src/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_template_sparc64_hpc_ace_double.pre

   1 /* ## */
   2 /* ## This file is part of the GROMACS molecular simulation package. */
   3 /* ## */
   4 /* ## Copyright (c) 2012, by the GROMACS development team, led by */
   5 /* ## David van der Spoel, Berk Hess, Erik Lindahl, and including many */
   6 /* ## others, as listed in the AUTHORS file in the top-level source */
   7 /* ## directory and at http://www.gromacs.org. */
   8 /* ## */
   9 /* ## GROMACS is free software; you can redistribute it and/or */
  10 /* ## modify it under the terms of the GNU Lesser General Public License */
  11 /* ## as published by the Free Software Foundation; either version 2.1 */
  12 /* ## of the License, or (at your option) any later version. */
  13 /* ## */
  14 /* ## GROMACS is distributed in the hope that it will be useful, */
  15 /* ## but WITHOUT ANY WARRANTY; without even the implied warranty of */
  16 /* ## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU */
  17 /* ## Lesser General Public License for more details. */
  18 /* ## */
  19 /* ## You should have received a copy of the GNU Lesser General Public */
  20 /* ## License along with GROMACS; if not, see */
  21 /* ## http://www.gnu.org/licenses, or write to the Free Software Foundation, */
  22 /* ## Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA. */
  23 /* ## */
  24 /* ## If you want to redistribute modifications to GROMACS, please */
  25 /* ## consider that scientific software is very special. Version */
  26 /* ## control is crucial - bugs must be traceable. We will be happy to */
  27 /* ## consider code for inclusion in the official distribution, but */
  28 /* ## derived work must not be called official GROMACS. Details are found */
  29 /* ## in the README & COPYING files - if they are missing, get the */
  30 /* ## official version at http://www.gromacs.org. */
  31 /* ## */
  32 /* ## To help us fund GROMACS development, we humbly ask that you cite */
  33 /* ## the research papers on the package. Check out http://www.gromacs.org. */
  34 /* ## */
  35 /* #if 0 */
  36 #error This file must be processed with the Gromacs pre-preprocessor
  37 /* #endif */
  38 /* #if INCLUDE_HEADER */
  39 #ifdef HAVE_CONFIG_H
  40 #include <config.h>
  41 #endif
  42
  43 #include <math.h>
  44
  45 #include "../nb_kernel.h"
  46 #include "types/simple.h"
  47 #include "vec.h"
  48 #include "nrnb.h"
  49
  50 #include "kernelutil_sparc64_hpc_ace_double.h"
  51 /* #endif */
  52
  53 /* ## List of variables set by the generating script:                                    */
  54 /* ##                                                                                    */
  55 /* ## Setttings that apply to the entire kernel:                                         */
  56 /* ## KERNEL_ELEC:           String, choice for electrostatic interactions               */
  57 /* ## KERNEL_VDW:            String, choice for van der Waals interactions               */
  58 /* ## KERNEL_NAME:           String, name of this kernel                                 */
  59 /* ## KERNEL_VF:             String telling if we calculate potential, force, or both    */
  60 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
  61 /* ##                                                                                    */
  62 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops:             */
  63 /* ## PARTICLES_I[]/         Arrays with lists of i/j particles to use in kernel. It is  */
  64 /* ## PARTICLES_J[]:         just [0] for particle geometry, but can be longer for water */
  65 /* ## PARTICLES_ELEC_I[]/    Arrays with lists of i/j particle that have electrostatics  */
  66 /* ## PARTICLES_ELEC_J[]:    interactions that should be calculated in this kernel.      */
  67 /* ## PARTICLES_VDW_I[]/     Arrays with the list of i/j particle that have VdW          */
  68 /* ## PARTICLES_VDW_J[]:     interactions that should be calculated in this kernel.      */
  69 /* ##                                                                                    */
  70 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle)   */
  71 /* ## PAIRS_IJ[]:            Array with (i,j) tuples of pairs for which interactions     */
  72 /* ##                        should be calculated in this kernel. Zero-charge particles  */
  73 /* ##                        do not have interactions with particles without vdw, and    */
  74 /* ##                        Vdw-only interactions are not evaluated in a no-vdw-kernel. */
  75 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
  76 /* ##                        For each i-j pair, the element [I][J] is a list of strings  */
  77 /* ##                        defining properties/flags of this interaction. Examples     */
  78 /* ##                        include 'electrostatics'/'vdw' if that type of interaction  */
  79 /* ##                        should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values  */
  80 /* ##                        are needed, and 'exactcutoff' or 'shift','switch' to        */
  81 /* ##                        decide if the force/potential should be modified. This way  */
  82 /* ##                        we only calculate values absolutely needed for each case.   */
  83
  84 /* ## Calculate the size and offset for (merged/interleaved) table data */
  85
  86 /*
  87  * Gromacs nonbonded kernel:   {KERNEL_NAME}
  88  * Electrostatics interaction: {KERNEL_ELEC}
  89  * VdW interaction:            {KERNEL_VDW}
  90  * Geometry:                   {GEOMETRY_I}-{GEOMETRY_J}
  91  * Calculate force/pot:        {KERNEL_VF}
  92  */
  93 void
  94 {KERNEL_NAME}
  95                     (t_nblist * gmx_restrict                nlist,
  96                      rvec * gmx_restrict                    xx,
  97                      rvec * gmx_restrict                    ff,
  98                      t_forcerec * gmx_restrict              fr,
  99                      t_mdatoms * gmx_restrict               mdatoms,
 100                      nb_kernel_data_t * gmx_restrict        kernel_data,
 101                      t_nrnb * gmx_restrict                  nrnb)
 102 {
 103     /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
 104     /* ## so there is no point in going to extremes to exclude variables that are not needed. */
 105     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 106      * just 0 for non-waters.
 107      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 108      * jnr indices corresponding to data put in the four positions in the SIMD register.
