src/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_template_sse4_1_single.pre

   1 /* #if 0 */
   2 #error This file must be processed with the Gromacs pre-preprocessor
   3 /* #endif */
   4 /* #if INCLUDE_HEADER */
   5 #ifdef HAVE_CONFIG_H
   6 #include <config.h>
   7 #endif
   8
   9 #include <math.h>
  10
  11 #include "../nb_kernel.h"
  12 #include "types/simple.h"
  13 #include "vec.h"
  14 #include "nrnb.h"
  15
  16 #include "gmx_math_x86_sse4_1_single.h"
  17 #include "kernelutil_x86_sse4_1_single.h"
  18 /* #endif */
  19
  20 /* ## List of variables set by the generating script:                                    */
  21 /* ##                                                                                    */
  22 /* ## Setttings that apply to the entire kernel:                                         */
  23 /* ## KERNEL_ELEC:           String, choice for electrostatic interactions               */
  24 /* ## KERNEL_VDW:            String, choice for van der Waals interactions               */
  25 /* ## KERNEL_NAME:           String, name of this kernel                                 */
  26 /* ## KERNEL_VF:             String telling if we calculate potential, force, or both    */
  27 /* ## GEOMETRY_I/GEOMETRY_J: String, name of each geometry, e.g. 'Water3' or '1Particle' */
  28 /* ##                                                                                    */
  29 /* ## Setttings that apply to particles in the outer (I) or inner (J) loops:             */
  30 /* ## PARTICLES_I[]/         Arrays with lists of i/j particles to use in kernel. It is  */
  31 /* ## PARTICLES_J[]:         just [0] for particle geometry, but can be longer for water */
  32 /* ## PARTICLES_ELEC_I[]/    Arrays with lists of i/j particle that have electrostatics  */
  33 /* ## PARTICLES_ELEC_J[]:    interactions that should be calculated in this kernel.      */
  34 /* ## PARTICLES_VDW_I[]/     Arrays with the list of i/j particle that have VdW          */
  35 /* ## PARTICLES_VDW_J[]:     interactions that should be calculated in this kernel.      */
  36 /* ##                                                                                    */
  37 /* ## Setttings for pairs of interactions (e.g. 2nd i particle against 1st j particle)   */
  38 /* ## PAIRS_IJ[]:            Array with (i,j) tuples of pairs for which interactions     */
  39 /* ##                        should be calculated in this kernel. Zero-charge particles  */
  40 /* ##                        do not have interactions with particles without vdw, and    */
  41 /* ##                        Vdw-only interactions are not evaluated in a no-vdw-kernel. */
  42 /* ## INTERACTION_FLAGS[][]: 2D matrix, dimension e.g. 3*3 for water-water interactions. */
  43 /* ##                        For each i-j pair, the element [I][J] is a list of strings  */
  44 /* ##                        defining properties/flags of this interaction. Examples     */
  45 /* ##                        include 'electrostatics'/'vdw' if that type of interaction  */
  46 /* ##                        should be evaluated, 'rsq'/'rinv'/'rinvsq' if those values  */
  47 /* ##                        are needed, and 'exactcutoff' or 'shift','switch' to        */
  48 /* ##                        decide if the force/potential should be modified. This way  */
  49 /* ##                        we only calculate values absolutely needed for each case.   */
  50
  51 /* ## Calculate the size and offset for (merged/interleaved) table data */
  52
  53 /*
  54  * Gromacs nonbonded kernel:   {KERNEL_NAME}
  55  * Electrostatics interaction: {KERNEL_ELEC}
  56  * VdW interaction:            {KERNEL_VDW}
  57  * Geometry:                   {GEOMETRY_I}-{GEOMETRY_J}
  58  * Calculate force/pot:        {KERNEL_VF}
  59  */
  60 void
  61 {KERNEL_NAME}
  62                     (t_nblist * gmx_restrict                nlist,
  63                      rvec * gmx_restrict                    xx,
  64                      rvec * gmx_restrict                    ff,
  65                      t_forcerec * gmx_restrict              fr,
  66                      t_mdatoms * gmx_restrict               mdatoms,
  67                      nb_kernel_data_t * gmx_restrict        kernel_data,
  68                      t_nrnb * gmx_restrict                  nrnb)
  69 {
  70     /* ## Not all variables are used for all kernels, but any optimizing compiler fixes that, */
  71     /* ## so there is no point in going to extremes to exclude variables that are not needed. */
  72     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  73      * just 0 for non-waters.
  74      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  75      * jnr indices corresponding to data put in the four positions in the SIMD register.
