src/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_template_sse4_1_single.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 "gmx_math_x86_sse4_1_single.h"
  51 #include "kernelutil_x86_sse4_1_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  * Gromacs nonbonded kernel:   {KERNEL_NAME}
  89  * Electrostatics interaction: {KERNEL_ELEC}
  90  * VdW interaction:            {KERNEL_VDW}
  91  * Geometry:                   {GEOMETRY_I}-{GEOMETRY_J}
  92  * Calculate force/pot:        {KERNEL_VF}
  93  */
  94 void
  95 {KERNEL_NAME}
  96                     (t_nblist * gmx_restrict                nlist,
  97                      rvec * gmx_restrict                    xx,
  98                      rvec * gmx_restrict                    ff,
  99                      t_forcerec * gmx_restrict              fr,
 100                      t_mdatoms * gmx_restrict               mdatoms,
 101                      nb_kernel_data_t * gmx_restrict        kernel_data,
 102                      t_nrnb * gmx_restrict                  nrnb)
 103 {
 104     /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
 105     /* ## so there is no point in going to extremes to exclude variables that are not needed. */
 106     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 107      * just 0 for non-waters.
 108      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
 109      * jnr indices corresponding to data put in the four positions in the SIMD register.
 110      */
 111     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 112     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 113     int              jnrA,jnrB,jnrC,jnrD;
 114     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 115     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 116     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 117     real             rcutoff_scalar;
 118     real             *shiftvec,*fshift,*x,*f;
 119     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 120     real             scratch[4*DIM];
 121     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 122     /* #for I in PARTICLES_I */
 123     int              vdwioffset{I};
 124     __m128           ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
 125     /* #endfor */
 126     /* #for J in PARTICLES_J */
 127     int              vdwjidx{J}A,vdwjidx{J}B,vdwjidx{J}C,vdwjidx{J}D;
 128     __m128           jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
 129     /* #endfor */
 130     /* #for I,J in PAIRS_IJ */
 131     __m128           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};
 132     /* #endfor */
 133     /* #if KERNEL_ELEC != 'None' */
 134     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 135     real             *charge;
 136     /* #endif */
 137     /* #if 'GeneralizedBorn' in KERNEL_ELEC */
 138     __m128i          gbitab;
 139     __m128           vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
 140     __m128           minushalf = _mm_set1_ps(-0.5);
 141     real             *invsqrta,*dvda,*gbtab;
 142     /* #endif */
 143     /* #if KERNEL_VDW != 'None' */
 144     int              nvdwtype;
 145     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 146     int              *vdwtype;
 147     real             *vdwparam;
 148     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 149     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 150     /* #endif */
 151     /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
 152     __m128i          vfitab;
 153     __m128i          ifour       = _mm_set1_epi32(4);
 154     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 155     real             *vftab;
 156     /* #endif */
 157     /* #if 'Ewald' in KERNEL_ELEC */
 158     __m128i          ewitab;
 159     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 160     real             *ewtab;
 161     /* #endif */
 162     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 163     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 164     real             rswitch_scalar,d_scalar;
 165     /* #endif */
 166     __m128           dummy_mask,cutoff_mask;
 167     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 168     __m128           one     = _mm_set1_ps(1.0);
 169     __m128           two     = _mm_set1_ps(2.0);
 170     x                = xx[0];
 171     f                = ff[0];
 172
 173     nri              = nlist->nri;
 174     iinr             = nlist->iinr;
 175     jindex           = nlist->jindex;
 176     jjnr             = nlist->jjnr;
 177     shiftidx         = nlist->shift;
 178     gid              = nlist->gid;
 179     shiftvec         = fr->shift_vec[0];
 180     fshift           = fr->fshift[0];
 181     /* #if KERNEL_ELEC != 'None' */
 182     facel            = _mm_set1_ps(fr->epsfac);
 183     charge           = mdatoms->chargeA;
 184     /*     #if 'ReactionField' in KERNEL_ELEC */
 185     krf              = _mm_set1_ps(fr->ic->k_rf);
 186     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
 187     crf              = _mm_set1_ps(fr->ic->c_rf);
 188     /*     #endif */
 189     /* #endif */
 190     /* #if KERNEL_VDW != 'None' */
 191     nvdwtype         = fr->ntype;
 192     vdwparam         = fr->nbfp;
 193     vdwtype          = mdatoms->typeA;
 194     /* #endif */
 195
 196     /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 197     vftab            = kernel_data->table_elec_vdw->data;
 198     vftabscale       = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
 199     /* #elif 'Table' in KERNEL_ELEC */
 200     vftab            = kernel_data->table_elec->data;
 201     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
 202     /* #elif 'Table' in KERNEL_VDW */
 203     vftab            = kernel_data->table_vdw->data;
 204     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 205     /* #endif */
 206
 207     /* #if 'Ewald' in KERNEL_ELEC */
 208     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 209     /*     #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
 210     ewtab            = fr->ic->tabq_coul_F;
 211     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 212     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 213     /*     #else */
 214     ewtab            = fr->ic->tabq_coul_FDV0;