 109      */
 110     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 111     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 112     int              jnrA,jnrB;
 113     int              j_coord_offsetA,j_coord_offsetB;
 114     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 115     real             rcutoff_scalar;
 116     real             *shiftvec,*fshift,*x,*f;
 117     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 118     /* #for I in PARTICLES_I */
 119     int              vdwioffset{I};
 120     _fjsp_v2r8       ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
 121     /* #endfor */
 122     /* #for J in PARTICLES_J */
 123     int              vdwjidx{J}A,vdwjidx{J}B;
 124     _fjsp_v2r8       jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
 125     /* #endfor */
 126     /* #for I,J in PAIRS_IJ */
 127     _fjsp_v2r8       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};
 128     /* #endfor */
 129     /* #if KERNEL_ELEC != 'None' */
 130     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 131     real             *charge;
 132     /* #endif */
 133     /* #if 'GeneralizedBorn' in KERNEL_ELEC */
 134     _fjsp_v2r8       vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
 135     _fjsp_v2r8       minushalf = gmx_fjsp_set1_v2r8(-0.5);
 136     real             *invsqrta,*dvda,*gbtab;
 137     /* #endif */
 138     /* #if KERNEL_VDW != 'None' */
 139     int              nvdwtype;
 140     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 141     int              *vdwtype;
 142     real             *vdwparam;
 143     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 144     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 145     /* #endif */
 146     /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
 147     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 148     real             *vftab;
 149     /* #endif */
 150     /* #if 'Ewald' in KERNEL_ELEC */
 151     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 152     real             *ewtab;
 153     /* #endif */
 154     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 155     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 156     real             rswitch_scalar,d_scalar;
 157     /* #endif */
 158     _fjsp_v2r8       itab_tmp;
 159     _fjsp_v2r8       dummy_mask,cutoff_mask;
 160     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 161     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 162     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 163
 164     x                = xx[0];
 165     f                = ff[0];
 166
 167     nri              = nlist->nri;
 168     iinr             = nlist->iinr;
 169     jindex           = nlist->jindex;
 170     jjnr             = nlist->jjnr;
 171     shiftidx         = nlist->shift;
 172     gid              = nlist->gid;
 173     shiftvec         = fr->shift_vec[0];
 174     fshift           = fr->fshift[0];
 175     /* #if KERNEL_ELEC != 'None' */
 176     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 177     charge           = mdatoms->chargeA;
 178     /*     #if 'ReactionField' in KERNEL_ELEC */
 179     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 180     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 181     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 182     /*     #endif */
 183     /* #endif */
 184     /* #if KERNEL_VDW != 'None' */
 185     nvdwtype         = fr->ntype;
 186     vdwparam         = fr->nbfp;
 187     vdwtype          = mdatoms->typeA;
 188     /* #endif */
 189
 190     /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 191     vftab            = kernel_data->table_elec_vdw->data;
 192     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 193     /* #elif 'Table' in KERNEL_ELEC */
 194     vftab            = kernel_data->table_elec->data;
 195     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 196     /* #elif 'Table' in KERNEL_VDW */
 197     vftab            = kernel_data->table_vdw->data;
 198     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 199     /* #endif */
 200
 201     /* #if 'Ewald' in KERNEL_ELEC */
 202     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 203     /*     #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
 204     ewtab            = fr->ic->tabq_coul_F;
 205     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 206     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 207     /*     #else */
 208     ewtab            = fr->ic->tabq_coul_FDV0;
 209     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 210     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 211      /*     #endif */
 212     /* #endif */
 213
 214     /* #if KERNEL_ELEC=='GeneralizedBorn' */
 215     invsqrta         = fr->invsqrta;
 216     dvda             = fr->dvda;
 217     gbtabscale       = gmx_fjsp_set1_v2r8(fr->gbtab.scale);
 218     gbtab            = fr->gbtab.data;
 219     gbinvepsdiff     = gmx_fjsp_set1_v2r8((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 220     /* #endif */
 221
 222     /* #if 'Water' in GEOMETRY_I */
 223     /* Setup water-specific parameters */
 224     inr              = nlist->iinr[0];
 225     /*     #for I in PARTICLES_ELEC_I */
 226     iq{I}              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+{I}]));
 227     /*     #endfor */
 228     /*     #for I in PARTICLES_VDW_I */
 229     vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 230     /*     #endfor */
 231     /* #endif */
 232
 233     /* #if 'Water' in GEOMETRY_J */
 234     /*     #for J in PARTICLES_ELEC_J */
 235     jq{J}              = gmx_fjsp_set1_v2r8(charge[inr+{J}]);
 236     /*     #endfor */
 237     /*     #for J in PARTICLES_VDW_J */
 238     vdwjidx{J}A        = 2*vdwtype[inr+{J}];
 239     /*     #endfor */
 240     /*     #for I,J in PAIRS_IJ */
 241     /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 242     qq{I}{J}             = _fjsp_mul_v2r8(iq{I},jq{J});
 243     /*         #endif */
 244     /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 245     c6_{I}{J}            = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset{I}+vdwjidx{J}A]);
 246     c12_{I}{J}           = gmx_fjsp_set1_v2r8(vdwparam[vdwioffset{I}+vdwjidx{J}A+1]);
 247     /*         #endif */
 248     /*     #endfor */
 249     /* #endif */
 250
 251     /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
 252     /*     #if KERNEL_ELEC!='None' */
 253     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 254     rcutoff_scalar   = fr->rcoulomb;
 255     /*     #else */
 256     rcutoff_scalar   = fr->rvdw;
 257     /*     #endif */
 258     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 259     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 260     /* #endif */
 261
 262     /* #if KERNEL_MOD_VDW=='PotentialShift' */
 263     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 264     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 265     /* #endif */
 266
 267     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 268     /*     #if KERNEL_MOD_ELEC=='PotentialSwitch'  */
 269     rswitch_scalar   = fr->rcoulomb_switch;
 270     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 271     /*     #else */
 272     rswitch_scalar   = fr->rvdw_switch;
 273     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 274     /*     #endif */
 275     /* Setup switch parameters */
 276     d_scalar         = rcutoff_scalar-rswitch_scalar;