  76      */
  77     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  78     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  79     int              jnrA,jnrB,jnrC,jnrD;
  80     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  81     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  86     real             scratch[4*DIM];
  87     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     /* #for I in PARTICLES_I */
  89     int              vdwioffset{I};
  90     __m128           ix{I},iy{I},iz{I},fix{I},fiy{I},fiz{I},iq{I},isai{I};
  91     /* #endfor */
  92     /* #for J in PARTICLES_J */
  93     int              vdwjidx{J}A,vdwjidx{J}B,vdwjidx{J}C,vdwjidx{J}D;
  94     __m128           jx{J},jy{J},jz{J},fjx{J},fjy{J},fjz{J},jq{J},isaj{J};
  95     /* #endfor */
  96     /* #for I,J in PAIRS_IJ */
  97     __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};
  98     /* #endfor */
  99     /* #if KERNEL_ELEC != 'None' */
 100     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 101     real             *charge;
 102     /* #endif */
 103     /* #if 'GeneralizedBorn' in KERNEL_ELEC */
 104     __m128i          gbitab;
 105     __m128           vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,gbeps,dvdatmp;
 106     __m128           minushalf = _mm_set1_ps(-0.5);
 107     real             *invsqrta,*dvda,*gbtab;
 108     /* #endif */
 109     /* #if KERNEL_VDW != 'None' */
 110     int              nvdwtype;
 111     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 115     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 116     /* #endif */
 117     /* #if 'Table' in KERNEL_ELEC or 'GeneralizedBorn' in KERNEL_ELEC or 'Table' in KERNEL_VDW */
 118     __m128i          vfitab;
 119     __m128i          ifour       = _mm_set1_epi32(4);
 120     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 121     real             *vftab;
 122     /* #endif */
 123     /* #if 'Ewald' in KERNEL_ELEC */
 124     __m128i          ewitab;
 125     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 126     real             *ewtab;
 127     /* #endif */
 128     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 129     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 130     real             rswitch_scalar,d_scalar;
 131     /* #endif */
 132     __m128           dummy_mask,cutoff_mask;
 133     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 134     __m128           one     = _mm_set1_ps(1.0);
 135     __m128           two     = _mm_set1_ps(2.0);
 136     x                = xx[0];
 137     f                = ff[0];
 138
 139     nri              = nlist->nri;
 140     iinr             = nlist->iinr;
 141     jindex           = nlist->jindex;
 142     jjnr             = nlist->jjnr;
 143     shiftidx         = nlist->shift;
 144     gid              = nlist->gid;
 145     shiftvec         = fr->shift_vec[0];
 146     fshift           = fr->fshift[0];
 147     /* #if KERNEL_ELEC != 'None' */
 148     facel            = _mm_set1_ps(fr->epsfac);
 149     charge           = mdatoms->chargeA;
 150     /*     #if 'ReactionField' in KERNEL_ELEC */
 151     krf              = _mm_set1_ps(fr->ic->k_rf);
 152     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
 153     crf              = _mm_set1_ps(fr->ic->c_rf);
 154     /*     #endif */
 155     /* #endif */
 156     /* #if KERNEL_VDW != 'None' */
 157     nvdwtype         = fr->ntype;
 158     vdwparam         = fr->nbfp;
 159     vdwtype          = mdatoms->typeA;
 160     /* #endif */
 161
 162     /* #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW */
 163     vftab            = kernel_data->table_elec_vdw->data;
 164     vftabscale       = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
 165     /* #elif 'Table' in KERNEL_ELEC */
 166     vftab            = kernel_data->table_elec->data;
 167     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
 168     /* #elif 'Table' in KERNEL_VDW */
 169     vftab            = kernel_data->table_vdw->data;
 170     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 171     /* #endif */
 172
 173     /* #if 'Ewald' in KERNEL_ELEC */
 174     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 175     /*     #if KERNEL_VF=='Force' and KERNEL_MOD_ELEC!='PotentialSwitch' */
 176     ewtab            = fr->ic->tabq_coul_F;
 177     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 178     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 179     /*     #else */
 180     ewtab            = fr->ic->tabq_coul_FDV0;
 181     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 182     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 183      /*     #endif */
 184     /* #endif */
 185
 186     /* #if KERNEL_ELEC=='GeneralizedBorn' */
 187     invsqrta         = fr->invsqrta;
 188     dvda             = fr->dvda;
 189     gbtabscale       = _mm_set1_ps(fr->gbtab.scale);
 190     gbtab            = fr->gbtab.data;
 191     gbinvepsdiff     = _mm_set1_ps((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 192     /* #endif */
 193
 194     /* #if 'Water' in GEOMETRY_I */
 195     /* Setup water-specific parameters */
 196     inr              = nlist->iinr[0];
 197     /*     #for I in PARTICLES_ELEC_I */
 198     iq{I}              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+{I}]));
 199     /*     #endfor */
 200     /*     #for I in PARTICLES_VDW_I */
 201     vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 202     /*     #endfor */
 203     /* #endif */
 204
 205     /* #if 'Water' in GEOMETRY_J */
 206     /*     #for J in PARTICLES_ELEC_J */
 207     jq{J}              = _mm_set1_ps(charge[inr+{J}]);
 208     /*     #endfor */
 209     /*     #for J in PARTICLES_VDW_J */
 210     vdwjidx{J}A        = 2*vdwtype[inr+{J}];
 211     /*     #endfor */
 212     /*     #for I,J in PAIRS_IJ */
 213     /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 214     qq{I}{J}             = _mm_mul_ps(iq{I},jq{J});
 215     /*         #endif */
 216     /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 217     c6_{I}{J}            = _mm_set1_ps(vdwparam[vdwioffset{I}+vdwjidx{J}A]);
 218     c12_{I}{J}           = _mm_set1_ps(vdwparam[vdwioffset{I}+vdwjidx{J}A+1]);
 219     /*         #endif */
 220     /*     #endfor */
 221     /* #endif */
 222
 223     /* #if KERNEL_MOD_ELEC!='None' or KERNEL_MOD_VDW!='None' */
 224     /*     #if KERNEL_ELEC!='None' */
 225     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 226     rcutoff_scalar   = fr->rcoulomb;
 227     /*     #else */
 228     rcutoff_scalar   = fr->rvdw;
 229     /*     #endif */
 230     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 231     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 232     /* #endif */
 233
 234     /* #if KERNEL_MOD_VDW=='PotentialShift' */
 235     sh_vdw_invrcut6  = _mm_set1_ps(fr->ic->sh_invrc6);
 236     rvdw             = _mm_set1_ps(fr->rvdw);
 237     /* #endif */
 238
 239     /* #if 'PotentialSwitch' in [KERNEL_MOD_ELEC,KERNEL_MOD_VDW] */
 240     /*     #if KERNEL_MOD_ELEC=='PotentialSwitch'  */
 241     rswitch_scalar   = fr->rcoulomb_switch;
 242     rswitch          = _mm_set1_ps(rswitch_scalar);
 243     /*     #else */
 244     rswitch_scalar   = fr->rvdw_switch;
 245     rswitch          = _mm_set1_ps(rswitch_scalar);
 246     /*     #endif */
 247     /* Setup switch parameters */
 248     d_scalar         = rcutoff_scalar-rswitch_scalar;
 249     d                = _mm_set1_ps(d_scalar);
 250     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 251     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 252     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 253     /*     #if 'Force' in KERNEL_VF */
 254     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 255     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 256     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 257     /*     #endif */
