src/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwLJ_GeomW3P1_sparc64_hpc_ace_double.c

   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 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Coulomb
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCoul_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist * gmx_restrict                nlist,
  61                      rvec * gmx_restrict                    xx,
  62                      rvec * gmx_restrict                    ff,
  63                      t_forcerec * gmx_restrict              fr,
  64                      t_mdatoms * gmx_restrict               mdatoms,
  65                      nb_kernel_data_t * gmx_restrict        kernel_data,
  66                      t_nrnb * gmx_restrict                  nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     /* Setup water-specific parameters */
 124     inr              = nlist->iinr[0];
 125     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 126     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 127     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 128     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 129
 130     /* Avoid stupid compiler warnings */
 131     jnrA = jnrB = 0;
 132     j_coord_offsetA = 0;
 133     j_coord_offsetB = 0;
 134
 135     outeriter        = 0;
 136     inneriter        = 0;
 137
 138     /* Start outer loop over neighborlists */
 139     for(iidx=0; iidx<nri; iidx++)
 140     {
 141         /* Load shift vector for this list */
 142         i_shift_offset   = DIM*shiftidx[iidx];
 143
 144         /* Load limits for loop over neighbors */
 145         j_index_start    = jindex[iidx];
 146         j_index_end      = jindex[iidx+1];
 147
 148         /* Get outer coordinate index */
 149         inr              = iinr[iidx];
 150         i_coord_offset   = DIM*inr;
 151
 152         /* Load i particle coords and add shift vector */
 153         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 154                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 155
 156         fix0             = _fjsp_setzero_v2r8();
 157         fiy0             = _fjsp_setzero_v2r8();
 158         fiz0             = _fjsp_setzero_v2r8();
 159         fix1             = _fjsp_setzero_v2r8();
 160         fiy1             = _fjsp_setzero_v2r8();
 161         fiz1             = _fjsp_setzero_v2r8();
 162         fix2             = _fjsp_setzero_v2r8();
 163         fiy2             = _fjsp_setzero_v2r8();
 164         fiz2             = _fjsp_setzero_v2r8();
 165
 166         /* Reset potential sums */
 167         velecsum         = _fjsp_setzero_v2r8();
 168         vvdwsum          = _fjsp_setzero_v2r8();
 169
 170         /* Start inner kernel loop */
 171         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 172         {
 173
 174             /* Get j neighbor index, and coordinate index */
 175             jnrA             = jjnr[jidx];
 176             jnrB             = jjnr[jidx+1];
 177             j_coord_offsetA  = DIM*jnrA;
 178             j_coord_offsetB  = DIM*jnrB;
 179
 180             /* load j atom coordinates */
 181             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 182                                               &jx0,&jy0,&jz0);
 183
 184             /* Calculate displacement vector */
 185             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 186             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 187             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 188             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 189             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 190             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 191             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 192             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 193             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 194
 195             /* Calculate squared distance and things based on it */
 196             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 197             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 198             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 199
 200             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 201             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 202             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 203
 204             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 205             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 206             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 207
 208             /* Load parameters for j particles */
 209             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 210             vdwjidx0A        = 2*vdwtype[jnrA+0];
 211             vdwjidx0B        = 2*vdwtype[jnrB+0];
 212
 213             fjx0             = _fjsp_setzero_v2r8();
 214             fjy0             = _fjsp_setzero_v2r8();
 215             fjz0             = _fjsp_setzero_v2r8();
 216
 217             /**************************
 218              * CALCULATE INTERACTIONS *
 219              **************************/
 220
 221             /* Compute parameters for interactions between i and j atoms */
 222             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 223             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 224                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 225
 226             /* COULOMB ELECTROSTATICS */
 227             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 228             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 229
 230             /* LENNARD-JONES DISPERSION/REPULSION */
 231
 232             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 233             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 234             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 235             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 236             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 237
 238             /* Update potential sum for this i atom from the interaction with this j atom. */
 239             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 240             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 241
 242             fscal            = _fjsp_add_v2r8(felec,fvdw);
 243
 244             /* Update vectorial force */
 245             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 246             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 247             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 248
 249             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 250             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 251             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 252
 253             /**************************
 254              * CALCULATE INTERACTIONS *
 255              **************************/
 256
 257             /* Compute parameters for interactions between i and j atoms */
 258             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 259
 260             /* COULOMB ELECTROSTATICS */
 261             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 262             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 263
 264             /* Update potential sum for this i atom from the interaction with this j atom. */
 265             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 266
 267             fscal            = felec;
 268
 269             /* Update vectorial force */
 270             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 271             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 272             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 273
 274             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 275             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 276             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 277
 278             /**************************
 279              * CALCULATE INTERACTIONS *
 280              **************************/
 281
 282             /* Compute parameters for interactions between i and j atoms */
 283             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 284
 285             /* COULOMB ELECTROSTATICS */
 286             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 287             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 291
 292             fscal            = felec;
 293
 294             /* Update vectorial force */
 295             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 296             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 297             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 298