 277     d                = gmx_fjsp_set1_v2r8(d_scalar);
 278     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 279     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 280     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 281     /*     #if 'Force' in KERNEL_VF */
 282     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 283     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 284     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 285     /*     #endif */
 286     /* #endif */
 287
 288     /* Avoid stupid compiler warnings */
 289     jnrA = jnrB = 0;
 290     j_coord_offsetA = 0;
 291     j_coord_offsetB = 0;
 292
 293     /* ## Keep track of the floating point operations we issue for reporting! */
 294     /* #define OUTERFLOPS 0 */
 295     outeriter        = 0;
 296     inneriter        = 0;
 297
 298     /* Start outer loop over neighborlists */
 299     for(iidx=0; iidx<nri; iidx++)
 300     {
 301         /* Load shift vector for this list */
 302         i_shift_offset   = DIM*shiftidx[iidx];
 303
 304         /* Load limits for loop over neighbors */
 305         j_index_start    = jindex[iidx];
 306         j_index_end      = jindex[iidx+1];
 307
 308         /* Get outer coordinate index */
 309         inr              = iinr[iidx];
 310         i_coord_offset   = DIM*inr;
 311
 312         /* Load i particle coords and add shift vector */
 313         /* #if GEOMETRY_I == 'Particle' */
 314         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 315         /* #elif GEOMETRY_I == 'Water3' */
 316         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 317                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 318         /* #elif GEOMETRY_I == 'Water4' */
 319         /*     #if 0 in PARTICLES_I                 */
 320         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 321                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 322         /*     #else                                */
 323         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 324                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 325         /*     #endif                               */
 326         /* #endif                                   */
 327
 328         /* #if 'Force' in KERNEL_VF */
 329         /*     #for I in PARTICLES_I */
 330         fix{I}             = _fjsp_setzero_v2r8();
 331         fiy{I}             = _fjsp_setzero_v2r8();
 332         fiz{I}             = _fjsp_setzero_v2r8();
 333         /*     #endfor */
 334         /* #endif */
 335
 336         /* ## For water we already preloaded parameters at the start of the kernel */
 337         /* #if not 'Water' in GEOMETRY_I */
 338         /* Load parameters for i particles */
 339         /*     #for I in PARTICLES_ELEC_I */
 340         iq{I}              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+{I}));
 341         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 342         /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 343         isai{I}            = gmx_fjsp_load1_v2r8(invsqrta+inr+{I});
 344         /*         #endif */
 345         /*     #endfor */
 346         /*     #for I in PARTICLES_VDW_I */
 347         vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 348         /*     #endfor */
 349         /* #endif */
 350
 351         /* #if 'Potential' in KERNEL_VF */
 352         /* Reset potential sums */
 353         /*     #if KERNEL_ELEC != 'None' */
 354         velecsum         = _fjsp_setzero_v2r8();
 355         /*     #endif */
 356         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
 357         vgbsum           = _fjsp_setzero_v2r8();
 358         /*     #endif */
 359         /*     #if KERNEL_VDW != 'None' */
 360         vvdwsum          = _fjsp_setzero_v2r8();
 361         /*     #endif */
 362         /* #endif */
 363         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
 364         dvdasum          = _fjsp_setzero_v2r8();
 365         /*     #endif */
 366
 367         /* #for ROUND in ['Loop','Epilogue'] */
 368
 369         /* #if ROUND =='Loop' */
 370         /* Start inner kernel loop */
 371         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 372         {
 373         /* ## First round is normal loop (next statement resets indentation) */
 374         /*     #if 0 */
 375         }
 376         /*     #endif */
 377         /* #else */
 378         if(jidx<j_index_end)
 379         {
 380         /* ## Second round is epilogue */
 381         /* #endif */
 382         /* #define INNERFLOPS 0 */
 383
 384             /* #if ROUND =='Loop' */
 385             /* Get j neighbor index, and coordinate index */
 386             jnrA             = jjnr[jidx];
 387             jnrB             = jjnr[jidx+1];
 388             j_coord_offsetA  = DIM*jnrA;
 389             j_coord_offsetB  = DIM*jnrB;
 390
 391             /* load j atom coordinates */
 392             /*     #if GEOMETRY_J == 'Particle'             */
 393             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 394                                               &jx0,&jy0,&jz0);
 395             /*     #elif GEOMETRY_J == 'Water3'             */
 396             gmx_fjsp_load_3rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 397                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 398             /*     #elif GEOMETRY_J == 'Water4'             */
 399             /*         #if 0 in PARTICLES_J                 */
 400             gmx_fjsp_load_4rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 401                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 402                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 403             /*         #else                                */
 404             gmx_fjsp_load_3rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 405                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 406             /*         #endif                               */
 407             /*     #endif                                   */
 408             /* #else */
 409             jnrA             = jjnr[jidx];
 410             j_coord_offsetA  = DIM*jnrA;
 411
 412             /* load j atom coordinates */
 413             /*     #if GEOMETRY_J == 'Particle'             */
 414             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 415                                               &jx0,&jy0,&jz0);
 416             /*     #elif GEOMETRY_J == 'Water3'             */
 417             gmx_fjsp_load_3rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 418                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 419             /*     #elif GEOMETRY_J == 'Water4'             */
 420             /*         #if 0 in PARTICLES_J                 */
 421             gmx_fjsp_load_4rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 422                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 423                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 424             /*         #else                                */
 425             gmx_fjsp_load_3rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA+DIM,
 426                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 427             /*         #endif                               */
 428             /*     #endif                                   */
 429             /* #endif */