 258     /* #endif */
 259
 260     /* Avoid stupid compiler warnings */
 261     jnrA = jnrB = jnrC = jnrD = 0;
 262     j_coord_offsetA = 0;
 263     j_coord_offsetB = 0;
 264     j_coord_offsetC = 0;
 265     j_coord_offsetD = 0;
 266
 267     /* ## Keep track of the floating point operations we issue for reporting! */
 268     /* #define OUTERFLOPS 0 */
 269     outeriter        = 0;
 270     inneriter        = 0;
 271
 272     for(iidx=0;iidx<4*DIM;iidx++)
 273     {
 274         scratch[iidx] = 0.0;
 275     }
 276
 277     /* Start outer loop over neighborlists */
 278     for(iidx=0; iidx<nri; iidx++)
 279     {
 280         /* Load shift vector for this list */
 281         i_shift_offset   = DIM*shiftidx[iidx];
 282
 283         /* Load limits for loop over neighbors */
 284         j_index_start    = jindex[iidx];
 285         j_index_end      = jindex[iidx+1];
 286
 287         /* Get outer coordinate index */
 288         inr              = iinr[iidx];
 289         i_coord_offset   = DIM*inr;
 290
 291         /* Load i particle coords and add shift vector */
 292         /* #if GEOMETRY_I == 'Particle' */
 293         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 294         /* #elif GEOMETRY_I == 'Water3' */
 295         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 296                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 297         /* #elif GEOMETRY_I == 'Water4' */
 298         /*     #if 0 in PARTICLES_I                 */
 299         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 300                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 301         /*     #else                                */
 302         gmx_mm_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 303                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 304         /*     #endif                               */
 305         /* #endif                                   */
 306
 307         /* #if 'Force' in KERNEL_VF */
 308         /*     #for I in PARTICLES_I */
 309         fix{I}             = _mm_setzero_ps();
 310         fiy{I}             = _mm_setzero_ps();
 311         fiz{I}             = _mm_setzero_ps();
 312         /*     #endfor */
 313         /* #endif */
 314
 315         /* ## For water we already preloaded parameters at the start of the kernel */
 316         /* #if not 'Water' in GEOMETRY_I */
 317         /* Load parameters for i particles */
 318         /*     #for I in PARTICLES_ELEC_I */
 319         iq{I}              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+{I}));
 320         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 321         /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 322         isai{I}            = _mm_load1_ps(invsqrta+inr+{I});
 323         /*         #endif */
 324         /*     #endfor */
 325         /*     #for I in PARTICLES_VDW_I */
 326         vdwioffset{I}      = 2*nvdwtype*vdwtype[inr+{I}];
 327         /*     #endfor */
 328         /* #endif */
 329
 330         /* #if 'Potential' in KERNEL_VF */
 331         /* Reset potential sums */
 332         /*     #if KERNEL_ELEC != 'None' */
 333         velecsum         = _mm_setzero_ps();
 334         /*     #endif */
 335         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
 336         vgbsum           = _mm_setzero_ps();
 337         /*     #endif */
 338         /*     #if KERNEL_VDW != 'None' */
 339         vvdwsum          = _mm_setzero_ps();
 340         /*     #endif */
 341         /* #endif */
 342         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
 343         dvdasum          = _mm_setzero_ps();
 344         /*     #endif */
 345
 346         /* #for ROUND in ['Loop','Epilogue'] */
 347
 348         /* #if ROUND =='Loop' */
 349         /* Start inner kernel loop */
 350         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 351         {
 352         /* ## First round is normal loop (next statement resets indentation) */
 353         /*     #if 0 */
 354         }
 355         /*     #endif */
 356         /* #else */
 357         if(jidx<j_index_end)
 358         {
 359         /* ## Second round is epilogue */
 360         /* #endif */
 361         /* #define INNERFLOPS 0 */
 362
 363             /* Get j neighbor index, and coordinate index */
 364             /* #if ROUND =='Loop' */
 365             jnrA             = jjnr[jidx];
 366             jnrB             = jjnr[jidx+1];
 367             jnrC             = jjnr[jidx+2];
 368             jnrD             = jjnr[jidx+3];
 369             /* #else */
 370             jnrlistA         = jjnr[jidx];
 371             jnrlistB         = jjnr[jidx+1];
 372             jnrlistC         = jjnr[jidx+2];
 373             jnrlistD         = jjnr[jidx+3];
 374             /* Sign of each element will be negative for non-real atoms.
 375              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 376              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 377              */
 378             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 379             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 380             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 381             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 382             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 383             /* #endif */
 384             j_coord_offsetA  = DIM*jnrA;
 385             j_coord_offsetB  = DIM*jnrB;
 386             j_coord_offsetC  = DIM*jnrC;
 387             j_coord_offsetD  = DIM*jnrD;
 388
 389             /* load j atom coordinates */
 390             /* #if GEOMETRY_J == 'Particle'             */
 391             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 392                                               x+j_coord_offsetC,x+j_coord_offsetD,
 393                                               &jx0,&jy0,&jz0);
 394             /* #elif GEOMETRY_J == 'Water3'             */
 395             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 396                                               x+j_coord_offsetC,x+j_coord_offsetD,
 397                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 398             /* #elif GEOMETRY_J == 'Water4'             */
 399             /*     #if 0 in PARTICLES_J                 */
 400             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 401                                               x+j_coord_offsetC,x+j_coord_offsetD,
 402                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 403                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 404             /*     #else                                */
 405             gmx_mm_load_3rvec_4ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 406                                               x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 407                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 408             /*     #endif                               */
 409             /* #endif                                   */
 410
 411             /* Calculate displacement vector */
 412             /* #for I,J in PAIRS_IJ */
 413             dx{I}{J}             = _mm_sub_ps(ix{I},jx{J});
 414             dy{I}{J}             = _mm_sub_ps(iy{I},jy{J});
 415             dz{I}{J}             = _mm_sub_ps(iz{I},jz{J});
 416             /*     #define INNERFLOPS INNERFLOPS+3 */
 417             /* #endfor */
 418
 419             /* Calculate squared distance and things based on it */
 420             /* #for I,J in PAIRS_IJ */
 421             rsq{I}{J}            = gmx_mm_calc_rsq_ps(dx{I}{J},dy{I}{J},dz{I}{J});
 422             /*     #define INNERFLOPS INNERFLOPS+5 */
 423             /* #endfor */
 424
 425             /* #for I,J in PAIRS_IJ */