 215     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 216     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 217      /*     #endif */
 218     /* #endif */
 219
 220     /* #if KERNEL_ELEC=='GeneralizedBorn' */
 221     invsqrta         = fr->invsqrta;
 222     dvda             = fr->dvda;
 223     gbtabscale       = _mm_set1_ps(fr->gbtab.scale);
 224     gbtab            = fr->gbtab.data;
 225     gbinvepsdiff     = _mm_set1_ps((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 226     /* #endif */
 227
 228     /* #if 'Water' in GEOMETRY_I */
 229     /* Setup water-specific parameters */
 230     inr              = nlist->iinr[0];
 231     /*     #for I in PARTICLES_ELEC_I */
 232     iq{I}              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+{I}]));
 233     /*     #endfor */
 234     /*     #for I in PARTICLES_VDW_I */
 235     vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 236     /*     #endfor */
 237     /* #endif */
 238
 239     /* #if 'Water' in GEOMETRY_J */
 240     /*     #for J in PARTICLES_ELEC_J */
 241     jq{J}              = _mm_set1_ps(charge[inr+{J}]);
 242     /*     #endfor */
 243     /*     #for J in PARTICLES_VDW_J */
 244     vdwjidx{J}A        = 2*vdwtype[inr+{J}];
 245     /*     #endfor */
 246     /*     #for I,J in PAIRS_IJ */
 247     /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 248     qq{I}{J}             = _mm_mul_ps(iq{I},jq{J});
 249     /*         #endif */
 250     /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 251     c6_{I}{J}            = _mm_set1_ps(vdwparam[vdwioffset{I}+vdwjidx{J}A]);
 252     c12_{I}{J}           = _mm_set1_ps(vdwparam[vdwioffset{I}+vdwjidx{J}A+1]);
 253     /*         #endif */
 254     /*     #endfor */
 255     /* #endif */
 256
 257     /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
 258     /*     #if KERNEL_ELEC!='None' */
 259     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 260     rcutoff_scalar   = fr->rcoulomb;
 261     /*     #else */
 262     rcutoff_scalar   = fr->rvdw;
 263     /*     #endif */
 264     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 265     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 266     /* #endif */
 267
 268     /* #if KERNEL_MOD_VDW=='PotentialShift' */
 269     sh_vdw_invrcut6  = _mm_set1_ps(fr->ic->sh_invrc6);
 270     rvdw             = _mm_set1_ps(fr->rvdw);
 271     /* #endif */
 272
 273     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 274     /*     #if KERNEL_MOD_ELEC=='PotentialSwitch'  */
 275     rswitch_scalar   = fr->rcoulomb_switch;
 276     rswitch          = _mm_set1_ps(rswitch_scalar);
 277     /*     #else */
 278     rswitch_scalar   = fr->rvdw_switch;
 279     rswitch          = _mm_set1_ps(rswitch_scalar);
 280     /*     #endif */
 281     /* Setup switch parameters */
 282     d_scalar         = rcutoff_scalar-rswitch_scalar;
 283     d                = _mm_set1_ps(d_scalar);
 284     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 285     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 286     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 287     /*     #if 'Force' in KERNEL_VF */
 288     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 289     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 290     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 291     /*     #endif */
 292     /* #endif */
 293
 294     /* Avoid stupid compiler warnings */
 295     jnrA = jnrB = jnrC = jnrD = 0;
 296     j_coord_offsetA = 0;
 297     j_coord_offsetB = 0;
 298     j_coord_offsetC = 0;
 299     j_coord_offsetD = 0;
 300
 301     /* ## Keep track of the floating point operations we issue for reporting! */
 302     /* #define OUTERFLOPS 0 */
 303     outeriter        = 0;
 304     inneriter        = 0;
 305
 306     for(iidx=0;iidx<4*DIM;iidx++)
 307     {
 308         scratch[iidx] = 0.0;
 309     }
 310
 311     /* Start outer loop over neighborlists */
 312     for(iidx=0; iidx<nri; iidx++)
 313     {
 314         /* Load shift vector for this list */
 315         i_shift_offset   = DIM*shiftidx[iidx];
 316
 317         /* Load limits for loop over neighbors */
 318         j_index_start    = jindex[iidx];
 319         j_index_end      = jindex[iidx+1];
 320
 321         /* Get outer coordinate index */
 322         inr              = iinr[iidx];
 323         i_coord_offset   = DIM*inr;
 324
 325         /* Load i particle coords and add shift vector */
 326         /* #if GEOMETRY_I == 'Particle' */
 327         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 328         /* #elif GEOMETRY_I == 'Water3' */
 329         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 330                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 331         /* #elif GEOMETRY_I == 'Water4' */
 332         /*     #if 0 in PARTICLES_I                 */
 333         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 334                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 335         /*     #else                                */
 336         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 337                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 338         /*     #endif                               */
 339         /* #endif                                   */
 340
 341         /* #if 'Force' in KERNEL_VF */
 342         /*     #for I in PARTICLES_I */
 343         fix{I}             = _mm_setzero_ps();
 344         fiy{I}             = _mm_setzero_ps();
 345         fiz{I}             = _mm_setzero_ps();
 346         /*     #endfor */
 347         /* #endif */
 348
 349         /* ## For water we already preloaded parameters at the start of the kernel */
 350         /* #if not 'Water' in GEOMETRY_I */
 351         /* Load parameters for i particles */
 352         /*     #for I in PARTICLES_ELEC_I */
 353         iq{I}              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+{I}));
 354         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 355         /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 356         isai{I}            = _mm_load1_ps(invsqrta+inr+{I});
 357         /*         #endif */
 358         /*     #endfor */
 359         /*     #for I in PARTICLES_VDW_I */
