNERFLOPS INNERFLOPS+4 */
 870             /*             #if 'Potential' in KERNEL_VF */
 871             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 872             vvdw6            = _mm256_mul_ps(c6_{I}{J},VV);
 873             /*                 #define INNERFLOPS INNERFLOPS+3 */
 874             /*             #endif */
 875             /*             #if 'Force' in KERNEL_VF */
 876             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 877             fvdw6            = _mm256_mul_ps(c6_{I}{J},FF);
 878             /*                 #define INNERFLOPS INNERFLOPS+4 */
 879             /*             #endif */
 880
 881             /* CUBIC SPLINE TABLE REPULSION */
 882             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 883             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 884             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 885                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 886             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 887                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 888             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 889                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 890             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 891                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 892             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 893             Heps             = _mm256_mul_ps(vfeps,H);
 894             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 895             /*             #define INNERFLOPS INNERFLOPS+4 */
 896             /*             #if 'Potential' in KERNEL_VF */
 897             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 898             vvdw12           = _mm256_mul_ps(c12_{I}{J},VV);
 899             /*                 #define INNERFLOPS INNERFLOPS+3 */
 900             /*             #endif */
 901             /*             #if 'Force' in KERNEL_VF */
 902             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 903             fvdw12           = _mm256_mul_ps(c12_{I}{J},FF);
 904             /*                 #define INNERFLOPS INNERFLOPS+5 */
 905             /*             #endif */
 906             /*             #if 'Potential' in KERNEL_VF */
 907             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 908             /*                 #define INNERFLOPS INNERFLOPS+1 */
 909             /*             #endif */
 910             /*             #if 'Force' in KERNEL_VF */
 911             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv{I}{J})));
 912             /*                 #define INNERFLOPS INNERFLOPS+4 */
 913             /*             #endif */
 914
 915             /*         #elif KERNEL_VDW=='LJEwald' */
 916
 917             /* Analytical LJ-PME */
 918             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 919             ewcljrsq         = _mm256_mul_ps(ewclj2,rsq{I}{J});
 920             ewclj6           = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
 921             exponent         = gmx_simd_exp_r(ewcljrsq);
 922             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 923             poly             = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
 924             /*                 #define INNERFLOPS INNERFLOPS+11 */
 925             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 926             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 927             vvdw6            = _mm256_mul_ps(_mm256_sub_ps(c6_{I}{J},_mm256_mul_ps(c6grid_{I}{J},_mm256_sub_ps(one,poly))),rinvsix);
 928             vvdw12           = _mm256_mul_ps(c12_{I}{J},_mm256_mul_ps(rinvsix,rinvsix));
 929             /*                 #define INNERFLOPS INNERFLOPS+6 */
 930             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 931             vvdw             = _mm256_sub_ps(_mm256_mul_ps( _mm256_sub_ps(vvdw12 , _mm256_mul_ps(c12_{I}{J},_mm256_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 932                                           _mm256_mul_ps( _mm256_sub_ps(vvdw6,_mm256_add_ps(_mm256_mul_ps(c6_{I}{J},sh_vdw_invrcut6),_mm256_mul_ps(c6grid_{I}{J},sh_lj_ewald))),one_sixth));
 933             /*                     #define INNERFLOPS INNERFLOPS+10 */
 934             /*                 #else */
 935             vvdw             = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
 936             /*                 #define INNERFLOPS INNERFLOPS+3 */
 937             /*                 #endif */
 938             /*                  ## Check for force inside potential check, i.e. this means we already did the potential part */
 939             /*                  #if 'Force' in KERNEL_VF */
 940             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 941             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_{I}{J},one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq{I}{J});
 942             /*                 #define INNERFLOPS INNERFLOPS+6 */
 943             /*                  #endif */
 944             /*              #elif KERNEL_VF=='Force' */
 945             /* f6A = 6 * C6grid * (1 - poly) */
 946             f6A              = _mm256_mul_ps(c6grid_{I}{J},_mm256_sub_ps(one,poly));
 947             /* f6B = C6grid * exponent * beta^6 */
 948             f6B              = _mm256_mul_ps(_mm256_mul_ps(c6grid_{I}{J},one_sixth),_mm256_mul_ps(exponent,ewclj6));