 426             /*     #if 'rinv' in INTERACTION_FLAGS[I][J] */
 427             rinv{I}{J}           = gmx_mm_invsqrt_ps(rsq{I}{J});
 428             /*         #define INNERFLOPS INNERFLOPS+5 */
 429             /*     #endif */
 430             /* #endfor */
 431
 432             /* #for I,J in PAIRS_IJ */
 433             /*     #if 'rinvsq' in INTERACTION_FLAGS[I][J] */
 434             /*         # if 'rinv' not in INTERACTION_FLAGS[I][J] */
 435             rinvsq{I}{J}         = gmx_mm_inv_ps(rsq{I}{J});
 436             /*             #define INNERFLOPS INNERFLOPS+4 */
 437             /*         #else */
 438             rinvsq{I}{J}         = _mm_mul_ps(rinv{I}{J},rinv{I}{J});
 439             /*             #define INNERFLOPS INNERFLOPS+1 */
 440             /*         #endif */
 441             /*     #endif */
 442             /* #endfor */
 443
 444             /* #if not 'Water' in GEOMETRY_J */
 445             /* Load parameters for j particles */
 446             /*     #for J in PARTICLES_ELEC_J */
 447             jq{J}              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+{J},charge+jnrB+{J},
 448                                                               charge+jnrC+{J},charge+jnrD+{J});
 449             /*         #if KERNEL_ELEC=='GeneralizedBorn' */
 450             isaj{J}            = gmx_mm_load_4real_swizzle_ps(invsqrta+jnrA+{J},invsqrta+jnrB+{J},
 451                                                               invsqrta+jnrC+{J},invsqrta+jnrD+{J});
 452             /*         #endif */
 453             /*     #endfor */
 454             /*     #for J in PARTICLES_VDW_J */
 455             vdwjidx{J}A        = 2*vdwtype[jnrA+{J}];
 456             vdwjidx{J}B        = 2*vdwtype[jnrB+{J}];
 457             vdwjidx{J}C        = 2*vdwtype[jnrC+{J}];
 458             vdwjidx{J}D        = 2*vdwtype[jnrD+{J}];
 459             /*     #endfor */
 460             /* #endif */
 461
 462             /* #if 'Force' in KERNEL_VF and not 'Particle' in GEOMETRY_J */
 463             /*     #for J in PARTICLES_J */
 464             fjx{J}             = _mm_setzero_ps();
 465             fjy{J}             = _mm_setzero_ps();
 466             fjz{J}             = _mm_setzero_ps();
 467             /*     #endfor */
 468             /* #endif */
 469
 470             /* #for I,J in PAIRS_IJ */
 471
 472             /**************************
 473              * CALCULATE INTERACTIONS *
 474              **************************/
 475
 476             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 477             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 478             /*         ## We always calculate rinv/rinvsq above to enable pipelineing in compilers (performance tested on x86) */
 479             if (gmx_mm_any_lt(rsq{I}{J},rcutoff2))
 480             {
 481                 /*     #if 0    ## this and the next two lines is a hack to maintain auto-indentation in template file */
 482             }
 483             /*         #endif */
 484             /*         #define INNERFLOPS INNERFLOPS+1 */
 485             /*     #endif */
 486
 487             /*     #if 'r' in INTERACTION_FLAGS[I][J] */
 488             r{I}{J}              = _mm_mul_ps(rsq{I}{J},rinv{I}{J});
 489             /*         #if ROUND == 'Epilogue' */
 490             r{I}{J}              = _mm_andnot_ps(dummy_mask,r{I}{J});
 491             /*             #define INNERFLOPS INNERFLOPS+1 */
 492             /*         #endif */
 493             /*         #define INNERFLOPS INNERFLOPS+1 */
 494             /*     #endif */
 495
 496             /*     ## For water geometries we already loaded parameters at the start of the kernel */
 497             /*     #if not 'Water' in GEOMETRY_J */
 498             /* Compute parameters for interactions between i and j atoms */
 499             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 500             qq{I}{J}             = _mm_mul_ps(iq{I},jq{J});
 501             /*             #define INNERFLOPS INNERFLOPS+1 */
 502             /*         #endif */
 503             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 504             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset{I}+vdwjidx{J}A,
 505                                          vdwparam+vdwioffset{I}+vdwjidx{J}B,
 506                                          vdwparam+vdwioffset{I}+vdwjidx{J}C,
 507                                          vdwparam+vdwioffset{I}+vdwjidx{J}D,
 508                                          &c6_{I}{J},&c12_{I}{J});
 509             /*         #endif */
 510             /*     #endif */
 511
 512             /*     #if 'table' in INTERACTION_FLAGS[I][J] */
 513             /* Calculate table index by multiplying r with table scale and truncate to integer */
 514             rt               = _mm_mul_ps(r{I}{J},vftabscale);
 515             vfitab           = _mm_cvttps_epi32(rt);
 516             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 517             /*         #define INNERFLOPS INNERFLOPS+4                          */
 518             /*         #if 'Table' in KERNEL_ELEC and 'Table' in KERNEL_VDW     */
 519             /*             ## 3 tables, 4 bytes per point: multiply index by 12 */
 520             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 521             /*         #elif 'Table' in KERNEL_ELEC                             */
 522             /*             ## 1 table, 4 bytes per point: multiply index by 4   */
 523             vfitab           = _mm_slli_epi32(vfitab,2);
 524             /*         #elif 'Table' in KERNEL_VDW                              */
 525             /*             ## 2 tables, 4 bytes per point: multiply index by 8  */
 526             vfitab           = _mm_slli_epi32(vfitab,3);
 527             /*         #endif                                                   */
 528             /*     #endif */
 529
 530             /*     ## ELECTROSTATIC INTERACTIONS */
 531             /*     #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 532
 533             /*         #if KERNEL_ELEC=='Coulomb' */
 534
 535             /* COULOMB ELECTROSTATICS */
 536             velec            = _mm_mul_ps(qq{I}{J},rinv{I}{J});
 537             /*             #define INNERFLOPS INNERFLOPS+1 */
 538             /*             #if 'Force' in KERNEL_VF */
 539             felec            = _mm_mul_ps(velec,rinvsq{I}{J});
 540             /*                 #define INNERFLOPS INNERFLOPS+2 */
 541             /*             #endif */
 542
 543             /*         #elif KERNEL_ELEC=='ReactionField' */
 544
 545             /* REACTION-FIELD ELECTROSTATICS */
 546             /*             #if 'Potential' in KERNEL_VF */
 547             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_add_ps(rinv{I}{J},_mm_mul_ps(krf,rsq{I}{J})),crf));
 548             /*                 #define INNERFLOPS INNERFLOPS+4 */
 549             /*             #endif */
 550             /*             #if 'Force' in KERNEL_VF */
 551             felec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_mul_ps(rinv{I}{J},rinvsq{I}{J}),krf2));
 552             /*                 #define INNERFLOPS INNERFLOPS+3 */
 553             /*             #endif */
 554
 555             /*         #elif KERNEL_ELEC=='GeneralizedBorn' */
 556
 557             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 558             isaprod          = _mm_mul_ps(isai{I},isaj{J});
 559             gbqqfactor       = _mm_xor_ps(signbit,_mm_mul_ps(qq{I}{J},_mm_mul_ps(isaprod,gbinvepsdiff)));
 560             gbscale          = _mm_mul_ps(isaprod,gbtabscale);
 561             /*             #define INNERFLOPS INNERFLOPS+5 */
 562
 563             /* Calculate generalized born table index - this is a separate table from the normal one,
 564              * but we use the same procedure by multiplying r with scale and truncating to integer.