 430
 431             /* Calculate displacement vector */
 432             /* #for I,J in PAIRS_IJ */
 433             dx{I}{J}             = _fjsp_sub_v2r8(ix{I},jx{J});
 434             dy{I}{J}             = _fjsp_sub_v2r8(iy{I},jy{J});
 435             dz{I}{J}             = _fjsp_sub_v2r8(iz{I},jz{J});
 436             /*     #define INNERFLOPS INNERFLOPS+3 */
 437             /* #endfor */
 438
 439             /* Calculate squared distance and things based on it */
 440             /* #for I,J in PAIRS_IJ */
 441             rsq{I}{J}            = gmx_fjsp_calc_rsq_v2r8(dx{I}{J},dy{I}{J},dz{I}{J});
 442             /*     #define INNERFLOPS INNERFLOPS+5 */
 443             /* #endfor */
 444
 445             /* #for I,J in PAIRS_IJ */
 446             /*     #if 'rinv' in INTERACTION_FLAGS[I][J] */
 447             rinv{I}{J}           = gmx_fjsp_invsqrt_v2r8(rsq{I}{J});
 448             /*         #define INNERFLOPS INNERFLOPS+5 */
 449             /*     #endif */
 450             /* #endfor */
 451
 452             /* #for I,J in PAIRS_IJ */
 453             /*     #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
 454             /*         # if 'rinv' not in INTERACTION_FLAGS[I][J] */
 455             rinvsq{I}{J}         = gmx_fjsp_inv_v2r8(rsq{I}{J});
 456             /*             #define INNERFLOPS INNERFLOPS+4 */
 457             /*         #else */
 458             rinvsq{I}{J}         = _fjsp_mul_v2r8(rinv{I}{J},rinv{I}{J});
 459             /*             #define INNERFLOPS INNERFLOPS+1 */
 460             /*         #endif */
 461             /*     #endif */
 462             /* #endfor */
 463
 464             /* #if not 'Water' in GEOMETRY_J */
 465             /* Load parameters for j particles */
 466             /*     #for J in PARTICLES_ELEC_J */
 467             /*         #if ROUND =='Loop' */
 468             jq{J}              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+{J},charge+jnrB+{J});
 469             /*         #else */
 470             jq{J}              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+{J});
 471             /*         #endif */
 472             /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 473             /*             #if ROUND =='Loop' */
 474             isaj{J}            = gmx_fjsp_load_2real_swizzle_v2r8(invsqrta+jnrA+{J},invsqrta+jnrB+{J});
 475             /*             #else */
 476             isaj{J}            = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),invsqrta+jnrA+{J});
 477             /*             #endif */
 478             /*         #endif */
 479             /*     #endfor */
 480             /*     #for J in PARTICLES_VDW_J */
 481             vdwjidx{J}A        = 2*vdwtype[jnrA+{J}];
 482             /*         #if ROUND =='Loop' */
 483             vdwjidx{J}B        = 2*vdwtype[jnrB+{J}];
 484             /*         #endif */
 485             /*     #endfor */
 486             /* #endif */
 487
 488             /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
 489             /*     #for J in PARTICLES_J */
 490             fjx{J}             = _fjsp_setzero_v2r8();
 491             fjy{J}             = _fjsp_setzero_v2r8();
 492             fjz{J}             = _fjsp_setzero_v2r8();
 493             /*     #endfor */
 494             /* #endif */
 495
 496             /* #for I,J in PAIRS_IJ */
 497
 498             /**************************
 499              * CALCULATE INTERACTIONS *
 500              **************************/
 501
 502             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 503             /*         ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
 504             if (gmx_fjsp_any_lt_v2r8(rsq{I}{J},rcutoff2))
 505             {
 506                 /*     #if 0    ## this and the next two lines is a hack to maintain auto-indentation in template file */
 507             }
 508             /*         #endif */
 509             /*         #define INNERFLOPS INNERFLOPS+1 */
 510             /*     #endif */
 511
 512             /*     #if 'r' in INTERACTION_FLAGS[I][J] */
 513             r{I}{J}              = _fjsp_mul_v2r8(rsq{I}{J},rinv{I}{J});
 514              /*         #define INNERFLOPS INNERFLOPS+1 */
 515             /*     #endif */
 516
 517             /*     ## For water geometries we already loaded parameters at the start of the kernel */
 518             /*     #if not 'Water' in GEOMETRY_J */
 519             /* Compute parameters for interactions between i and j atoms */
 520             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 521             qq{I}{J}             = _fjsp_mul_v2r8(iq{I},jq{J});
 522             /*             #define INNERFLOPS INNERFLOPS+1 */
 523             /*         #endif */
 524             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 525             /*             #if ROUND == 'Loop' */
 526             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset{I}+vdwjidx{J}A,
 527                                          vdwparam+vdwioffset{I}+vdwjidx{J}B,&c6_{I}{J},&c12_{I}{J});
 528             /*             #else */
 529             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset{I}+vdwjidx{J}A,&c6_{I}{J},&c12_{I}{J});
 530             /*             #endif */
 531             /*         #endif */
 532             /*     #endif */
 533
 534             /*     #if 'table' in INTERACTION_FLAGS[I][J] */
 535             /* Calculate table index by multiplying r with table scale and truncate to integer */
 536             rt               = _fjsp_mul_v2r8(r{I}{J},vftabscale);
 537             itab_tmp         = _fjsp_dtox_v2r8(rt);
 538             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 539             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 540             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 541
 542             /*         #define INNERFLOPS INNERFLOPS+4                          */
 543             /*         #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW     */
 544             /*             ## 3 tables, 4 data per point: multiply index by 12 */
 545             vfconv.i[0]     *= 12;
 546             vfconv.i[1]     *= 12;
 547             /*         #elif 'Table' in KERNEL_ELEC                             */
 548             /*             ## 1 table, 4 data per point: multiply index by 4   */
 549             vfconv.i[0]     *= 4;
 550             vfconv.i[1]     *= 4;
 551             /*         #elif 'Table' in KERNEL_VDW                              */
 552             /*             ## 2 tables, 4 data per point: multiply index by 8  */
 553             vfconv.i[0]     *= 8;
 554             vfconv.i[1]     *= 8;
 555             /*         #endif                                                   */
 556             /*     #endif */
 557
 558             /*     ## ELECTROSTATIC INTERACTIONS */
 559             /*     #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 560
 561             /*         #if KERNEL_ELEC=='Coulomb' */
 562
 563             /* COULOMB ELECTROSTATICS */
 564             velec            = _fjsp_mul_v2r8(qq{I}{J},rinv{I}{J});
 565             /*             #define INNERFLOPS INNERFLOPS+1 */
 566             /*             #if 'Force' in KERNEL_VF */
 567             felec            = _fjsp_mul_v2r8(velec,rinvsq{I}{J});
 568             /*                 #define INNERFLOPS INNERFLOPS+2 */
 569             /*             #endif */
 570