 299             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 300             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 301             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 302
 303             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 304
 305             /* Inner loop uses 108 flops */
 306         }
 307
 308         if(jidx<j_index_end)
 309         {
 310
 311             jnrA             = jjnr[jidx];
 312             j_coord_offsetA  = DIM*jnrA;
 313
 314             /* load j atom coordinates */
 315             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 316                                               &jx0,&jy0,&jz0);
 317
 318             /* Calculate displacement vector */
 319             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 320             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 321             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 322             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 323             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 324             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 325             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 326             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 327             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 328
 329             /* Calculate squared distance and things based on it */
 330             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 331             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 332             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 333
 334             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 335             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 336             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 337
 338             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 339             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 340             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 341
 342             /* Load parameters for j particles */
 343             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 344             vdwjidx0A        = 2*vdwtype[jnrA+0];
 345
 346             fjx0             = _fjsp_setzero_v2r8();
 347             fjy0             = _fjsp_setzero_v2r8();
 348             fjz0             = _fjsp_setzero_v2r8();
 349
 350             /**************************
 351              * CALCULATE INTERACTIONS *
 352              **************************/
 353
 354             /* Compute parameters for interactions between i and j atoms */
 355             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 356             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 357
 358             /* COULOMB ELECTROSTATICS */
 359             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 360             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 361
 362             /* LENNARD-JONES DISPERSION/REPULSION */
 363
 364             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 365             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 366             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 367             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 368             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 369
 370             /* Update potential sum for this i atom from the interaction with this j atom. */
 371             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 372             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 373             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 374             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 375
 376             fscal            = _fjsp_add_v2r8(felec,fvdw);
 377
 378             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 379
 380             /* Update vectorial force */
 381             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 382             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 383             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 384
 385             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 386             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 387             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 388
 389             /**************************
 390              * CALCULATE INTERACTIONS *
 391              **************************/
 392
 393             /* Compute parameters for interactions between i and j atoms */
 394             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 395
 396             /* COULOMB ELECTROSTATICS */
 397             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 398             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 399
 400             /* Update potential sum for this i atom from the interaction with this j atom. */
 401             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 402             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 403
 404             fscal            = felec;
 405
 406             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 407
 408             /* Update vectorial force */
 409             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 410             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 411             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 412
 413             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 414             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 415             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 416
 417             /**************************
 418              * CALCULATE INTERACTIONS *
 419              **************************/
 420
 421             /* Compute parameters for interactions between i and j atoms */
 422             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 423
 424             /* COULOMB ELECTROSTATICS */
 425             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 426             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 427
 428             /* Update potential sum for this i atom from the interaction with this j atom. */
 429             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 430             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 431
 432             fscal            = felec;
 433
 434             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 435
 436             /* Update vectorial force */
 437             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 438             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 439             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 440
 441             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 442             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 443             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 444
 445             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 446
 447             /* Inner loop uses 108 flops */
 448         }
 449
 450         /* End of innermost loop */
 451
 452         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 453                                               f+i_coord_offset,fshift+i_shift_offset);
 454
 455         ggid                        = gid[iidx];
 456         /* Update potential energies */
 457         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 458         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 459
 460         /* Increment number of inner iterations */
 461         inneriter                  += j_index_end - j_index_start;
 462
 463         /* Outer loop uses 20 flops */
 464     }
 465
 466     /* Increment number of outer iterations */
 467     outeriter        += nri;
 468
 469     /* Update outer/inner flops */
 470
 471     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*108);
 472 }
 473 /*
 474  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 475  * Electrostatics interaction: Coulomb
 476  * VdW interaction:            LennardJones
 477  * Geometry:                   Water3-Particle
 478  * Calculate force/pot:        Force
 479  */
 480 void
 481 nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 482                     (t_nblist * gmx_restrict                nlist,
 483                      rvec * gmx_restrict                    xx,
 484                      rvec * gmx_restrict                    ff,
 485                      t_forcerec * gmx_restrict              fr,
 486                      t_mdatoms * gmx_restrict               mdatoms,
 487                      nb_kernel_data_t * gmx_restrict        kernel_data,
 488                      t_nrnb * gmx_restrict                  nrnb)
 489 {
 490     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 491      * just 0 for non-waters.