 565              */
 566             rt               = _mm_mul_ps(r{I}{J},gbscale);
 567             gbitab           = _mm_cvttps_epi32(rt);
 568             gbeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 569             gbitab           = _mm_slli_epi32(gbitab,2);
 570             Y                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,0) );
 571             F                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,1) );
 572             G                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,2) );
 573             H                = _mm_load_ps( gbtab + gmx_mm_extract_epi32(gbitab,3) );
 574             _MM_TRANSPOSE4_PS(Y,F,G,H);
 575             Heps             = _mm_mul_ps(gbeps,H);
 576             Fp               = _mm_add_ps(F,_mm_mul_ps(gbeps,_mm_add_ps(G,Heps)));
 577             VV               = _mm_add_ps(Y,_mm_mul_ps(gbeps,Fp));
 578             vgb              = _mm_mul_ps(gbqqfactor,VV);
 579             /*             #define INNERFLOPS INNERFLOPS+10 */
 580
 581             /*             #if 'Force' in KERNEL_VF */
 582             FF               = _mm_add_ps(Fp,_mm_mul_ps(gbeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 583             fgb              = _mm_mul_ps(gbqqfactor,_mm_mul_ps(FF,gbscale));
 584             dvdatmp          = _mm_mul_ps(minushalf,_mm_add_ps(vgb,_mm_mul_ps(fgb,r{I}{J})));
 585             dvdasum          = _mm_add_ps(dvdasum,dvdatmp);
 586             /*                 #if ROUND == 'Loop' */
 587             fjptrA           = dvda+jnrA;
 588             fjptrB           = dvda+jnrB;
 589             fjptrC           = dvda+jnrC;
 590             fjptrD           = dvda+jnrD;
 591             /*                 #else */
 592             /* 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. */
 593             fjptrA             = (jnrlistA>=0) ? dvda+jnrA : scratch;
 594             fjptrB             = (jnrlistB>=0) ? dvda+jnrB : scratch;
 595             fjptrC             = (jnrlistC>=0) ? dvda+jnrC : scratch;
 596             fjptrD             = (jnrlistD>=0) ? dvda+jnrD : scratch;
 597             /*                 #endif */
 598             gmx_mm_increment_4real_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,_mm_mul_ps(dvdatmp,_mm_mul_ps(isaj{J},isaj{J})));
 599             /*                 #define INNERFLOPS INNERFLOPS+13 */
 600             /*             #endif */
 601             velec            = _mm_mul_ps(qq{I}{J},rinv{I}{J});
 602             /*                 #define INNERFLOPS INNERFLOPS+1 */
 603             /*             #if 'Force' in KERNEL_VF */
 604             felec            = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(velec,rinv{I}{J}),fgb),rinv{I}{J});
 605             /*                 #define INNERFLOPS INNERFLOPS+3 */
 606             /*             #endif */
 607
 608             /*         #elif KERNEL_ELEC=='Ewald' */
 609             /* EWALD ELECTROSTATICS */
 610
 611             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 612             ewrt             = _mm_mul_ps(r{I}{J},ewtabscale);
 613             ewitab           = _mm_cvttps_epi32(ewrt);
 614             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 615             /*             #define INNERFLOPS INNERFLOPS+4 */
 616             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_ELEC=='PotentialSwitch' */
 617             ewitab           = _mm_slli_epi32(ewitab,2);
 618             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 619             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 620             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 621             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 622             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 623             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 624             /*                 #define INNERFLOPS INNERFLOPS+2 */
 625             /*                 #if KERNEL_MOD_ELEC=='PotentialShift' */
 626             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 627             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(_mm_sub_ps(rinv{I}{J},sh_ewald),velec));
 628             /*                     #define INNERFLOPS INNERFLOPS+7 */
 629             /*                 #else */
 630             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 631             velec            = _mm_mul_ps(qq{I}{J},_mm_sub_ps(rinv{I}{J},velec));
 632             /*                     #define INNERFLOPS INNERFLOPS+6 */
 633             /*                 #endif */
 634             /*                 #if 'Force' in KERNEL_VF */
 635             felec            = _mm_mul_ps(_mm_mul_ps(qq{I}{J},rinv{I}{J}),_mm_sub_ps(rinvsq{I}{J},felec));
 636             /*                      #define INNERFLOPS INNERFLOPS+3 */
 637             /*                 #endif */
 638             /*             #elif KERNEL_VF=='Force' */
 639             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
 640                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
 641                                          &ewtabF,&ewtabFn);
 642             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
 643             felec            = _mm_mul_ps(_mm_mul_ps(qq{I}{J},rinv{I}{J}),_mm_sub_ps(rinvsq{I}{J},felec));
 644             /*                 #define INNERFLOPS INNERFLOPS+7 */
 645             /*             #endif */
 646
 647             /*         #elif KERNEL_ELEC=='CubicSplineTable' */
 648
 649             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 650             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 651             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 652             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 653             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 654             _MM_TRANSPOSE4_PS(Y,F,G,H);
 655             Heps             = _mm_mul_ps(vfeps,H);
 656             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 657             /*             #define INNERFLOPS INNERFLOPS+4 */
 658             /*             #if 'Potential' in KERNEL_VF */
 659             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 660             velec            = _mm_mul_ps(qq{I}{J},VV);
 661             /*                 #define INNERFLOPS INNERFLOPS+3 */
 662             /*             #endif */
 663             /*             #if 'Force' in KERNEL_VF */
 664             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 665             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq{I}{J},FF),_mm_mul_ps(vftabscale,rinv{I}{J})));
 666             /*                 #define INNERFLOPS INNERFLOPS+7 */
 667             /*             #endif */
 668             /*         #endif */
 669             /*         ## End of check for electrostatics interaction forms */