 571             /*         #elif KERNEL_ELEC=='ReactionField' */
 572
 573             /* REACTION-FIELD ELECTROSTATICS */
 574             /*             #if 'Potential' in KERNEL_VF */
 575             velec            = _fjsp_mul_v2r8(qq{I}{J},_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq{I}{J},rinv{I}{J}),crf));
 576             /*                 #define INNERFLOPS INNERFLOPS+4 */
 577             /*             #endif */
 578             /*             #if 'Force' in KERNEL_VF */
 579             felec            = _fjsp_mul_v2r8(qq{I}{J},_fjsp_msub_v2r8(rinv{I}{J},rinvsq{I}{J},krf2));
 580             /*                 #define INNERFLOPS INNERFLOPS+3 */
 581             /*             #endif */
 582
 583             /*         #elif KERNEL_ELEC=='GeneralizedBorn' */
 584
 585             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 586             isaprod          = _fjsp_mul_v2r8(isai{I},isaj{J});
 587             gbqqfactor       = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq{I}{J},_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
 588             gbscale          = _fjsp_mul_v2r8(isaprod,gbtabscale);
 589             /*             #define INNERFLOPS INNERFLOPS+5 */
 590
 591             /* Calculate generalized born table index - this is a separate table from the normal one,
 592              * but we use the same procedure by multiplying r with scale and truncating to integer.
 593              */
 594             rt               = _fjsp_mul_v2r8(r{I}{J},gbscale);
 595             itab_tmp         = _fjsp_dtox_v2r8(rt);
 596             gbeps            = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
 597             _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
 598
 599             Y                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
 600             /*             #if ROUND == 'Loop' */
 601             F                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] );
 602             /*             #else */
 603             F                = _fjsp_setzero_v2r8();
 604             /*             #endif */
 605             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 606             G                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
 607             /*             #if ROUND == 'Loop' */
 608             H                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] +2);
 609             /*             #else */
 610             H                = _fjsp_setzero_v2r8();
 611             /*             #endif */
 612             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 613             Fp               = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
 614             VV               = _fjsp_madd_v2r8(gbeps,Fp,Y);
 615             vgb              = _fjsp_mul_v2r8(gbqqfactor,VV);
 616             /*             #define INNERFLOPS INNERFLOPS+10 */
 617
 618             /*             #if 'Force' in KERNEL_VF */
 619             twogbeps         = _fjsp_add_v2r8(gbeps,gbeps);
 620             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
 621             fgb              = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
 622             dvdatmp          = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r{I}{J},vgb));
 623             dvdasum          = _fjsp_add_v2r8(dvdasum,dvdatmp);
 624             /*             #if ROUND == 'Loop' */
 625             gmx_fjsp_increment_2real_swizzle_v2r8(dvda+jnrA,dvda+jnrB,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj{J},isaj{J})));
 626             /*             #else */
 627             gmx_fjsp_increment_1real_v2r8(dvda+jnrA,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj{J},isaj{J})));
 628             /*             #endif */
 629             /*                 #define INNERFLOPS INNERFLOPS+13 */
 630             /*             #endif */
 631             velec            = _fjsp_mul_v2r8(qq{I}{J},rinv{I}{J});
 632             /*                 #define INNERFLOPS INNERFLOPS+1 */
 633             /*             #if 'Force' in KERNEL_VF */
 634             felec            = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv{I}{J},fgb),rinv{I}{J});
 635             /*                 #define INNERFLOPS INNERFLOPS+3 */
 636             /*             #endif */
 637
 638             /*         #elif KERNEL_ELEC=='Ewald' */
 639             /* EWALD ELECTROSTATICS */
 640
 641             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 642             ewrt             = _fjsp_mul_v2r8(r{I}{J},ewtabscale);
 643             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 644             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 645             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 646
 647             /*             #define INNERFLOPS INNERFLOPS+4 */
 648             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
 649             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 650             /*                 #if ROUND == 'Loop' */
 651             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 652             /*                 #else */
 653             ewtabD           = _fjsp_setzero_v2r8();
 654             /*                 #endif */
 655             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 656             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 657             /*                 #if ROUND == 'Loop' */
 658             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 659             /*                 #else */
 660             ewtabFn          = _fjsp_setzero_v2r8();
 661             /*                 #endif */
 662             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 663             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 664             /*                 #define INNERFLOPS INNERFLOPS+2 */
 665             /*                 #if KERNEL_MOD_ELEC=='PotentialShift' */
 666             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 667             velec            = _fjsp_mul_v2r8(qq{I}{J},_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv{I}{J},sh_ewald),velec));
 668             /*                     #define INNERFLOPS INNERFLOPS+7 */
 669             /*                 #else */
 670             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 671             velec            = _fjsp_mul_v2r8(qq{I}{J},_fjsp_sub_v2r8(rinv{I}{J},velec));
 672             /*                     #define INNERFLOPS INNERFLOPS+6 */
 673             /*                 #endif */
 674             /*                 #if 'Force' in KERNEL_VF */
 675             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq{I}{J},rinv{I}{J}),_fjsp_sub_v2r8(rinvsq{I}{J},felec));
 676             /*                      #define INNERFLOPS INNERFLOPS+3 */
 677             /*                 #endif */
 678             /*             #elif KERNEL_VF=='Force' */
 679             /*                 #if ROUND == 'Loop' */
 680             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 681                                          &ewtabF,&ewtabFn);
 682             /*                 #else */
 683             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 684             /*                 #endif */
 685             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 686             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq{I}{J},rinv{I}{J}),_fjsp_sub_v2r8(rinvsq{I}{J},felec));
