src/gromacs/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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source 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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/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_unused * 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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 357                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 358
 359             /* COULOMB ELECTROSTATICS */
 360             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 361             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 362
 363             /* LENNARD-JONES DISPERSION/REPULSION */
 364
 365             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 366             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 367             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 368             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 369             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 370
 371             /* Update potential sum for this i atom from the interaction with this j atom. */
 372             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 373             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 374             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 375             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 376
 377             fscal            = _fjsp_add_v2r8(felec,fvdw);
 378
 379             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 380
 381             /* Update vectorial force */
 382             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 383             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 384             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 385
 386             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 387             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 388             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 389
 390             /**************************
 391              * CALCULATE INTERACTIONS *
 392              **************************/
 393
 394             /* Compute parameters for interactions between i and j atoms */
 395             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 396
 397             /* COULOMB ELECTROSTATICS */
 398             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 399             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 400
 401             /* Update potential sum for this i atom from the interaction with this j atom. */
 402             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 403             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 404
 405             fscal            = felec;
 406
 407             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 408
 409             /* Update vectorial force */
 410             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 411             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 412             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 413
 414             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 415             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 416             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 417
 418             /**************************
 419              * CALCULATE INTERACTIONS *
 420              **************************/
 421
 422             /* Compute parameters for interactions between i and j atoms */
 423             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 424
 425             /* COULOMB ELECTROSTATICS */
 426             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 427             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 428
 429             /* Update potential sum for this i atom from the interaction with this j atom. */
 430             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 431             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 432
 433             fscal            = felec;
 434
 435             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 436
 437             /* Update vectorial force */
 438             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 439             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 440             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 441
 442             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 443             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 444             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 445
 446             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 447
 448             /* Inner loop uses 108 flops */
 449         }
 450
 451         /* End of innermost loop */
 452
 453         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 454                                               f+i_coord_offset,fshift+i_shift_offset);
 455
 456         ggid                        = gid[iidx];
 457         /* Update potential energies */
 458         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 459         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 460
 461         /* Increment number of inner iterations */
 462         inneriter                  += j_index_end - j_index_start;
 463
 464         /* Outer loop uses 20 flops */
 465     }
 466
 467     /* Increment number of outer iterations */
 468     outeriter        += nri;
 469
 470     /* Update outer/inner flops */
 471
 472     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*108);
 473 }
 474 /*
 475  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 476  * Electrostatics interaction: Coulomb
 477  * VdW interaction:            LennardJones
 478  * Geometry:                   Water3-Particle
 479  * Calculate force/pot:        Force
 480  */
 481 void
 482 nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 483                     (t_nblist                    * gmx_restrict       nlist,
 484                      rvec                        * gmx_restrict          xx,
 485                      rvec                        * gmx_restrict          ff,
 486                      t_forcerec                  * gmx_restrict          fr,
 487                      t_mdatoms                   * gmx_restrict     mdatoms,
 488                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 489                      t_nrnb                      * gmx_restrict        nrnb)
 490 {
 491     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 492      * just 0 for non-waters.
 493      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 494      * jnr indices corresponding to data put in the four positions in the SIMD register.