 670             /*     #endif */
 671             /*     ## END OF ELECTROSTATIC INTERACTION CHECK FOR PAIR I-J */
 672
 673             /*     #if 'vdw' in INTERACTION_FLAGS[I][J] */
 674
 675             /*         #if KERNEL_VDW=='LennardJones' */
 676
 677             /* LENNARD-JONES DISPERSION/REPULSION */
 678
 679             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq{I}{J},rinvsq{I}{J}),rinvsq{I}{J});
 680             /*             #define INNERFLOPS INNERFLOPS+2 */
 681             /*             #if 'Potential' in KERNEL_VF or KERNEL_MOD_VDW=='PotentialSwitch' */
 682             vvdw6            = _mm_mul_ps(c6_{I}{J},rinvsix);
 683             vvdw12           = _mm_mul_ps(c12_{I}{J},_mm_mul_ps(rinvsix,rinvsix));
 684             /*                 #define INNERFLOPS INNERFLOPS+3 */
 685             /*                 #if KERNEL_MOD_VDW=='PotentialShift' */
 686             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) ,
 687                                           _mm_mul_ps( _mm_sub_ps(vvdw6,_mm_mul_ps(c6_{I}{J},sh_vdw_invrcut6)),one_sixth));
 688             /*                     #define INNERFLOPS INNERFLOPS+8 */
 689             /*                 #else */
 690             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 691             /*                     #define INNERFLOPS INNERFLOPS+3 */
 692             /*                 #endif */
 693             /*                 ## Check for force inside potential check, i.e. this means we already did the potential part */
 694             /*                 #if 'Force' in KERNEL_VF */
 695             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq{I}{J});
 696             /*                     #define INNERFLOPS INNERFLOPS+2 */
 697             /*                 #endif */
 698             /*             #elif KERNEL_VF=='Force' */
 699             /*                 ## Force-only LennardJones makes it possible to save 1 flop (they do add up...) */
 700             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_{I}{J},rinvsix),c6_{I}{J}),_mm_mul_ps(rinvsix,rinvsq{I}{J}));
 701             /*                 #define INNERFLOPS INNERFLOPS+4 */
 702             /*             #endif */
 703
 704             /*         #elif KERNEL_VDW=='CubicSplineTable' */
 705
 706             /* CUBIC SPLINE TABLE DISPERSION */
 707             /*             #if 'Table' in KERNEL_ELEC */
 708             vfitab           = _mm_add_epi32(vfitab,ifour);
 709             /*             #endif                     */
 710             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 711             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 712             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 713             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 714             _MM_TRANSPOSE4_PS(Y,F,G,H);
 715             Heps             = _mm_mul_ps(vfeps,H);
 716             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 717             /*             #define INNERFLOPS INNERFLOPS+4 */
 718             /*             #if 'Potential' in KERNEL_VF */
 719             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 720             vvdw6            = _mm_mul_ps(c6_{I}{J},VV);
 721             /*                 #define INNERFLOPS INNERFLOPS+3 */
 722             /*             #endif */
 723             /*             #if 'Force' in KERNEL_VF */
 724             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 725             fvdw6            = _mm_mul_ps(c6_{I}{J},FF);
 726             /*                 #define INNERFLOPS INNERFLOPS+4 */
 727             /*             #endif */
 728
 729             /* CUBIC SPLINE TABLE REPULSION */
 730             vfitab           = _mm_add_epi32(vfitab,ifour);
 731             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 732             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 733             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 734             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 735             _MM_TRANSPOSE4_PS(Y,F,G,H);
 736             Heps             = _mm_mul_ps(vfeps,H);
 737             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 738             /*             #define INNERFLOPS INNERFLOPS+4 */
 739             /*             #if 'Potential' in KERNEL_VF */
 740             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 741             vvdw12           = _mm_mul_ps(c12_{I}{J},VV);
 742             /*                 #define INNERFLOPS INNERFLOPS+3 */
 743             /*             #endif */
 744             /*             #if 'Force' in KERNEL_VF */
 745             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 746             fvdw12           = _mm_mul_ps(c12_{I}{J},FF);
 747             /*                 #define INNERFLOPS INNERFLOPS+5 */
 748             /*             #endif */
 749             /*             #if 'Potential' in KERNEL_VF */
 750             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 751             /*                 #define INNERFLOPS INNERFLOPS+1 */
 752             /*             #endif */
 753             /*             #if 'Force' in KERNEL_VF */
 754             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv{I}{J})));
 755             /*                 #define INNERFLOPS INNERFLOPS+4 */
 756             /*             #endif */
 757             /*         #endif */
 758             /*         ## End of check for vdw interaction forms */
 759             /*     #endif */
 760             /*     ## END OF VDW INTERACTION CHECK FOR PAIR I-J */
 761
 762             /*     #if 'switch' in INTERACTION_FLAGS[I][J] */
 763             d                = _mm_sub_ps(r{I}{J},rswitch);
 764             d                = _mm_max_ps(d,_mm_setzero_ps());
 765             d2               = _mm_mul_ps(d,d);
 766             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)))))));
 767             /*         #define INNERFLOPS INNERFLOPS+10 */
 768
 769             /*         #if 'Force' in KERNEL_VF */
 770             dsw              = _mm_mul_ps(d2,_mm_add_ps(swF2,_mm_mul_ps(d,_mm_add_ps(swF3,_mm_mul_ps(d,swF4)))));
 771             /*             #define INNERFLOPS INNERFLOPS+5 */
 772             /*         #endif */
 773
 774             /* Evaluate switch function */
 775             /*         #if 'Force' in KERNEL_VF */
 776             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 777             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 778             felec            = _mm_sub_ps( _mm_mul_ps(felec,sw) , _mm_mul_ps(rinv{I}{J},_mm_mul_ps(velec,dsw)) );
 779             /*                 #define INNERFLOPS INNERFLOPS+4 */
 780             /*             #endif */
 781             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 782             fvdw             = _mm_sub_ps( _mm_mul_ps(fvdw,sw) , _mm_mul_ps(rinv{I}{J},_mm_mul_ps(vvdw,dsw)) );