 687             /*                 #define INNERFLOPS INNERFLOPS+7 */
 688             /*             #endif */
 689
 690             /*         #elif KERNEL_ELEC=='CubicSplineTable' */
 691
 692             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 693             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 694             /*             #if ROUND == 'Loop' */
 695             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 696             /*             #else */
 697             F                = _fjsp_setzero_v2r8();
 698             /*             #endif */
 699             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 700             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 701             /*             #if ROUND == 'Loop' */
 702             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 703             /*             #else */
 704             H                = _fjsp_setzero_v2r8();
 705             /*             #endif */
 706             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 707             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 708             /*             #define INNERFLOPS INNERFLOPS+4 */
 709             /*             #if 'Potential' in KERNEL_VF */
 710             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 711             velec            = _fjsp_mul_v2r8(qq{I}{J},VV);
 712             /*                 #define INNERFLOPS INNERFLOPS+3 */
 713             /*             #endif */
 714             /*             #if 'Force' in KERNEL_VF */
 715             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 716             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq{I}{J},FF),_fjsp_mul_v2r8(vftabscale,rinv{I}{J})));
 717             /*                 #define INNERFLOPS INNERFLOPS+7 */
 718             /*             #endif */
 719             /*         #endif */
 720             /*         ## End of check for electrostatics interaction forms */
 721             /*     #endif */
 722             /*     ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
 723
 724             /*     #if 'vdw' in INTERACTION_FLAGS[I][J] */
 725
 726             /*         #if KERNEL_VDW=='LennardJones' */
 727
 728             /* LENNARD-JONES DISPERSION/REPULSION */
 729
 730             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 731             /*             #define INNERFLOPS INNERFLOPS+2 */
 732             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 733             vvdw6            = _fjsp_mul_v2r8(c6_{I}{J},rinvsix);
 734             vvdw12           = _fjsp_mul_v2r8(c12_{I}{J},_fjsp_mul_v2r8(rinvsix,rinvsix));
 735             /*                 #define INNERFLOPS INNERFLOPS+3 */
 736             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 737             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_{I}{J},_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 738                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_{I}{J},sh_vdw_invrcut6,vvdw6),one_sixth));
 739             /*                     #define INNERFLOPS INNERFLOPS+8 */
 740             /*                 #else */
 741             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 742             /*                     #define INNERFLOPS INNERFLOPS+3 */
 743             /*                 #endif */
 744             /*                 ## Check for force inside potential check, i.e. this means we already did the potential part */
 745             /*                 #if 'Force' in KERNEL_VF */
 746             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq{I}{J});
 747             /*                     #define INNERFLOPS INNERFLOPS+2 */
 748             /*                 #endif */
 749             /*             #elif KERNEL_VF=='Force' */
 750             /*                 ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
 751             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_{I}{J},rinvsix,c6_{I}{J}),_fjsp_mul_v2r8(rinvsix,rinvsq{I}{J}));
 752             /*                 #define INNERFLOPS INNERFLOPS+4 */
 753             /*             #endif */
 754
 755             /*         #elif KERNEL_VDW=='CubicSplineTable' */
 756
 757             /* CUBIC SPLINE TABLE DISPERSION */
 758             /*             #if 'Table' in KERNEL_ELEC */
 759             vfconv.i[0]       += 4;
 760             vfconv.i[1]       += 4;
 761             /*             #endif                     */
 762             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 763             /*             #if ROUND == 'Loop' */
 764             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 765             /*             #else */
 766             F                = _fjsp_setzero_v2r8();
 767             /*             #endif */
 768             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 769             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 770             /*             #if ROUND == 'Loop' */
 771             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 772             /*             #else */
 773             H                = _fjsp_setzero_v2r8();
 774             /*             #endif */
 775             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 776             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 777             /*             #define INNERFLOPS INNERFLOPS+4 */
 778             /*             #if 'Potential' in KERNEL_VF */
 779             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 780             vvdw6            = _fjsp_mul_v2r8(c6_{I}{J},VV);
 781             /*                 #define INNERFLOPS INNERFLOPS+3 */
 782             /*             #endif */
 783             /*             #if 'Force' in KERNEL_VF */
 784             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 785             fvdw6            = _fjsp_mul_v2r8(c6_{I}{J},FF);
 786             /*                 #define INNERFLOPS INNERFLOPS+4 */
 787             /*             #endif */
 788
 789             /* CUBIC SPLINE TABLE REPULSION */
 790             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 791             /*             #if ROUND == 'Loop' */
 792             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 793             /*             #else */
 794             F                = _fjsp_setzero_v2r8();
 795             /*             #endif */
 796             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 797             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 798             /*             #if ROUND == 'Loop' */
 799             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 800             /*             #else */
 801             H                = _fjsp_setzero_v2r8();
 802             /*             #endif */
 803             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 804             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 805             /*             #define INNERFLOPS INNERFLOPS+4 */
 806             /*             #if 'Potential' in KERNEL_VF */
 807             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 808             vvdw12           = _fjsp_mul_v2r8(c12_{I}{J},VV);
 809             /*                 #define INNERFLOPS INNERFLOPS+3 */
 810             /*             #endif */
 811             /*             #if 'Force' in KERNEL_VF */
