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