 783             /*                 #define INNERFLOPS INNERFLOPS+4 */
 784             /*             #endif */
 785             /*         #endif */
 786             /*         #if 'Potential' in KERNEL_VF */
 787             /*             #if 'electrostatics' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_ELEC=='PotentialSwitch' */
 788             velec            = _mm_mul_ps(velec,sw);
 789             /*                 #define INNERFLOPS INNERFLOPS+1 */
 790             /*             #endif */
 791             /*             #if 'vdw' in INTERACTION_FLAGS[I][J] and KERNEL_MOD_VDW=='PotentialSwitch' */
 792             vvdw             = _mm_mul_ps(vvdw,sw);
 793             /*                 #define INNERFLOPS INNERFLOPS+1 */
 794             /*             #endif */
 795             /*         #endif */
 796             /*     #endif */
 797             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 798             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 799             cutoff_mask      = _mm_cmplt_ps(rsq{I}{J},rcutoff2);
 800             /*         #define INNERFLOPS INNERFLOPS+1 */
 801             /*     #endif */
 802
 803             /*     #if 'Potential' in KERNEL_VF */
 804             /* Update potential sum for this i atom from the interaction with this j atom. */
 805             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] */
 806             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 807             velec            = _mm_and_ps(velec,cutoff_mask);
 808             /*                 #define INNERFLOPS INNERFLOPS+1 */
 809             /*             #endif                                       */
 810             /*             #if ROUND == 'Epilogue' */
 811             velec            = _mm_andnot_ps(dummy_mask,velec);
 812             /*             #endif */
 813             velecsum         = _mm_add_ps(velecsum,velec);
 814             /*             #define INNERFLOPS INNERFLOPS+1 */
 815             /*             #if KERNEL_ELEC=='GeneralizedBorn' */
 816             /*             #if 'exactcutoff' in INTERACTION_FLAGS[I][J] */
 817             vgb              = _mm_and_ps(vgb,cutoff_mask);
 818             /*                 #define INNERFLOPS INNERFLOPS+1 */
 819             /*             #endif                                       */
 820             /*             #if ROUND == 'Epilogue' */
 821             vgb              = _mm_andnot_ps(dummy_mask,vgb);
 822             /*             #endif */
 823             vgbsum           = _mm_add_ps(vgbsum,vgb);
 824             /*                 #define INNERFLOPS INNERFLOPS+1 */
 825             /*             #endif */
 826             /*         #endif */
 827             /*         #if 'vdw' in INTERACTION_FLAGS[I][J] */
 828             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 829             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 830             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 831             /*                 #define INNERFLOPS INNERFLOPS+1 */
 832             /*             #endif                                       */
 833             /*             #if ROUND == 'Epilogue' */
 834             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 835             /*             #endif */
 836             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 837             /*             #define INNERFLOPS INNERFLOPS+1 */
 838             /*         #endif */
 839             /*     #endif */
 840
 841             /*     #if 'Force' in KERNEL_VF */
 842
 843             /*         #if 'electrostatics' in INTERACTION_FLAGS[I][J] and 'vdw' in INTERACTION_FLAGS[I][J] */
 844             fscal            = _mm_add_ps(felec,fvdw);
 845             /*             #define INNERFLOPS INNERFLOPS+1 */
 846             /*         #elif 'electrostatics' in INTERACTION_FLAGS[I][J] */
 847             fscal            = felec;
 848             /*         #elif 'vdw' in INTERACTION_FLAGS[I][J] */
 849             fscal            = fvdw;
 850             /*        #endif */
 851
 852             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 853             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 854             fscal            = _mm_and_ps(fscal,cutoff_mask);
 855             /*                 #define INNERFLOPS INNERFLOPS+1 */
 856             /*             #endif                                       */
 857
 858             /*             #if ROUND == 'Epilogue' */
 859             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 860             /*             #endif */
 861
 862             /* Calculate temporary vectorial force */
 863             tx               = _mm_mul_ps(fscal,dx{I}{J});
 864             ty               = _mm_mul_ps(fscal,dy{I}{J});
 865             tz               = _mm_mul_ps(fscal,dz{I}{J});
 866
 867             /* Update vectorial force */
 868             fix{I}             = _mm_add_ps(fix{I},tx);
 869             fiy{I}             = _mm_add_ps(fiy{I},ty);
 870             fiz{I}             = _mm_add_ps(fiz{I},tz);
 871             /*             #define INNERFLOPS INNERFLOPS+6 */
 872
 873             /* #if GEOMETRY_J == 'Particle'             */
 874             /*     #if ROUND == 'Loop' */
 875             fjptrA             = f+j_coord_offsetA;
 876             fjptrB             = f+j_coord_offsetB;
 877             fjptrC             = f+j_coord_offsetC;
 878             fjptrD             = f+j_coord_offsetD;
 879             /*     #else */
 880             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 881             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 882             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 883             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 884             /*     #endif */
 885             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 886             /*     #define INNERFLOPS INNERFLOPS+3      */
 887             /* #else                                    */
 888             fjx{J}             = _mm_add_ps(fjx{J},tx);
 889             fjy{J}             = _mm_add_ps(fjy{J},ty);
 890             fjz{J}             = _mm_add_ps(fjz{J},tz);
 891             /*     #define INNERFLOPS INNERFLOPS+3      */
 892             /* #endif                                   */
 893
 894             /*     #endif */
 895
 896             /*     ## Note special check for TIP4P-TIP4P. Since we are cutting of all hydrogen interactions we also cut the LJ-only O-O interaction */
 897             /*     #if 'exactcutoff' in INTERACTION_FLAGS[I][J] or (GEOMETRY_I=='Water4' and GEOMETRY_J=='Water4' and 'exactcutoff' in INTERACTION_FLAGS[1][1]) */
 898             /*         #if 0    ## This and next two lines is a hack to maintain indentation in template file */
 899             {