 495      */
 496     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 497     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 498     int              jnrA,jnrB;
 499     int              j_coord_offsetA,j_coord_offsetB;
 500     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 501     real             rcutoff_scalar;
 502     real             *shiftvec,*fshift,*x,*f;
 503     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 504     int              vdwioffset0;
 505     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 506     int              vdwioffset1;
 507     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 508     int              vdwioffset2;
 509     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 510     int              vdwjidx0A,vdwjidx0B;
 511     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 512     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 513     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 514     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 515     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 516     real             *charge;
 517     int              nvdwtype;
 518     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 519     int              *vdwtype;
 520     real             *vdwparam;
 521     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 522     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 523     _fjsp_v2r8       itab_tmp;
 524     _fjsp_v2r8       dummy_mask,cutoff_mask;
 525     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 526     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 527     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 528
 529     x                = xx[0];
 530     f                = ff[0];
 531
 532     nri              = nlist->nri;
 533     iinr             = nlist->iinr;
 534     jindex           = nlist->jindex;
 535     jjnr             = nlist->jjnr;
 536     shiftidx         = nlist->shift;
 537     gid              = nlist->gid;
 538     shiftvec         = fr->shift_vec[0];
 539     fshift           = fr->fshift[0];
 540     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 541     charge           = mdatoms->chargeA;
 542     nvdwtype         = fr->ntype;
 543     vdwparam         = fr->nbfp;
 544     vdwtype          = mdatoms->typeA;
 545
 546     /* Setup water-specific parameters */
 547     inr              = nlist->iinr[0];
 548     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 549     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 550     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 551     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 552
 553     /* Avoid stupid compiler warnings */
 554     jnrA = jnrB = 0;
 555     j_coord_offsetA = 0;
 556     j_coord_offsetB = 0;
 557
 558     outeriter        = 0;
 559     inneriter        = 0;
 560
 561     /* Start outer loop over neighborlists */
 562     for(iidx=0; iidx<nri; iidx++)
 563     {
 564         /* Load shift vector for this list */
 565         i_shift_offset   = DIM*shiftidx[iidx];
 566
 567         /* Load limits for loop over neighbors */
 568         j_index_start    = jindex[iidx];
 569         j_index_end      = jindex[iidx+1];
 570
 571         /* Get outer coordinate index */
 572         inr              = iinr[iidx];
 573         i_coord_offset   = DIM*inr;
 574
 575         /* Load i particle coords and add shift vector */
 576         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 577                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 578
 579         fix0             = _fjsp_setzero_v2r8();
 580         fiy0             = _fjsp_setzero_v2r8();
 581         fiz0             = _fjsp_setzero_v2r8();
 582         fix1             = _fjsp_setzero_v2r8();
 583         fiy1             = _fjsp_setzero_v2r8();
 584         fiz1             = _fjsp_setzero_v2r8();
 585         fix2             = _fjsp_setzero_v2r8();
 586         fiy2             = _fjsp_setzero_v2r8();
 587         fiz2             = _fjsp_setzero_v2r8();
 588
 589         /* Start inner kernel loop */
 590         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 591         {
 592
 593             /* Get j neighbor index, and coordinate index */
 594             jnrA             = jjnr[jidx];
 595             jnrB             = jjnr[jidx+1];
 596             j_coord_offsetA  = DIM*jnrA;
 597             j_coord_offsetB  = DIM*jnrB;
 598
 599             /* load j atom coordinates */
 600             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 601                                               &jx0,&jy0,&jz0);
 602
 603             /* Calculate displacement vector */
 604             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 605             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 606             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 607             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 608             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 609             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 610             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 611             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 612             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 613
 614             /* Calculate squared distance and things based on it */
 615             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 616             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 617             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 618
 619             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 620             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 621             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 622
 623             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 624             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 625             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 626
 627             /* Load parameters for j particles */
 628             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 629             vdwjidx0A        = 2*vdwtype[jnrA+0];
 630             vdwjidx0B        = 2*vdwtype[jnrB+0];
 631
 632             fjx0             = _fjsp_setzero_v2r8();
 633             fjy0             = _fjsp_setzero_v2r8();
 634             fjz0             = _fjsp_setzero_v2r8();
 635
 636             /**************************
 637              * CALCULATE INTERACTIONS *
 638              **************************/
 639
 640             /* Compute parameters for interactions between i and j atoms */
 641             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 642             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 643                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 644
 645             /* COULOMB ELECTROSTATICS */
 646             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 647             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 648
 649             /* LENNARD-JONES DISPERSION/REPULSION */
 650
 651             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 652             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 653
 654             fscal            = _fjsp_add_v2r8(felec,fvdw);
 655
 656             /* Update vectorial force */
 657             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 658             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 659             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 660
 661             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 662             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 663             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 664
 665             /**************************
 666              * CALCULATE INTERACTIONS *
 667              **************************/
 668
 669             /* Compute parameters for interactions between i and j atoms */
 670             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 671
 672             /* COULOMB ELECTROSTATICS */