 492      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 493      * jnr indices corresponding to data put in the four positions in the SIMD register.
 494      */
 495     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 496     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 497     int              jnrA,jnrB;
 498     int              j_coord_offsetA,j_coord_offsetB;
 499     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 500     real             rcutoff_scalar;
 501     real             *shiftvec,*fshift,*x,*f;
 502     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 503     int              vdwioffset0;
 504     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 505     int              vdwioffset1;
 506     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 507     int              vdwioffset2;
 508     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 509     int              vdwjidx0A,vdwjidx0B;
 510     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 511     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 512     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 513     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 514     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 515     real             *charge;
 516     int              nvdwtype;
 517     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 518     int              *vdwtype;
 519     real             *vdwparam;
 520     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 521     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 522     _fjsp_v2r8       itab_tmp;
 523     _fjsp_v2r8       dummy_mask,cutoff_mask;
 524     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 525     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 526     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 527
 528     x                = xx[0];
 529     f                = ff[0];
 530
 531     nri              = nlist->nri;
 532     iinr             = nlist->iinr;
 533     jindex           = nlist->jindex;
 534     jjnr             = nlist->jjnr;
 535     shiftidx         = nlist->shift;
 536     gid              = nlist->gid;
 537     shiftvec         = fr->shift_vec[0];
 538     fshift           = fr->fshift[0];
 539     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 540     charge           = mdatoms->chargeA;
 541     nvdwtype         = fr->ntype;
 542     vdwparam         = fr->nbfp;
 543     vdwtype          = mdatoms->typeA;
 544
 545     /* Setup water-specific parameters */
 546     inr              = nlist->iinr[0];
 547     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 548     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 549     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 550     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 551
 552     /* Avoid stupid compiler warnings */
 553     jnrA = jnrB = 0;
 554     j_coord_offsetA = 0;
 555     j_coord_offsetB = 0;
 556
 557     outeriter        = 0;
 558     inneriter        = 0;
 559
 560     /* Start outer loop over neighborlists */
 561     for(iidx=0; iidx<nri; iidx++)
 562     {
 563         /* Load shift vector for this list */
 564         i_shift_offset   = DIM*shiftidx[iidx];
 565
 566         /* Load limits for loop over neighbors */
 567         j_index_start    = jindex[iidx];
 568         j_index_end      = jindex[iidx+1];
 569
 570         /* Get outer coordinate index */
 571         inr              = iinr[iidx];
 572         i_coord_offset   = DIM*inr;
 573
 574         /* Load i particle coords and add shift vector */
 575         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 576                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 577
 578         fix0             = _fjsp_setzero_v2r8();
 579         fiy0             = _fjsp_setzero_v2r8();
 580         fiz0             = _fjsp_setzero_v2r8();
 581         fix1             = _fjsp_setzero_v2r8();
 582         fiy1             = _fjsp_setzero_v2r8();
 583         fiz1             = _fjsp_setzero_v2r8();
 584         fix2             = _fjsp_setzero_v2r8();
 585         fiy2             = _fjsp_setzero_v2r8();
 586         fiz2             = _fjsp_setzero_v2r8();
 587
 588         /* Start inner kernel loop */
 589         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 590         {
 591
 592             /* Get j neighbor index, and coordinate index */
 593             jnrA             = jjnr[jidx];
 594             jnrB             = jjnr[jidx+1];
 595             j_coord_offsetA  = DIM*jnrA;
 596             j_coord_offsetB  = DIM*jnrB;
 597
 598             /* load j atom coordinates */
 599             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 600                                               &jx0,&jy0,&jz0);
 601
 602             /* Calculate displacement vector */
 603             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 604             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 605             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 606             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 607             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 608             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 609             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 610             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 611             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 612
 613             /* Calculate squared distance and things based on it */
 614             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 615             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 616             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 617
 618             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 619             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 620             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 621
 622             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 623             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 624             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 625
 626             /* Load parameters for j particles */
 627             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 628             vdwjidx0A        = 2*vdwtype[jnrA+0];
 629             vdwjidx0B        = 2*vdwtype[jnrB+0];
 630
 631             fjx0             = _fjsp_setzero_v2r8();
 632             fjy0             = _fjsp_setzero_v2r8();
 633             fjz0             = _fjsp_setzero_v2r8();
 634
 635             /**************************
 636              * CALCULATE INTERACTIONS *
 637              **************************/
 638
 639             /* Compute parameters for interactions between i and j atoms */
 640             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 641             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 642                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 643
 644             /* COULOMB ELECTROSTATICS */
 645             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 646             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 647
 648             /* LENNARD-JONES DISPERSION/REPULSION */
 649
 650             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 651             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 652
 653             fscal            = _fjsp_add_v2r8(felec,fvdw);
 654
 655             /* Update vectorial force */
 656             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 657             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 658             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 659
 660             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 661             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 662             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 663
 664             /**************************
 665              * CALCULATE INTERACTIONS *
 666              **************************/
 667
 668             /* Compute parameters for interactions between i and j atoms */