 900                 /*     #endif */
 901             }
 902             /*     #endif */
 903             /*    ## End of check for the interaction being outside the cutoff */
 904
 905             /* #endfor */
 906             /* ## End of loop over i-j interaction pairs */
 907
 908             /* #if GEOMETRY_J != 'Particle' */
 909             /*     #if ROUND == 'Loop' */
 910             fjptrA             = f+j_coord_offsetA;
 911             fjptrB             = f+j_coord_offsetB;
 912             fjptrC             = f+j_coord_offsetC;
 913             fjptrD             = f+j_coord_offsetD;
 914             /*     #else */
 915             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 916             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 917             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 918             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 919             /*     #endif */
 920             /* #endif */
 921
 922             /* #if GEOMETRY_J == 'Water3'               */
 923             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 924                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 925             /*     #define INNERFLOPS INNERFLOPS+9      */
 926             /* #elif GEOMETRY_J == 'Water4'             */
 927             /*     #if 0 in PARTICLES_J                 */
 928             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 929                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 930                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 931             /*     #define INNERFLOPS INNERFLOPS+12     */
 932             /*     #else                                */
 933             gmx_mm_decrement_3rvec_4ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 934                                                    fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 935             /*     #define INNERFLOPS INNERFLOPS+9      */
 936             /*     #endif                               */
 937             /* #endif                                   */
 938
 939             /* Inner loop uses {INNERFLOPS} flops */
 940         }
 941
 942         /* #endfor */
 943
 944         /* End of innermost loop */
 945
 946         /* #if 'Force' in KERNEL_VF */
 947         /*     #if GEOMETRY_I == 'Particle'            */
 948         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 949                                               f+i_coord_offset,fshift+i_shift_offset);
 950         /*         #define OUTERFLOPS OUTERFLOPS+6     */
 951         /*     #elif GEOMETRY_I == 'Water3'            */
 952         gmx_mm_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 953                                               f+i_coord_offset,fshift+i_shift_offset);
 954         /*         #define OUTERFLOPS OUTERFLOPS+18    */
 955         /*     #elif GEOMETRY_I == 'Water4'            */
 956         /*         #if 0 in PARTICLES_I                */
 957         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 958                                               f+i_coord_offset,fshift+i_shift_offset);
 959         /*             #define OUTERFLOPS OUTERFLOPS+24    */
 960         /*         #else                               */
 961         gmx_mm_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 962                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 963         /*             #define OUTERFLOPS OUTERFLOPS+18    */
 964         /*         #endif                              */
 965         /*     #endif                                  */
 966         /* #endif                                      */
 967
 968         /* #if 'Potential' in KERNEL_VF */
 969         ggid                        = gid[iidx];
 970         /* Update potential energies */
 971         /*     #if KERNEL_ELEC != 'None' */
 972         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 973         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 974         /*     #endif */
 975         /*     #if 'GeneralizedBorn' in KERNEL_ELEC */
 976         gmx_mm_update_1pot_ps(vgbsum,kernel_data->energygrp_polarization+ggid);
 977         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 978         /*     #endif */
 979         /*     #if KERNEL_VDW != 'None' */
 980         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 981         /*         #define OUTERFLOPS OUTERFLOPS+1 */
 982         /*     #endif */
 983         /* #endif */
 984         /*     #if 'GeneralizedBorn' in KERNEL_ELEC and 'Force' in KERNEL_VF */
 985         dvdasum = _mm_mul_ps(dvdasum, _mm_mul_ps(isai{I},isai{I}));
 986         gmx_mm_update_1pot_ps(dvdasum,dvda+inr);
 987         /*     #endif */
 988
 989         /* Increment number of inner iterations */
 990         inneriter                  += j_index_end - j_index_start;
 991
 992         /* Outer loop uses {OUTERFLOPS} flops */
 993     }
 994
 995     /* Increment number of outer iterations */
 996     outeriter        += nri;
 997
 998     /* Update outer/inner flops */
 999     /* ## NB: This is not important, it just affects the flopcount. However, since our preprocessor is */
1000     /* ## primitive and replaces aggressively even in strings inside these directives, we need to      */
1001     /* ## assemble the main part of the name (containing KERNEL/ELEC/VDW) directly in the source.      */
1002     /* #if GEOMETRY_I == 'Water3'            */
1003     /*     #define ISUFFIX '_W3'             */
1004     /* #elif GEOMETRY_I == 'Water4'          */
1005     /*     #define ISUFFIX '_W4'             */
1006     /* #else                                 */
1007     /*     #define ISUFFIX ''                */
1008     /* #endif                                */
1009     /* #if GEOMETRY_J == 'Water3'            */
1010     /*     #define JSUFFIX 'W3'              */
1011     /* #elif GEOMETRY_J == 'Water4'          */
1012     /*     #define JSUFFIX 'W4'              */
1013     /* #else                                 */
1014     /*     #define JSUFFIX ''                */
1015     /* #endif                                */
1016     /* #if 'PotentialAndForce' in KERNEL_VF  */
1017     /*     #define VFSUFFIX  '_VF'           */
1018     /* #elif 'Potential' in KERNEL_VF        */
1019     /*     #define VFSUFFIX '_V'             */
1020     /* #else                                 */
1021     /*     #define VFSUFFIX '_F'             */
1022     /* #endif                                */
1023
1024     /* #if KERNEL_ELEC != 'None' and KERNEL_VDW != 'None' */
1025     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1026     /* #elif KERNEL_ELEC != 'None' */
1027     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1028     /* #else */
1029     inc_nrnb(nrnb,eNR_NBKERNEL_VDW{ISUFFIX}{JSUFFIX}{VFSUFFIX},outeriter*{OUTERFLOPS} + inneriter*{INNERFLOPS});
1030     /* #endif  */
1031 }