 360         vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 361         /*     #endfor */
 362         /* #endif */
 363
 364         /* #if 'Potential' in KERNEL_VF */
 365         /* Reset potential sums */
 366         /*     #if KERNEL_ELEC != 'None' */
 367         velecsum         = _mm_setzero_ps();
 368         /*     #endif */
 369         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
 370         vgbsum           = _mm_setzero_ps();
 371         /*     #endif */
 372         /*     #if KERNEL_VDW != 'None' */
 373         vvdwsum          = _mm_setzero_ps();
 374         /*     #endif */
 375         /* #endif */
 376         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
 377         dvdasum          = _mm_setzero_ps();
 378         /*     #endif */
 379
 380         /* #for ROUND in ['Loop','Epilogue'] */
 381
 382         /* #if ROUND =='Loop' */
 383         /* Start inner kernel loop */
 384         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 385         {
 386         /* ## First round is normal loop (next statement resets indentation) */
 387         /*     #if 0 */
 388         }
 389         /*     #endif */
 390         /* #else */
 391         if(jidx<j_index_end)
 392         {
 393         /* ## Second round is epilogue */
 394         /* #endif */
 395         /* #define INNERFLOPS 0 */
 396
 397             /* Get j neighbor index, and coordinate index */
 398             /* #if ROUND =='Loop' */
 399             jnrA             = jjnr[jidx];
 400             jnrB             = jjnr[jidx+1];
 401             jnrC             = jjnr[jidx+2];
 402             jnrD             = jjnr[jidx+3];
 403             /* #else */
 404             jnrlistA         = jjnr[jidx];
 405             jnrlistB         = jjnr[jidx+1];
 406             jnrlistC         = jjnr[jidx+2];
 407             jnrlistD         = jjnr[jidx+3];
 408             /* Sign of each element will be negative for non-real atoms.
 409              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 410              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 411              */
 412             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 413             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 414             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 415             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 416             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 417             /* #endif */
 418             j_coord_offsetA  = DIM*jnrA;
 419             j_coord_offsetB  = DIM*jnrB;
 420             j_coord_offsetC  = DIM*jnrC;
 421             j_coord_offsetD  = DIM*jnrD;
 422
 423             /* load j atom coordinates */
 424             /* #if GEOMETRY_J == 'Particle'             */
 425             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 426                                               x+j_coord_offsetC,x+j_coord_offsetD,
 427                                               &jx0,&jy0,&jz0);
 428             /* #elif GEOMETRY_J == 'Water3'             */
 429             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 430                                               x+j_coord_offsetC,x+j_coord_offsetD,
 431                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 432             /* #elif GEOMETRY_J == 'Water4'             */
 433             /*     #if 0 in PARTICLES_J                 */
 434             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 435                                               x+j_coord_offsetC,x+j_coord_offsetD,
 436                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 437                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 438             /*     #else                                */
 439             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 440                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 441                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 442             /*     #endif                               */
 443             /* #endif                                   */
 444
 445             /* Calculate displacement vector */
 446             /* #for I,J in PAIRS_IJ */
 447             dx{I}{J}             = _mm_sub_ps(ix{I},jx{J});
 448             dy{I}{J}             = _mm_sub_ps(iy{I},jy{J});
 449             dz{I}{J}             = _mm_sub_ps(iz{I},jz{J});
 450             /*     #define INNERFLOPS INNERFLOPS+3 */
 451             /* #endfor */
 452
 453             /* Calculate squared distance and things based on it */
 454             /* #for I,J in PAIRS_IJ */
 455             rsq{I}{J}            = gmx_mm_calc_rsq_ps(dx{I}{J},dy{I}{J},dz{I}{J});
 456             /*     #define INNERFLOPS INNERFLOPS+5 */
 457             /* #endfor */
 458
 459             /* #for I,J in PAIRS_IJ */
 460             /*     #if 'rinv' in INTERACTION_FLAGS[I][J] */
 461             rinv{I}{J}           = gmx_mm_invsqrt_ps(rsq{I}{J});
 462             /*         #define INNERFLOPS INNERFLOPS+5 */
 463             /*     #endif */
 464             /* #endfor */
 465
 466             /* #for I,J in PAIRS_IJ */
 467             /*     #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
 468             /*         # if 'rinv' not in INTERACTION_FLAGS[I][J] */
 469             rinvsq{I}{J}         = gmx_mm_inv_ps(rsq{I}{J});
 470             /*             #define INNERFLOPS INNERFLOPS+4 */
 471             /*         #else */
 472             rinvsq{I}{J}         = _mm_mul_ps(rinv{I}{J},rinv{I}{J});
 473             /*             #define INNERFLOPS INNERFLOPS+1 */
 474             /*         #endif */
 475             /*     #endif */
 476             /* #endfor */
 477
 478             /* #if not 'Water' in GEOMETRY_J */
 479             /* Load parameters for j particles */
 480             /*     #for J in PARTICLES_ELEC_J */
 481             jq{J}              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+{J},charge+jnrB+{J},
 482                                                               charge+jnrC+{J},charge+jnrD+{J});
 483             /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 484             isaj{J}            = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+{J},invsqrta+jnrB+{J},