 812             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 813             fvdw12           = _fjsp_mul_v2r8(c12_{I}{J},FF);
 814             /*                 #define INNERFLOPS INNERFLOPS+5 */
 815             /*             #endif */
 816             /*             #if 'Potential' in KERNEL_VF */
 817             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 818             /*                 #define INNERFLOPS INNERFLOPS+1 */
 819             /*             #endif */
 820             /*             #if 'Force' in KERNEL_VF */
 821             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv{I}{J})));
 822             /*                 #define INNERFLOPS INNERFLOPS+4 */
 823             /*             #endif */
 824             /*         #endif */
 825             /*         ## End of check for vdw interaction forms */
 826             /*     #endif */
 827             /*     ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
 828
 829             /*     #if 'switch' in INTERACTION_FLAGS[I][J] */
 830             d                = _fjsp_sub_v2r8(r{I}{J},rswitch);
 831             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 832             d2               = _fjsp_mul_v2r8(d,d);
 833             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 834             /*         #define INNERFLOPS INNERFLOPS+10 */
 835
 836             /*         #if 'Force' in KERNEL_VF */
 837             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 838             /*             #define INNERFLOPS INNERFLOPS+5 */
 839             /*         #endif */
 840
 841             /* Evaluate switch function */
 842             /*         #if 'Force' in KERNEL_VF */
 843             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 844             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 845             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv{I}{J},_fjsp_mul_v2r8(velec,dsw)) );
 846             /*                 #define INNERFLOPS INNERFLOPS+4 */
 847             /*             #endif */
 848             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 849             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv{I}{J},_fjsp_mul_v2r8(vvdw,dsw)) );
 850             /*                 #define INNERFLOPS INNERFLOPS+4 */
 851             /*             #endif */
 852             /*         #endif */
 853             /*         #if 'Potential' in KERNEL_VF */
 854             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 855             velec            = _fjsp_mul_v2r8(velec,sw);
 856             /*                 #define INNERFLOPS INNERFLOPS+1 */
 857             /*             #endif */
 858             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 859             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 860             /*                 #define INNERFLOPS INNERFLOPS+1 */
 861             /*             #endif */
 862             /*         #endif */
 863             /*     #endif */
 864             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 865             cutoff_mask      = _fjsp_cmplt_v2r8(rsq{I}{J},rcutoff2);
 866             /*         #define INNERFLOPS INNERFLOPS+1 */
 867             /*     #endif */
 868
 869             /*     #if 'Potential' in KERNEL_VF */
 870             /* Update potential sum for this i atom from the interaction with this j atom. */
 871             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 872             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 873             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 874             /*                 #define INNERFLOPS INNERFLOPS+1 */
 875             /*             #endif                                       */
 876             /*             #if ROUND == 'Epilogue' */
 877             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 878             /*             #endif */
 879             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 880             /*             #define INNERFLOPS INNERFLOPS+1 */
 881             /*             #if KERNEL_ELEC=='GeneralizedBorn' */
 882             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 883             vgb              = _fjsp_and_v2r8(vgb,cutoff_mask);
 884             /*                 #define INNERFLOPS INNERFLOPS+1 */
 885             /*             #endif                                       */
 886             /*             #if ROUND == 'Epilogue' */
 887             vgb              = _fjsp_unpacklo_v2r8(vgb,_fjsp_setzero_v2r8());
 888             /*             #endif */
 889             vgbsum           = _fjsp_add_v2r8(vgbsum,vgb);
 890             /*                 #define INNERFLOPS INNERFLOPS+1 */
 891             /*             #endif */
 892             /*         #endif */
 893             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 894             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 895             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 896             /*                 #define INNERFLOPS INNERFLOPS+1 */
 897             /*             #endif                                       */
 898             /*             #if ROUND == 'Epilogue' */
 899             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 900             /*             #endif */
 901             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 902             /*             #define INNERFLOPS INNERFLOPS+1 */
 903             /*         #endif */
 904             /*     #endif */
 905
 906             /*     #if 'Force' in KERNEL_VF */
 907
 908             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
 909             fscal            = _fjsp_add_v2r8(felec,fvdw);
 910             /*             #define INNERFLOPS INNERFLOPS+1 */
 911             /*         #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
 912             fscal            = felec;
 913             /*         #elif 'vdw' in INTERACTION_FLAGS[I][J] */
 914             fscal            = fvdw;
 915             /*        #endif */
 916
 917             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 918             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 919             /*                 #define INNERFLOPS INNERFLOPS+1 */
 920             /*             #endif                                       */
 921
 922             /*             #if ROUND == 'Epilogue' */
 923             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 924             /*             #endif */
 925
 926             /* ## Construction of vectorial force built into FMA instructions now */
 927             /* #define INNERFLOPS INNERFLOPS+3      */
 928
 929             /* Update vectorial force */
 930             fix{I}             = _fjsp_madd_v2r8(dx{I}{J},fscal,fix{I});
 931             fiy{I}             = _fjsp_madd_v2r8(dy{I}{J},fscal,fiy{I});
 932             fiz{I}             = _fjsp_madd_v2r8(dz{I}{J},fscal,fiz{I});
 933             /*             #define INNERFLOPS INNERFLOPS+6 */
 934
 935             /* #if GEOMETRY_I == 'Particle'             */
 936             /*     #if ROUND == 'Loop' */
 937             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx{I}{J},dy{I}{J},dz{I}{J});
 938             /*     #else */
 939             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx{I}{J},dy{I}{J},dz{I}{J});
 940             /*     #endif */
 941             /*     #define INNERFLOPS INNERFLOPS+3      */
