src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwLJ_GeomW3P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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/gmxlib/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                      struct 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->ic->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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 355                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 356
 357             /* COULOMB ELECTROSTATICS */
 358             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 359             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 360
 361             /* LENNARD-JONES DISPERSION/REPULSION */
 362
 363             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 364             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 365             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 366             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 367             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 368
 369             /* Update potential sum for this i atom from the interaction with this j atom. */
 370             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 371             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 372             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 373             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 374
 375             fscal            = _fjsp_add_v2r8(felec,fvdw);
 376
 377             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 378
 379             /* Update vectorial force */
 380             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 381             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 382             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 383
 384             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 385             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 386             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 387
 388             /**************************
 389              * CALCULATE INTERACTIONS *
 390              **************************/
 391
 392             /* Compute parameters for interactions between i and j atoms */
 393             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 394
 395             /* COULOMB ELECTROSTATICS */
 396             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 397             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 398
 399             /* Update potential sum for this i atom from the interaction with this j atom. */
 400             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 401             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 402
 403             fscal            = felec;
 404
 405             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 406
 407             /* Update vectorial force */
 408             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 409             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 410             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 411
 412             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 413             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 414             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             /* Compute parameters for interactions between i and j atoms */
 421             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 422
 423             /* COULOMB ELECTROSTATICS */
 424             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 425             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 426
 427             /* Update potential sum for this i atom from the interaction with this j atom. */
 428             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 429             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 434
 435             /* Update vectorial force */
 436             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 437             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 438             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 439
 440             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 441             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 442             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 443
 444             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 445
 446             /* Inner loop uses 108 flops */
 447         }
 448
 449         /* End of innermost loop */
 450
 451         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 452                                               f+i_coord_offset,fshift+i_shift_offset);
 453
 454         ggid                        = gid[iidx];
 455         /* Update potential energies */
 456         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 457         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 458
 459         /* Increment number of inner iterations */
 460         inneriter                  += j_index_end - j_index_start;
 461
 462         /* Outer loop uses 20 flops */
 463     }
 464
 465     /* Increment number of outer iterations */
 466     outeriter        += nri;
 467
 468     /* Update outer/inner flops */
 469
 470     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*108);
 471 }
 472 /*
 473  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 474  * Electrostatics interaction: Coulomb
 475  * VdW interaction:            LennardJones
 476  * Geometry:                   Water3-Particle
 477  * Calculate force/pot:        Force
 478  */
 479 void
 480 nb_kernel_ElecCoul_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 481                     (t_nblist                    * gmx_restrict       nlist,
 482                      rvec                        * gmx_restrict          xx,
 483                      rvec                        * gmx_restrict          ff,
 484                      struct t_forcerec           * gmx_restrict          fr,
 485                      t_mdatoms                   * gmx_restrict     mdatoms,
 486                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 487                      t_nrnb                      * gmx_restrict        nrnb)
 488 {
 489     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 490      * just 0 for non-waters.
 491      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 492      * jnr indices corresponding to data put in the four positions in the SIMD register.
 493      */
 494     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 495     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 496     int              jnrA,jnrB;
 497     int              j_coord_offsetA,j_coord_offsetB;
 498     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 499     real             rcutoff_scalar;
 500     real             *shiftvec,*fshift,*x,*f;
 501     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 502     int              vdwioffset0;
 503     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 504     int              vdwioffset1;
 505     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 506     int              vdwioffset2;
 507     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 508     int              vdwjidx0A,vdwjidx0B;
 509     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 510     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 511     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 512     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 513     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 514     real             *charge;
 515     int              nvdwtype;
 516     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 517     int              *vdwtype;
 518     real             *vdwparam;
 519     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 520     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 521     _fjsp_v2r8       itab_tmp;
 522     _fjsp_v2r8       dummy_mask,cutoff_mask;
 523     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 524     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 525     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 526
 527     x                = xx[0];
 528     f                = ff[0];
 529
 530     nri              = nlist->nri;
 531     iinr             = nlist->iinr;
 532     jindex           = nlist->jindex;
 533     jjnr             = nlist->jjnr;
 534     shiftidx         = nlist->shift;
 535     gid              = nlist->gid;
 536     shiftvec         = fr->shift_vec[0];
 537     fshift           = fr->fshift[0];
 538     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 539     charge           = mdatoms->chargeA;
 540     nvdwtype         = fr->ntype;
 541     vdwparam         = fr->nbfp;
 542     vdwtype          = mdatoms->typeA;
 543
 544     /* Setup water-specific parameters */
 545     inr              = nlist->iinr[0];
 546     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 547     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 548     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 549     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 550
 551     /* Avoid stupid compiler warnings */
 552     jnrA = jnrB = 0;
 553     j_coord_offsetA = 0;
 554     j_coord_offsetB = 0;
 555
 556     outeriter        = 0;
 557     inneriter        = 0;
 558
 559     /* Start outer loop over neighborlists */
 560     for(iidx=0; iidx<nri; iidx++)