 949             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 950             fvdw              = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_{I}{J},rinvsix),_mm256_sub_ps(c6_{I}{J},f6A)),rinvsix),f6B),rinvsq{I}{J});
 951             /*                 #define INNERFLOPS INNERFLOPS+11 */
 952             /*              #endif */
 953             /*         #endif */
 954             /*         ## End of check for vdw interaction forms */
 955             /*     #endif */
 956             /*     ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
 957
 958             /*     #if 'switch' in INTERACTION_FLAGS[I][J] */
 959             d                = _mm256_sub_ps(r{I}{J},rswitch);
 960             d                = _mm256_max_ps(d,_mm256_setzero_ps());
 961             d2               = _mm256_mul_ps(d,d);
 962             sw               = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
 963             /*         #define INNERFLOPS INNERFLOPS+10 */
 964
 965             /*         #if 'Force' in KERNEL_VF */
 966             dsw              = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
 967             /*             #define INNERFLOPS INNERFLOPS+5 */
 968             /*         #endif */
 969
 970             /* Evaluate switch function */
 971             /*         #if 'Force' in KERNEL_VF */
 972             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 973             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 974             felec            = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv{I}{J},_mm256_mul_ps(velec,dsw)) );
 975             /*                 #define INNERFLOPS INNERFLOPS+4 */
 976             /*             #endif */
 977             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 978             fvdw             = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv{I}{J},_mm256_mul_ps(vvdw,dsw)) );
 979             /*                 #define INNERFLOPS INNERFLOPS+4 */
 980             /*             #endif */
 981             /*         #endif */
 982             /*         #if 'Potential' in KERNEL_VF */
 983             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 984             velec            = _mm256_mul_ps(velec,sw);
 985             /*                 #define INNERFLOPS INNERFLOPS+1 */
 986             /*             #endif */
 987             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 988             vvdw             = _mm256_mul_ps(vvdw,sw);
 989             /*                 #define INNERFLOPS INNERFLOPS+1 */
 990             /*             #endif */
 991             /*         #endif */
 992             /*     #endif */
 993             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 994             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 995             cutoff_mask      = _mm256_cmp_ps(rsq{I}{J},rcutoff2,_CMP_LT_OQ);
 996             /*         #define INNERFLOPS INNERFLOPS+1 */
 997             /*     #endif */
 998
 999             /*     #if 'Potential' in KERNEL_VF */
1000             /* Update potential sum for this i atom from the interaction with this j atom. */
1001             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
1002             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
1003             velec            = _mm256_and_ps(velec,cutoff_mask);
1004             /*                 #define INNERFLOPS INNERFLOPS+1 */
1005             /*             #endif                                       */
1006             /*             #if ROUND == 'Epilogue' */
1007             velec            = _mm256_andnot_ps(dummy_mask,velec);
1008             /*             #endif */
1009             velecsum         = _mm256_add_ps(velecsum,velec);
1010             /*             #define INNERFLOPS INNERFLOPS+1 */
1011             /*             #if KERNEL_ELEC=='GeneralizedBorn' */
1012             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
1013             vgb              = _mm256_and_ps(vgb,cutoff_mask);
1014             /*                 #define INNERFLOPS INNERFLOPS+1 */
1015             /*             #endif                                       */
1016             /*             #if ROUND == 'Epilogue' */
1017             vgb              = _mm256_andnot_ps(dummy_mask,vgb);
1018             /*             #endif */
1019             vgbsum           = _mm256_add_ps(vgbsum,vgb);
1020             /*                 #define INNERFLOPS INNERFLOPS+1 */
1021             /*             #endif */
1022             /*         #endif */
1023             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
1024             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
1025             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
1026             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
1027             /*                 #define INNERFLOPS INNERFLOPS+1 */
1028             /*             #endif                                       */
1029             /*             #if ROUND == 'Epilogue' */
1030             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
1031             /*             #endif */
1032             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
1033             /*             #define INNERFLOPS INNERFLOPS+1 */
1034             /*         #endif */
1035             /*     #endif */
1036
1037             /*     #if 'Force' in KERNEL_VF */
1038
1039             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