 673             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 674             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 675
 676             fscal            = felec;
 677
 678             /* Update vectorial force */
 679             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 680             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 681             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 682
 683             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 684             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 685             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 686
 687             /**************************
 688              * CALCULATE INTERACTIONS *
 689              **************************/
 690
 691             /* Compute parameters for interactions between i and j atoms */
 692             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 693
 694             /* COULOMB ELECTROSTATICS */
 695             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 696             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 697
 698             fscal            = felec;
 699
 700             /* Update vectorial force */
 701             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 702             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 703             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 704
 705             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 706             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 707             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 708
 709             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 710
 711             /* Inner loop uses 100 flops */
 712         }
 713
 714         if(jidx<j_index_end)
 715         {
 716
 717             jnrA             = jjnr[jidx];
 718             j_coord_offsetA  = DIM*jnrA;
 719
 720             /* load j atom coordinates */
 721             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 722                                               &jx0,&jy0,&jz0);
 723
 724             /* Calculate displacement vector */
 725             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 726             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 727             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 728             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 729             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 730             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 731             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 732             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 733             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 734
 735             /* Calculate squared distance and things based on it */
 736             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 737             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 738             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 739
 740             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 741             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 742             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 743
 744             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 745             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 746             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 747
 748             /* Load parameters for j particles */
 749             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 750             vdwjidx0A        = 2*vdwtype[jnrA+0];
 751
 752             fjx0             = _fjsp_setzero_v2r8();
 753             fjy0             = _fjsp_setzero_v2r8();
 754             fjz0             = _fjsp_setzero_v2r8();
 755
 756             /**************************
 757              * CALCULATE INTERACTIONS *
 758              **************************/
 759
 760             /* Compute parameters for interactions between i and j atoms */
 761             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 762             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 763                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 764
 765             /* COULOMB ELECTROSTATICS */
 766             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 767             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 768
 769             /* LENNARD-JONES DISPERSION/REPULSION */
 770
 771             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 772             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 773
 774             fscal            = _fjsp_add_v2r8(felec,fvdw);
 775
 776             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 777
 778             /* Update vectorial force */
 779             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 780             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 781             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 782
 783             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 784             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 785             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 786
 787             /**************************
 788              * CALCULATE INTERACTIONS *
 789              **************************/
 790
 791             /* Compute parameters for interactions between i and j atoms */
 792             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 793
 794             /* COULOMB ELECTROSTATICS */
 795             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 796             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 797
 798             fscal            = felec;
 799
 800             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 801
 802             /* Update vectorial force */
 803             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 804             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 805             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 806
 807             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 808             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 809             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 810
 811             /**************************
 812              * CALCULATE INTERACTIONS *
 813              **************************/
 814
 815             /* Compute parameters for interactions between i and j atoms */
 816             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 817
 818             /* COULOMB ELECTROSTATICS */
 819             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 820             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 821
 822             fscal            = felec;
 823
 824             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 825
 826             /* Update vectorial force */
 827             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 828             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 829             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 830
 831             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 832             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 833             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 834
 835             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 836
 837             /* Inner loop uses 100 flops */
 838         }
 839
 840         /* End of innermost loop */
 841
 842         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 843                                               f+i_coord_offset,fshift+i_shift_offset);
 844
 845         /* Increment number of inner iterations */
 846         inneriter                  += j_index_end - j_index_start;
 847
 848         /* Outer loop uses 18 flops */
 849     }
 850
 851     /* Increment number of outer iterations */
 852     outeriter        += nri;
 853
 854     /* Update outer/inner flops */
 855
 856     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*100);
 857 }