 669             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 670
 671             /* COULOMB ELECTROSTATICS */
 672             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 673             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 674
 675             fscal            = felec;
 676
 677             /* Update vectorial force */
 678             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 679             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 680             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 681
 682             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 683             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 684             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 685
 686             /**************************
 687              * CALCULATE INTERACTIONS *
 688              **************************/
 689
 690             /* Compute parameters for interactions between i and j atoms */
 691             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 692
 693             /* COULOMB ELECTROSTATICS */
 694             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 695             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 696
 697             fscal            = felec;
 698
 699             /* Update vectorial force */
 700             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 701             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 702             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 703
 704             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 705             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 706             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 707
 708             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 709
 710             /* Inner loop uses 100 flops */
 711         }
 712
 713         if(jidx<j_index_end)
 714         {
 715
 716             jnrA             = jjnr[jidx];
 717             j_coord_offsetA  = DIM*jnrA;
 718
 719             /* load j atom coordinates */
 720             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 721                                               &jx0,&jy0,&jz0);
 722
 723             /* Calculate displacement vector */
 724             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 725             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 726             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 727             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 728             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 729             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 730             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 731             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 732             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 733
 734             /* Calculate squared distance and things based on it */
 735             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 736             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 737             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 738
 739             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 740             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 741             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 742
 743             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 744             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 745             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 746
 747             /* Load parameters for j particles */
 748             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 749             vdwjidx0A        = 2*vdwtype[jnrA+0];
 750
 751             fjx0             = _fjsp_setzero_v2r8();
 752             fjy0             = _fjsp_setzero_v2r8();
 753             fjz0             = _fjsp_setzero_v2r8();
 754
 755             /**************************
 756              * CALCULATE INTERACTIONS *
 757              **************************/
 758
 759             /* Compute parameters for interactions between i and j atoms */
 760             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 761             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 762
 763             /* COULOMB ELECTROSTATICS */
 764             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 765             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 766
 767             /* LENNARD-JONES DISPERSION/REPULSION */
 768
 769             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 770             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 771
 772             fscal            = _fjsp_add_v2r8(felec,fvdw);
 773
 774             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 775
 776             /* Update vectorial force */
 777             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 778             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 779             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 780
 781             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 782             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 783             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 784
 785             /**************************
 786              * CALCULATE INTERACTIONS *
 787              **************************/
 788
 789             /* Compute parameters for interactions between i and j atoms */
 790             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 794             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 795
 796             fscal            = felec;
 797
 798             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 799
 800             /* Update vectorial force */
 801             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 802             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 803             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 804
 805             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 806             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 807             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 808
 809             /**************************
 810              * CALCULATE INTERACTIONS *
 811              **************************/
 812
 813             /* Compute parameters for interactions between i and j atoms */
 814             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 815
 816             /* COULOMB ELECTROSTATICS */
 817             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 818             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 819
 820             fscal            = felec;
 821
 822             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 823
 824             /* Update vectorial force */
 825             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 826             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 827             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 828
 829             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 830             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 831             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 832
 833             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 834
 835             /* Inner loop uses 100 flops */
 836         }
 837
 838         /* End of innermost loop */
 839
 840         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 841                                               f+i_coord_offset,fshift+i_shift_offset);
 842
 843         /* Increment number of inner iterations */
 844         inneriter                  += j_index_end - j_index_start;
 845
 846         /* Outer loop uses 18 flops */
 847     }
 848
 849     /* Increment number of outer iterations */
 850     outeriter        += nri;
 851
 852     /* Update outer/inner flops */
 853
 854     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*100);
 855 }