 485                                                               invsqrta+jnrC+{J},invsqrta+jnrD+{J});
 486             /*         #endif */
 487             /*     #endfor */
 488             /*     #for J in PARTICLES_VDW_J */
 489             vdwjidx{J}A        = 2*vdwtype[jnrA+{J}];
 490             vdwjidx{J}B        = 2*vdwtype[jnrB+{J}];
 491             vdwjidx{J}C        = 2*vdwtype[jnrC+{J}];
 492             vdwjidx{J}D        = 2*vdwtype[jnrD+{J}];
 493             /*     #endfor */
 494             /* #endif */
 495
 496             /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_I */
 497             /*     #for J in PARTICLES_J */
 498             fjx{J}             = _mm_setzero_ps();
 499             fjy{J}             = _mm_setzero_ps();
 500             fjz{J}             = _mm_setzero_ps();
 501             /*     #endfor */
 502             /* #endif */
 503
 504             /* #for I,J in PAIRS_IJ */
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 511             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 512             /*         ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
 513             if (gmx_mm_any_lt(rsq{I}{J},rcutoff2))
 514             {
 515                 /*     #if 0    ## this and the next two lines is a hack to maintain auto-indentation in template file */
 516             }
 517             /*         #endif */
 518             /*         #define INNERFLOPS INNERFLOPS+1 */
 519             /*     #endif */
 520
 521             /*     #if 'r' in INTERACTION_FLAGS[I][J] */
 522             r{I}{J}              = _mm_mul_ps(rsq{I}{J},rinv{I}{J});
 523             /*         #if ROUND == 'Epilogue' */
 524             r{I}{J}              = _mm_andnot_ps(dummy_mask,r{I}{J});
 525             /*             #define INNERFLOPS INNERFLOPS+1 */
 526             /*         #endif */
 527             /*         #define INNERFLOPS INNERFLOPS+1 */
 528             /*     #endif */
 529
 530             /*     ## For water geometries we already loaded parameters at the start of the kernel */
 531             /*     #if not 'Water' in GEOMETRY_J */
 532             /* Compute parameters for interactions between i and j atoms */
 533             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 534             qq{I}{J}             = _mm_mul_ps(iq{I},jq{J});
 535             /*             #define INNERFLOPS INNERFLOPS+1 */
 536             /*         #endif */
 537             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 538             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset{I}+vdwjidx{J}A,
 539                                          vdwparam+vdwioffset{I}+vdwjidx{J}B,
 540                                          vdwparam+vdwioffset{I}+vdwjidx{J}C,
 541                                          vdwparam+vdwioffset{I}+vdwjidx{J}D,
 542                                          &c6_{I}{J},&c12_{I}{J});
 543             /*         #endif */
 544             /*     #endif */
 545
 546             /*     #if 'table' in INTERACTION_FLAGS[I][J] */
 547             /* Calculate table index by multiplying r with table scale and truncate to integer */
 548             rt               = _mm_mul_ps(r{I}{J},vftabscale);
 549             vfitab           = _mm_cvttps_epi32(rt);
 550             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 551             /*         #define INNERFLOPS INNERFLOPS+4                          */
 552             /*         #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW     */
 553             /*             ## 3 tables, 4 bytes per point: multiply index by 12 */
 554             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 555             /*         #elif 'Table' in KERNEL_ELEC                             */
 556             /*             ## 1 table, 4 bytes per point: multiply index by 4   */
 557             vfitab           = _mm_slli_epi32(vfitab,2);
 558             /*         #elif 'Table' in KERNEL_VDW                              */
 559             /*             ## 2 tables, 4 bytes per point: multiply index by 8  */
 560             vfitab           = _mm_slli_epi32(vfitab,3);
 561             /*         #endif                                                   */
 562             /*     #endif */
 563
 564             /*     ## ELECTROSTATIC INTERACTIONS */
 565             /*     #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 566
 567             /*         #if KERNEL_ELEC=='Coulomb' */
 568
 569             /* COULOMB ELECTROSTATICS */
 570             velec            = _mm_mul_ps(qq{I}{J},rinv{I}{J});
 571             /*             #define INNERFLOPS INNERFLOPS+1 */
 572             /*             #if 'Force' in KERNEL_VF */
 573             felec            = _mm_mul_ps(velec,rinvsq{I}{J});
 574             /*                 #define INNERFLOPS INNERFLOPS+2 */
 575             /*             #endif */
 576
 577             /*         #elif KERNEL_ELEC=='ReactionField' */
 578
 579             /* REACTION-FIELD ELECTROSTATICS */
 580             /*             #if 'Potential' in KERNEL_VF */
 581             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_add_ps(rinv{I}{J},_mm_mul_ps(krf,rsq{I}{J})),crf));
 582             /*                 #define INNERFLOPS INNERFLOPS+4 */
 583             /*             #endif */
 584             /*             #if 'Force' in KERNEL_VF */
 585             felec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_mul_ps(rinv{I}{J},rinvsq{I}{J}),krf2));
 586             /*                 #define INNERFLOPS INNERFLOPS+3 */
 587             /*             #endif */
 588
 589             /*         #elif KERNEL_ELEC=='GeneralizedBorn' */
 590
 591             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 592             isaprod          = _mm_mul_ps(isai{I},isaj{J});
 593             gbqqfactor       = _mm_xor_ps(signbit,_mm_mul_ps(qq{I}{J},_mm_mul_ps(isaprod,gbinvepsdiff)));
 594             gbscale          = _mm_mul_ps(isaprod,gbtabscale);
 595             /*             #define INNERFLOPS INNERFLOPS+5 */
 596
 597             /* Calculate generalized born table index - this is a separate table from the normal one,
 598              * but we use the same procedure by multiplying r with scale and truncating to integer.