1040             fscal            = _mm256_add_ps(felec,fvdw);
1041             /*             #define INNERFLOPS INNERFLOPS+1 */
1042             /*         #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
1043             fscal            = felec;
1044             /*         #elif 'vdw' in INTERACTION_FLAGS[I][J] */
1045             fscal            = fvdw;
1046             /*        #endif */
1047
1048             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
1049             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
1050             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1051             /*                 #define INNERFLOPS INNERFLOPS+1 */
1052             /*             #endif                                       */
1053
1054             /*             #if ROUND == 'Epilogue' */
1055             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1056             /*             #endif */
1057
1058             /* Calculate temporary vectorial force */
1059             tx               = _mm256_mul_ps(fscal,dx{I}{J});
1060             ty               = _mm256_mul_ps(fscal,dy{I}{J});
1061             tz               = _mm256_mul_ps(fscal,dz{I}{J});
1062
1063             /* Update vectorial force */
1064             fix{I}             = _mm256_add_ps(fix{I},tx);
1065             fiy{I}             = _mm256_add_ps(fiy{I},ty);
1066             fiz{I}             = _mm256_add_ps(fiz{I},tz);
1067             /*             #define INNERFLOPS INNERFLOPS+6 */
1068
1069             /* #if GEOMETRY_I == 'Particle'             */
1070             /*     #if ROUND == 'Loop' */
1071             fjptrA             = f+j_coord_offsetA;
1072             fjptrB             = f+j_coord_offsetB;
1073             fjptrC             = f+j_coord_offsetC;
1074             fjptrD             = f+j_coord_offsetD;
1075             fjptrE             = f+j_coord_offsetE;
1076             fjptrF             = f+j_coord_offsetF;
1077             fjptrG             = f+j_coord_offsetG;
1078             fjptrH             = f+j_coord_offsetH;
1079             /*     #else */
1080             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1081             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1082             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1083             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1084             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1085             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1086             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1087             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1088             /*     #endif */
1089             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
1090             /*     #define INNERFLOPS INNERFLOPS+3      */
1091             /* #else                                    */
1092             fjx{J}             = _mm256_add_ps(fjx{J},tx);
1093             fjy{J}             = _mm256_add_ps(fjy{J},ty);
1094             fjz{J}             = _mm256_add_ps(fjz{J},tz);
1095             /*     #define INNERFLOPS INNERFLOPS+3      */
1096             /* #endif                                   */
1097
1098             /*     #endif */
1099
1100             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
1101             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
1102             /*         #if 0    ## This and next two lines is a hack to maintain indentation in template file */
1103             {
1104                 /*     #endif */
1105             }
1106             /*     #endif */
1107             /*    ## End of check for the interaction being outside the cutoff */
1108
1109             /* #endfor */
1110             /* ## End of loop over i-j interaction pairs */
1111
1112             /* #if GEOMETRY_I != 'Particle' */
1113             /*     #if ROUND == 'Loop' */
1114             fjptrA             = f+j_coord_offsetA;
1115             fjptrB             = f+j_coord_offsetB;
1116             fjptrC             = f+j_coord_offsetC;
1117             fjptrD             = f+j_coord_offsetD;
1118             fjptrE             = f+j_coord_offsetE;
1119             fjptrF             = f+j_coord_offsetF;
1120             fjptrG             = f+j_coord_offsetG;
1121             fjptrH             = f+j_coord_offsetH;
1122             /*     #else */
1123             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1124             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1125             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1126             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1127             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1128             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1129             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1130             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1131             /*     #endif */
1132             /* #endif */
1133
1134             /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
1135             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1136             /*     #define INNERFLOPS INNERFLOPS+3      */
1137             /* #elif GEOMETRY_J == 'Water3'             */
1138             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1139                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1140             /*     #define INNERFLOPS INNERFLOPS+9      */
1141             /* #elif GEOMETRY_J == 'Water4'             */