 561     {
 562         /* Load shift vector for this list */
 563         i_shift_offset   = DIM*shiftidx[iidx];
 564
 565         /* Load limits for loop over neighbors */
 566         j_index_start    = jindex[iidx];
 567         j_index_end      = jindex[iidx+1];
 568
 569         /* Get outer coordinate index */
 570         inr              = iinr[iidx];
 571         i_coord_offset   = DIM*inr;
 572
 573         /* Load i particle coords and add shift vector */
 574         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 575                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 576
 577         fix0             = _fjsp_setzero_v2r8();
 578         fiy0             = _fjsp_setzero_v2r8();
 579         fiz0             = _fjsp_setzero_v2r8();
 580         fix1             = _fjsp_setzero_v2r8();
 581         fiy1             = _fjsp_setzero_v2r8();
 582         fiz1             = _fjsp_setzero_v2r8();
 583         fix2             = _fjsp_setzero_v2r8();
 584         fiy2             = _fjsp_setzero_v2r8();
 585         fiz2             = _fjsp_setzero_v2r8();
 586
 587         /* Start inner kernel loop */
 588         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 589         {
 590
 591             /* Get j neighbor index, and coordinate index */
 592             jnrA             = jjnr[jidx];
 593             jnrB             = jjnr[jidx+1];
 594             j_coord_offsetA  = DIM*jnrA;
 595             j_coord_offsetB  = DIM*jnrB;
 596
 597             /* load j atom coordinates */
 598             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 599                                               &jx0,&jy0,&jz0);
 600
 601             /* Calculate displacement vector */
 602             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 603             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 604             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 605             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 606             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 607             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 608             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 609             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 610             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 611
 612             /* Calculate squared distance and things based on it */
 613             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 614             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 615             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 616
 617             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 618             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 619             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 620
 621             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 622             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 623             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 624
 625             /* Load parameters for j particles */
 626             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 627             vdwjidx0A        = 2*vdwtype[jnrA+0];
 628             vdwjidx0B        = 2*vdwtype[jnrB+0];
 629
 630             fjx0             = _fjsp_setzero_v2r8();
 631             fjy0             = _fjsp_setzero_v2r8();
 632             fjz0             = _fjsp_setzero_v2r8();
 633
 634             /**************************
 635              * CALCULATE INTERACTIONS *
 636              **************************/
 637
 638             /* Compute parameters for interactions between i and j atoms */
 639             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 640             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 641                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 642
 643             /* COULOMB ELECTROSTATICS */
 644             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 645             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 646
 647             /* LENNARD-JONES DISPERSION/REPULSION */
 648
 649             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 650             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 651
 652             fscal            = _fjsp_add_v2r8(felec,fvdw);
 653
 654             /* Update vectorial force */
 655             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 656             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 657             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 658
 659             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 660             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 661             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 662
 663             /**************************
 664              * CALCULATE INTERACTIONS *
 665              **************************/
 666
 667             /* Compute parameters for interactions between i and j atoms */
 668             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 669
 670             /* COULOMB ELECTROSTATICS */
 671             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 672             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 673
 674             fscal            = felec;
 675
 676             /* Update vectorial force */
 677             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 678             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 679             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 680
 681             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 682             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 683             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 684
 685             /**************************
 686              * CALCULATE INTERACTIONS *
 687              **************************/
 688
 689             /* Compute parameters for interactions between i and j atoms */
 690             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 691
 692             /* COULOMB ELECTROSTATICS */
 693             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 694             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 695
 696             fscal            = felec;
 697
 698             /* Update vectorial force */
 699             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 700             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 701             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 702
 703             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 704             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 705             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 706
 707             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 708
 709             /* Inner loop uses 100 flops */
 710         }
 711
 712         if(jidx<j_index_end)
 713         {
 714
 715             jnrA             = jjnr[jidx];
 716             j_coord_offsetA  = DIM*jnrA;
 717
 718             /* load j atom coordinates */
 719             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 720                                               &jx0,&jy0,&jz0);
 721
 722             /* Calculate displacement vector */
 723             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 724             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 725             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 726             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 727             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 728             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 729             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 730             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 731             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 732
 733             /* Calculate squared distance and things based on it */
 734             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 735             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 736             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 737
 738             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 739             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 740             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 741
 742             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 743             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 744             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 745
 746             /* Load parameters for j particles */
 747             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 748             vdwjidx0A        = 2*vdwtype[jnrA+0];
 749
 750             fjx0             = _fjsp_setzero_v2r8();
 751             fjy0             = _fjsp_setzero_v2r8();
 752             fjz0             = _fjsp_setzero_v2r8();
 753
 754             /**************************
 755              * CALCULATE INTERACTIONS *
 756              **************************/
 757
 758             /* Compute parameters for interactions between i and j atoms */
 759             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 760             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 761                                          vdwparam+vdwioffset0+vdwjidx0B,&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 }