 599              */
 600             rt               = _mm_mul_ps(r{I}{J},gbscale);
 601             gbitab           = _mm_cvttps_epi32(rt);
 602             gbeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 603             gbitab           = _mm_slli_epi32(gbitab,2);
 604             Y                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
 605             F                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
 606             G                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
 607             H                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
 608             _MM_TRANSPOSE4_PS(Y,F,G,H);
 609             Heps             = _mm_mul_ps(gbeps,H);
 610             Fp               = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
 611             VV               = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
 612             vgb              = _mm_mul_ps(gbqqfactor,VV);
 613             /*             #define INNERFLOPS INNERFLOPS+10 */
 614
 615             /*             #if 'Force' in KERNEL_VF */
 616             FF               = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 617             fgb              = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
 618             dvdatmp          = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r{I}{J})));
 619             dvdasum          = _mm_add_ps(dvdasum,dvdatmp);
 620             /*                 #if ROUND == 'Loop' */
 621             fjptrA           = dvda+jnrA;
 622             fjptrB           = dvda+jnrB;
 623             fjptrC           = dvda+jnrC;
 624             fjptrD           = dvda+jnrD;
 625             /*                 #else */
 626             /* 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. */
 627             fjptrA             = (jnrlistA>=0) ? dvda+jnrA : scratch;
 628             fjptrB             = (jnrlistB>=0) ? dvda+jnrB : scratch;
 629             fjptrC             = (jnrlistC>=0) ? dvda+jnrC : scratch;
 630             fjptrD             = (jnrlistD>=0) ? dvda+jnrD : scratch;
 631             /*                 #endif */
 632             gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj{J},isaj{J})));
 633             /*                 #define INNERFLOPS INNERFLOPS+13 */
 634             /*             #endif */
 635             velec            = _mm_mul_ps(qq{I}{J},rinv{I}{J});
 636             /*                 #define INNERFLOPS INNERFLOPS+1 */
 637             /*             #if 'Force' in KERNEL_VF */
 638             felec            = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv{I}{J}),fgb),rinv{I}{J});
 639             /*                 #define INNERFLOPS INNERFLOPS+3 */
 640             /*             #endif */
 641
 642             /*         #elif KERNEL_ELEC=='Ewald' */
 643             /* EWALD ELECTROSTATICS */
 644
 645             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 646             ewrt             = _mm_mul_ps(r{I}{J},ewtabscale);
 647             ewitab           = _mm_cvttps_epi32(ewrt);
 648             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 649             /*             #define INNERFLOPS INNERFLOPS+4 */
 650             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
 651             ewitab           = _mm_slli_epi32(ewitab,2);
 652             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 653             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 654             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 655             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 656             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 657             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 658             /*                 #define INNERFLOPS INNERFLOPS+2 */
 659             /*                 #if KERNEL_MOD_ELEC=='PotentialShift' */
 660             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 661             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_sub_ps(rinv{I}{J},sh_ewald),velec));
 662             /*                     #define INNERFLOPS INNERFLOPS+7 */
 663             /*                 #else */
 664             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 665             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(rinv{I}{J},velec));
 666             /*                     #define INNERFLOPS INNERFLOPS+6 */
 667             /*                 #endif */
 668             /*                 #if 'Force' in KERNEL_VF */
 669             felec            = _mm_mul_ps(_mm_mul_ps(qq{I}{J},rinv{I}{J}),_mm_sub_ps(rinvsq{I}{J},felec));
 670             /*                      #define INNERFLOPS INNERFLOPS+3 */
 671             /*                 #endif */
 672             /*             #elif KERNEL_VF=='Force' */
 673             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
 674                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
 675                                          &ewtabF,&ewtabFn);
 676             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
 677             felec            = _mm_mul_ps(_mm_mul_ps(qq{I}{J},rinv{I}{J}),_mm_sub_ps(rinvsq{I}{J},felec));
 678             /*                 #define INNERFLOPS INNERFLOPS+7 */
 679             /*             #endif */
 680
 681             /*         #elif KERNEL_ELEC=='CubicSplineTable' */
 682
 683             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 684             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 685             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 686             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 687             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 688             _MM_TRANSPOSE4_PS(Y,F,G,H);
 689             Heps             = _mm_mul_ps(vfeps,H);
 690             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 691             /*             #define INNERFLOPS INNERFLOPS+4 */
 692             /*             #if 'Potential' in KERNEL_VF */
 693             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 694             velec            = _mm_mul_ps(qq{I}{J},VV);
 695             /*                 #define INNERFLOPS INNERFLOPS+3 */
 696             /*             #endif */
 697             /*             #if 'Force' in KERNEL_VF */
 698             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 699             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq{I}{J},FF),_mm_mul_ps(vftabscale,rinv{I}{J})));
 700             /*                 #define INNERFLOPS INNERFLOPS+7 */
 701             /*             #endif */
 702             /*         #endif */
 703             /*         ## End of check for electrostatics interaction forms */