 942             /* #else                                    */
 943             fjx{J}             = _fjsp_madd_v2r8(dx{I}{J},fscal,fjx{J});
 944             fjy{J}             = _fjsp_madd_v2r8(dy{I}{J},fscal,fjy{J});
 945             fjz{J}             = _fjsp_madd_v2r8(dz{I}{J},fscal,fjz{J});
 946             /*     #define INNERFLOPS INNERFLOPS+3      */
 947             /* #endif                                   */
 948
 949             /*     #endif */
 950
 951             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 952             /*         #if 0    ## This and next two lines is a hack to maintain indentation in template file */
 953             {
 954                 /*     #endif */
 955             }
 956             /*     #endif */
 957             /*    ## End of check for the interaction being outside the cutoff */
 958
 959             /* #endfor */
 960             /* ## End of loop over i-j interaction pairs */
 961
 962             /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
 963             /*     #if ROUND == 'Loop' */
 964             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 965             /*     #else */
 966             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 967             /*     #endif */
 968             /*     #define INNERFLOPS INNERFLOPS+3      */
 969             /* #elif GEOMETRY_J == 'Water3'             */
 970             /*     #if ROUND == 'Loop' */
 971             gmx_fjsp_decrement_3rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 972             /*     #else */
 973             gmx_fjsp_decrement_3rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 974             /*     #endif */
 975             /*     #define INNERFLOPS INNERFLOPS+9      */
 976             /* #elif GEOMETRY_J == 'Water4'             */
 977             /*     #if 0 in PARTICLES_J                 */
 978             /*         #if ROUND == 'Loop' */
 979             gmx_fjsp_decrement_4rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 980             /*         #else */
 981             gmx_fjsp_decrement_4rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 982             /*         #endif */
 983             /*         #define INNERFLOPS INNERFLOPS+12 */
 984             /*     #else                                */
 985             /*         #if ROUND == 'Loop' */
 986             gmx_fjsp_decrement_3rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 987             /*         #else */
 988             gmx_fjsp_decrement_3rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 989             /*         #endif */
 990             /*         #define INNERFLOPS INNERFLOPS+9  */
 991             /*     #endif                               */
 992             /* #endif                                   */
 993
 994             /* Inner loop uses {INNERFLOPS} flops */
 995         }
 996
 997         /* #endfor */
 998
 999         /* End of innermost loop */
1000
1001         /* #if 'Force' in KERNEL_VF */
1002         /*     #if GEOMETRY_I == 'Particle'            */
1003         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
1004                                               f+i_coord_offset,fshift+i_shift_offset);
1005         /*         #define OUTERFLOPS OUTERFLOPS+6     */
1006         /*     #elif GEOMETRY_I == 'Water3'            */
1007         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1008                                               f+i_coord_offset,fshift+i_shift_offset);
1009         /*         #define OUTERFLOPS OUTERFLOPS+18    */
1010         /*     #elif GEOMETRY_I == 'Water4'            */
1011         /*         #if 0 in PARTICLES_I                */
1012         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1013                                               f+i_coord_offset,fshift+i_shift_offset);
1014         /*             #define OUTERFLOPS OUTERFLOPS+24    */
1015         /*         #else                               */
1016         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1017                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1018         /*             #define OUTERFLOPS OUTERFLOPS+18    */
1019         /*         #endif                              */
1020         /*     #endif                                  */
1021         /* #endif                                      */
1022
1023         /* #if 'Potential' in KERNEL_VF */
1024         ggid                        = gid[iidx];
1025         /* Update potential energies */
1026         /*     #if KERNEL_ELEC != 'None' */
1027         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
1028         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1029         /*     #endif */
1030         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
1031         gmx_fjsp_update_1pot_v2r8(vgbsum,kernel_data->energygrp_polarization+ggid);
1032         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1033         /*     #endif */
1034         /*     #if KERNEL_VDW != 'None' */
1035         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
1036         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1037         /*     #endif */
1038         /* #endif */
1039         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
1040         dvdasum = _fjsp_mul_v2r8(dvdasum, _fjsp_mul_v2r8(isai{I},isai{I}));
1041         gmx_fjsp_update_1pot_v2r8(dvdasum,dvda+inr);
1042         /*     #endif */
1043
1044         /* Increment number of inner iterations */
1045         inneriter                  += j_index_end - j_index_start;
1046
1047         /* Outer loop uses {OUTERFLOPS} flops */
1048     }
1049
1050     /* Increment number of outer iterations */
1051     outeriter        += nri;
1052
1053     /* Update outer/inner flops */
1054     /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1055     /* ## primitive and replaces aggressively even in strings inside these directives, we need to      */
1056     /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source.      */
1057     /* #if GEOMETRY_I == 'Water3'            */
1058     /*     #define ISUFFIX '_W3'             */
1059     /* #elif GEOMETRY_I == 'Water4'          */
1060     /*     #define ISUFFIX '_W4'             */
1061     /* #else                                 */
1062     /*     #define ISUFFIX ''                */
1063     /* #endif                                */
1064     /* #if GEOMETRY_J == 'Water3'            */
1065     /*     #define JSUFFIX 'W3'              */
1066     /* #elif GEOMETRY_J == 'Water4'          */
1067     /*     #define JSUFFIX 'W4'              */
1068     /* #else                                 */
1069     /*     #define JSUFFIX ''                */
1070     /* #endif                                */
1071     /* #if 'PotentialAndForce' in KERNEL_VF  */
1072     /*     #define VFSUFFIX  '_VF'           */
1073     /* #elif 'Potential' in KERNEL_VF        */
1074     /*     #define VFSUFFIX '_V'             */
1075     /* #else                                 */
1076     /*     #define VFSUFFIX '_F'             */
1077     /* #endif                                */
1078
1079     /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1080     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1081     /* #elif KERNEL_ELEC != 'None' */
1082     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1083     /* #else */
1084     inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1085     /* #endif  */
1086 }