1142             /*     #if 0 in PARTICLES_J                 */
1143             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1144                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1145                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1146             /*     #define INNERFLOPS INNERFLOPS+12     */
1147             /*     #else                                */
1148             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1149                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1150                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1151             /*     #define INNERFLOPS INNERFLOPS+9      */
1152             /*     #endif                               */
1153             /* #endif                                   */
1154
1155             /* Inner loop uses {INNERFLOPS} flops */
1156         }
1157
1158         /* #endfor */
1159
1160         /* End of innermost loop */
1161
1162         /* #if 'Force' in KERNEL_VF */
1163         /*     #if GEOMETRY_I == 'Particle'            */
1164         gmx_mm256_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
1165                                                  f+i_coord_offset,fshift+i_shift_offset);
1166         /*         #define OUTERFLOPS OUTERFLOPS+6     */
1167         /*     #elif GEOMETRY_I == 'Water3'            */
1168         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1169                                                  f+i_coord_offset,fshift+i_shift_offset);
1170         /*         #define OUTERFLOPS OUTERFLOPS+18    */
1171         /*     #elif GEOMETRY_I == 'Water4'            */
1172         /*         #if 0 in PARTICLES_I                */
1173         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1174                                                  f+i_coord_offset,fshift+i_shift_offset);
1175         /*             #define OUTERFLOPS OUTERFLOPS+24    */
1176         /*         #else                               */
1177         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1178                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1179         /*             #define OUTERFLOPS OUTERFLOPS+18    */
1180         /*         #endif                              */
1181         /*     #endif                                  */
1182         /* #endif                                      */
1183
1184         /* #if 'Potential' in KERNEL_VF */
1185         ggid                        = gid[iidx];
1186         /* Update potential energies */
1187         /*     #if KERNEL_ELEC != 'None' */
1188         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1189         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1190         /*     #endif */
1191         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
1192         gmx_mm256_update_1pot_ps(vgbsum,kernel_data->energygrp_polarization+ggid);
1193         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1194         /*     #endif */
1195         /*     #if KERNEL_VDW != 'None' */
1196         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1197         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1198         /*     #endif */
1199         /* #endif */
1200         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
1201         dvdasum = _mm256_mul_ps(dvdasum, _mm256_mul_ps(isai{I},isai{I}));
1202         gmx_mm256_update_1pot_ps(dvdasum,dvda+inr);
1203         /*     #endif */
1204
1205         /* Increment number of inner iterations */
1206         inneriter                  += j_index_end - j_index_start;
1207
1208         /* Outer loop uses {OUTERFLOPS} flops */
1209     }
1210
1211     /* Increment number of outer iterations */
1212     outeriter        += nri;
1213
1214     /* Update outer/inner flops */
1215     /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1216     /* ## primitive and replaces aggressively even in strings inside these directives, we need to      */
1217     /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source.      */
1218     /* #if GEOMETRY_I == 'Water3'            */
1219     /*     #define ISUFFIX '_W3'             */
1220     /* #elif GEOMETRY_I == 'Water4'          */
1221     /*     #define ISUFFIX '_W4'             */
1222     /* #else                                 */
1223     /*     #define ISUFFIX ''                */
1224     /* #endif                                */
1225     /* #if GEOMETRY_J == 'Water3'            */
1226     /*     #define JSUFFIX 'W3'              */
1227     /* #elif GEOMETRY_J == 'Water4'          */
1228     /*     #define JSUFFIX 'W4'              */
1229     /* #else                                 */
1230     /*     #define JSUFFIX ''                */
1231     /* #endif                                */
1232     /* #if 'PotentialAndForce' in KERNEL_VF  */
1233     /*     #define VFSUFFIX  '_VF'           */
1234     /* #elif 'Potential' in KERNEL_VF        */
1235     /*     #define VFSUFFIX '_V'             */
1236     /* #else                                 */
1237     /*     #define VFSUFFIX '_F'             */
1238     /* #endif                                */
1239
1240     /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1241     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1242     /* #elif KERNEL_ELEC != 'None' */
1243     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1244     /* #else */
1245     inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1246     /* #endif  */
1247 }