 704             /*     #endif */
 705             /*     ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
 706
 707             /*     #if 'vdw' in INTERACTION_FLAGS[I][J] */
 708
 709             /*         #if KERNEL_VDW=='LennardJones' */
 710
 711             /* LENNARD-JONES DISPERSION/REPULSION */
 712
 713             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 714             /*             #define INNERFLOPS INNERFLOPS+2 */
 715             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 716             vvdw6            = _mm_mul_ps(c6_{I}{J},rinvsix);
 717             vvdw12           = _mm_mul_ps(c12_{I}{J},_mm_mul_ps(rinvsix,rinvsix));
 718             /*                 #define INNERFLOPS INNERFLOPS+3 */
 719             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 720             vvdw             = _mm_sub_ps(_mm_mul_ps( _mm_sub_ps(vvdw12 , _mm_mul_ps(c12_{I}{J},_mm_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 721                                           _mm_mul_ps( _mm_sub_ps(vvdw6,_mm_mul_ps(c6_{I}{J},sh_vdw_invrcut6)),one_sixth));
 722             /*                     #define INNERFLOPS INNERFLOPS+8 */
 723             /*                 #else */
 724             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 725             /*                     #define INNERFLOPS INNERFLOPS+3 */
 726             /*                 #endif */
 727             /*                 ## Check for force inside potential check, i.e. this means we already did the potential part */
 728             /*                 #if 'Force' in KERNEL_VF */
 729             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq{I}{J});
 730             /*                     #define INNERFLOPS INNERFLOPS+2 */
 731             /*                 #endif */
 732             /*             #elif KERNEL_VF=='Force' */
 733             /*                 ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
 734             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_{I}{J},rinvsix),c6_{I}{J}),_mm_mul_ps(rinvsix,rinvsq{I}{J}));
 735             /*                 #define INNERFLOPS INNERFLOPS+4 */
 736             /*             #endif */
 737
 738             /*         #elif KERNEL_VDW=='CubicSplineTable' */
 739
 740             /* CUBIC SPLINE TABLE DISPERSION */
 741             /*             #if 'Table' in KERNEL_ELEC */
 742             vfitab           = _mm_add_epi32(vfitab,ifour);
 743             /*             #endif                     */
 744             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 745             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 746             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 747             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 748             _MM_TRANSPOSE4_PS(Y,F,G,H);
 749             Heps             = _mm_mul_ps(vfeps,H);
 750             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 751             /*             #define INNERFLOPS INNERFLOPS+4 */
 752             /*             #if 'Potential' in KERNEL_VF */
 753             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 754             vvdw6            = _mm_mul_ps(c6_{I}{J},VV);
 755             /*                 #define INNERFLOPS INNERFLOPS+3 */
 756             /*             #endif */
 757             /*             #if 'Force' in KERNEL_VF */
 758             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 759             fvdw6            = _mm_mul_ps(c6_{I}{J},FF);
 760             /*                 #define INNERFLOPS INNERFLOPS+4 */
 761             /*             #endif */
 762
 763             /* CUBIC SPLINE TABLE REPULSION */
 764             vfitab           = _mm_add_epi32(vfitab,ifour);
 765             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 766             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 767             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 768             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 769             _MM_TRANSPOSE4_PS(Y,F,G,H);
 770             Heps             = _mm_mul_ps(vfeps,H);
 771             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 772             /*             #define INNERFLOPS INNERFLOPS+4 */
 773             /*             #if 'Potential' in KERNEL_VF */
 774             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 775             vvdw12           = _mm_mul_ps(c12_{I}{J},VV);
 776             /*                 #define INNERFLOPS INNERFLOPS+3 */
 777             /*             #endif */
 778             /*             #if 'Force' in KERNEL_VF */
 779             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 780             fvdw12           = _mm_mul_ps(c12_{I}{J},FF);
 781             /*                 #define INNERFLOPS INNERFLOPS+5 */
 782             /*             #endif */
 783             /*             #if 'Potential' in KERNEL_VF */
 784             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 785             /*                 #define INNERFLOPS INNERFLOPS+1 */
 786             /*             #endif */
 787             /*             #if 'Force' in KERNEL_VF */
 788             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv{I}{J})));
 789             /*                 #define INNERFLOPS INNERFLOPS+4 */
 790             /*             #endif */
 791             /*         #endif */
 792             /*         ## End of check for vdw interaction forms */
 793             /*     #endif */
 794             /*     ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
 795
 796             /*     #if 'switch' in INTERACTION_FLAGS[I][J] */
 797             d                = _mm_sub_ps(r{I}{J},rswitch);
 798             d                = _mm_max_ps(d,_mm_setzero_ps());
 799             d2               = _mm_mul_ps(d,d);
 800             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_add_ps(swV3,_mm_mul_ps(d,_mm_add_ps(swV4,_mm_mul_ps(d,swV5)))))));
 801             /*         #define INNERFLOPS INNERFLOPS+10 */
 802
 803             /*         #if 'Force' in KERNEL_VF */
 804             dsw              = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
 805             /*             #define INNERFLOPS INNERFLOPS+5 */
 806             /*         #endif */
 807
 808             /* Evaluate switch function */
 809             /*         #if 'Force' in KERNEL_VF */
 810             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 811             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 812             felec            = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv{I}{J},_mm_mul_ps(velec,dsw)) );
 813             /*                 #define INNERFLOPS INNERFLOPS+4 */
 814             /*             #endif */
 815             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 816             fvdw             = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv{I}{J},_mm_mul_ps(vvdw,dsw)) );
 817             /*                 #define INNERFLOPS INNERFLOPS+4 */
 818             /*             #endif */
 819             /*         #endif */
 820             /*         #if 'Potential' in KERNEL_VF */
 821             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 822             velec            = _mm_mul_ps(velec,sw);
 823             /*                 #define INNERFLOPS INNERFLOPS+1 */
 824             /*             #endif */
 825             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 826             vvdw             = _mm_mul_ps(vvdw,sw);
 827             /*                 #define INNERFLOPS INNERFLOPS+1 */
 828             /*             #endif */
 829             /*         #endif */
 830             /*     #endif */
 831             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 832             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 833             cutoff_mask      = _mm_cmplt_ps(rsq{I}{J},rcutoff2);
 834             /*         #define INNERFLOPS INNERFLOPS+1 */
 835             /*     #endif */
 836
 837             /*     #if 'Potential' in KERNEL_VF */
 838             /* Update potential sum for this i atom from the interaction with this j atom. */
 839             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 840             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 841             velec            = _mm_and_ps(velec,cutoff_mask);
 842             /*                 #define INNERFLOPS INNERFLOPS+1 */
 843             /*             #endif                                       */
 844             /*             #if ROUND == 'Epilogue' */
 845             velec            = _mm_andnot_ps(dummy_mask,velec);
 846             /*             #endif */
 847             velecsum         = _mm_add_ps(velecsum,velec);
 848             /*             #define INNERFLOPS INNERFLOPS+1 */
 849             /*             #if KERNEL_ELEC=='GeneralizedBorn' */
 850             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 851             vgb              = _mm_and_ps(vgb,cutoff_mask);
 852             /*                 #define INNERFLOPS INNERFLOPS+1 */
 853             /*             #endif                                       */
 854             /*             #if ROUND == 'Epilogue' */
 855             vgb              = _mm_andnot_ps(dummy_mask,vgb);
 856             /*             #endif */
 857             vgbsum           = _mm_add_ps(vgbsum,vgb);
 858             /*                 #define INNERFLOPS INNERFLOPS+1 */
 859             /*             #endif */
 860             /*         #endif */
 861             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 862             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 863             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 864             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 865             /*                 #define INNERFLOPS INNERFLOPS+1 */
 866             /*             #endif                                       */
 867             /*             #if ROUND == 'Epilogue' */
 868             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 869             /*             #endif */
 870             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 871             /*             #define INNERFLOPS INNERFLOPS+1 */
 872             /*         #endif */
 873             /*     #endif */
 874
 875             /*     #if 'Force' in KERNEL_VF */
 876
 877             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
 878             fscal            = _mm_add_ps(felec,fvdw);
 879             /*             #define INNERFLOPS INNERFLOPS+1 */
 880             /*         #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
 881             fscal            = felec;
 882             /*         #elif 'vdw' in INTERACTION_FLAGS[I][J] */
 883             fscal            = fvdw;
 884             /*        #endif */
 885
 886             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 887             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 888             fscal            = _mm_and_ps(fscal,cutoff_mask);
 889             /*                 #define INNERFLOPS INNERFLOPS+1 */
 890             /*             #endif                                       */
 891
 892             /*             #if ROUND == 'Epilogue' */
 893             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 894             /*             #endif */
 895
 896             /* Calculate temporary vectorial force */
 897             tx               = _mm_mul_ps(fscal,dx{I}{J});
 898             ty               = _mm_mul_ps(fscal,dy{I}{J});
 899             tz               = _mm_mul_ps(fscal,dz{I}{J});
 900
 901             /* Update vectorial force */
 902             fix{I}             = _mm_add_ps(fix{I},tx);
 903             fiy{I}             = _mm_add_ps(fiy{I},ty);
 904             fiz{I}             = _mm_add_ps(fiz{I},tz);
 905             /*             #define INNERFLOPS INNERFLOPS+6 */
 906
 907             /* #if GEOMETRY_I == 'Particle'             */
 908             /*     #if ROUND == 'Loop' */
 909             fjptrA             = f+j_coord_offsetA;
 910             fjptrB             = f+j_coord_offsetB;
 911             fjptrC             = f+j_coord_offsetC;
 912             fjptrD             = f+j_coord_offsetD;
 913             /*     #else */
 914             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 915             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 916             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 917             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 918             /*     #endif */
 919             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 920             /*     #define INNERFLOPS INNERFLOPS+3      */
 921             /* #else                                    */
 922             fjx{J}             = _mm_add_ps(fjx{J},tx);
 923             fjy{J}             = _mm_add_ps(fjy{J},ty);
 924             fjz{J}             = _mm_add_ps(fjz{J},tz);
 925             /*     #define INNERFLOPS INNERFLOPS+3      */
 926             /* #endif                                   */
 927
 928             /*     #endif */
 929
 930             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 931             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 932             /*         #if 0    ## This and next two lines is a hack to maintain indentation in template file */
 933             {
 934                 /*     #endif */
 935             }
 936             /*     #endif */
 937             /*    ## End of check for the interaction being outside the cutoff */
 938
 939             /* #endfor */
 940             /* ## End of loop over i-j interaction pairs */
 941
 942             /* #if GEOMETRY_I != 'Particle' */
 943             /*     #if ROUND == 'Loop' */
 944             fjptrA             = f+j_coord_offsetA;
 945             fjptrB             = f+j_coord_offsetB;
 946             fjptrC             = f+j_coord_offsetC;
 947             fjptrD             = f+j_coord_offsetD;
 948             /*     #else */
 949             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 950             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 951             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 952             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 953             /*     #endif */
 954             /* #endif */
 955
 956             /* #if 'Water' in GEOMETRY_I and GEOMETRY_J == 'Particle' */
 957             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 958             /* #elif GEOMETRY_J == 'Water3'               */
 959             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 960                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 961             /*     #define INNERFLOPS INNERFLOPS+9      */
 962             /* #elif GEOMETRY_J == 'Water4'             */
 963             /*     #if 0 in PARTICLES_J                 */
 964             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 965                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 966                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 967             /*     #define INNERFLOPS INNERFLOPS+12     */
 968             /*     #else                                */
 969             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 970                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 971             /*     #define INNERFLOPS INNERFLOPS+9      */
 972             /*     #endif                               */
 973             /* #endif                                   */
 974
 975             /* Inner loop uses {INNERFLOPS} flops */
 976         }
 977
 978         /* #endfor */
 979
 980         /* End of innermost loop */
 981
 982         /* #if 'Force' in KERNEL_VF */
 983         /*     #if GEOMETRY_I == 'Particle'            */
 984         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 985                                               f+i_coord_offset,fshift+i_shift_offset);
 986         /*         #define OUTERFLOPS OUTERFLOPS+6     */
 987         /*     #elif GEOMETRY_I == 'Water3'            */
 988         gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 989                                               f+i_coord_offset,fshift+i_shift_offset);
 990         /*         #define OUTERFLOPS OUTERFLOPS+18    */
 991         /*     #elif GEOMETRY_I == 'Water4'            */
 992         /*         #if 0 in PARTICLES_I                */
 993         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 994                                               f+i_coord_offset,fshift+i_shift_offset);
 995         /*             #define OUTERFLOPS OUTERFLOPS+24    */
 996         /*         #else                               */
 997         gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 998                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 999         /*             #define OUTERFLOPS OUTERFLOPS+18    */
1000         /*         #endif                              */
1001         /*     #endif                                  */
1002         /* #endif                                      */
1003
1004         /* #if 'Potential' in KERNEL_VF */
1005         ggid                        = gid[iidx];
1006         /* Update potential energies */
1007         /*     #if KERNEL_ELEC != 'None' */
1008         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1009         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1010         /*     #endif */
1011         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
1012         gmx_mm_update_1pot_ps(vgbsum,kernel_data->energygrp_polarization+ggid);
1013         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1014         /*     #endif */
1015         /*     #if KERNEL_VDW != 'None' */
1016         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1017         /*         #define OUTERFLOPS OUTERFLOPS+1 */
1018         /*     #endif */
1019         /* #endif */
1020         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
1021         dvdasum = _mm_mul_ps(dvdasum, _mm_mul_ps(isai{I},isai{I}));
1022         gmx_mm_update_1pot_ps(dvdasum,dvda+inr);
1023         /*     #endif */
1024
1025         /* Increment number of inner iterations */
1026         inneriter                  += j_index_end - j_index_start;
1027
1028         /* Outer loop uses {OUTERFLOPS} flops */
1029     }
1030
1031     /* Increment number of outer iterations */
1032     outeriter        += nri;
1033
1034     /* Update outer/inner flops */
1035     /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1036     /* ## primitive and replaces aggressively even in strings inside these directives, we need to      */
1037     /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source.      */
1038     /* #if GEOMETRY_I == 'Water3'            */
1039     /*     #define ISUFFIX '_W3'             */
1040     /* #elif GEOMETRY_I == 'Water4'          */
1041     /*     #define ISUFFIX '_W4'             */
1042     /* #else                                 */
1043     /*     #define ISUFFIX ''                */
1044     /* #endif                                */
1045     /* #if GEOMETRY_J == 'Water3'            */
1046     /*     #define JSUFFIX 'W3'              */
1047     /* #elif GEOMETRY_J == 'Water4'          */
1048     /*     #define JSUFFIX 'W4'              */
1049     /* #else                                 */
1050     /*     #define JSUFFIX ''                */
1051     /* #endif                                */
1052     /* #if 'PotentialAndForce' in KERNEL_VF  */
1053     /*     #define VFSUFFIX  '_VF'           */
1054     /* #elif 'Potential' in KERNEL_VF        */
1055     /*     #define VFSUFFIX '_V'             */
1056     /* #else                                 */
1057     /*     #define VFSUFFIX '_F'             */
1058     /* #endif                                */
1059
1060     /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1061     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1062     /* #elif KERNEL_ELEC != 'None' */
1063     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1064     /* #else */
1065     inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1066     /* #endif  */
1067 }