src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwNone_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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Coulomb
  54  * VdW interaction:            None
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCoul_VdwNone_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     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113
 114     /* Setup water-specific parameters */
 115     inr              = nlist->iinr[0];
 116     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 117     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 118     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 119
 120     /* Avoid stupid compiler warnings */
 121     jnrA = jnrB = 0;
 122     j_coord_offsetA = 0;
 123     j_coord_offsetB = 0;
 124
 125     outeriter        = 0;
 126     inneriter        = 0;
 127
 128     /* Start outer loop over neighborlists */
 129     for(iidx=0; iidx<nri; iidx++)
 130     {
 131         /* Load shift vector for this list */
 132         i_shift_offset   = DIM*shiftidx[iidx];
 133
 134         /* Load limits for loop over neighbors */
 135         j_index_start    = jindex[iidx];
 136         j_index_end      = jindex[iidx+1];
 137
 138         /* Get outer coordinate index */
 139         inr              = iinr[iidx];
 140         i_coord_offset   = DIM*inr;
 141
 142         /* Load i particle coords and add shift vector */
 143         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 144                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 145
 146         fix0             = _fjsp_setzero_v2r8();
 147         fiy0             = _fjsp_setzero_v2r8();
 148         fiz0             = _fjsp_setzero_v2r8();
 149         fix1             = _fjsp_setzero_v2r8();
 150         fiy1             = _fjsp_setzero_v2r8();
 151         fiz1             = _fjsp_setzero_v2r8();
 152         fix2             = _fjsp_setzero_v2r8();
 153         fiy2             = _fjsp_setzero_v2r8();
 154         fiz2             = _fjsp_setzero_v2r8();
 155
 156         /* Reset potential sums */
 157         velecsum         = _fjsp_setzero_v2r8();
 158
 159         /* Start inner kernel loop */
 160         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 161         {
 162
 163             /* Get j neighbor index, and coordinate index */
 164             jnrA             = jjnr[jidx];
 165             jnrB             = jjnr[jidx+1];
 166             j_coord_offsetA  = DIM*jnrA;
 167             j_coord_offsetB  = DIM*jnrB;
 168
 169             /* load j atom coordinates */
 170             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 171                                               &jx0,&jy0,&jz0);
 172
 173             /* Calculate displacement vector */
 174             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 175             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 176             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 177             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 178             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 179             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 180             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 181             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 182             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 183
 184             /* Calculate squared distance and things based on it */
 185             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 186             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 187             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 188
 189             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 190             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 191             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 192
 193             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 194             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 195             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 196
 197             /* Load parameters for j particles */
 198             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 199
 200             fjx0             = _fjsp_setzero_v2r8();
 201             fjy0             = _fjsp_setzero_v2r8();
 202             fjz0             = _fjsp_setzero_v2r8();
 203
 204             /**************************
 205              * CALCULATE INTERACTIONS *
 206              **************************/
 207
 208             /* Compute parameters for interactions between i and j atoms */
 209             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 210
 211             /* COULOMB ELECTROSTATICS */
 212             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 213             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 214
 215             /* Update potential sum for this i atom from the interaction with this j atom. */
 216             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 217
 218             fscal            = felec;
 219
 220             /* Update vectorial force */
 221             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 222             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 223             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 224
 225             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 226             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 227             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 228
 229             /**************************
 230              * CALCULATE INTERACTIONS *
 231              **************************/
 232
 233             /* Compute parameters for interactions between i and j atoms */
 234             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 235
 236             /* COULOMB ELECTROSTATICS */
 237             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 238             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 239
 240             /* Update potential sum for this i atom from the interaction with this j atom. */
 241             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 242
 243             fscal            = felec;
 244
 245             /* Update vectorial force */
 246             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 247             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 248             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 249
 250             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 251             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 252             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 253
 254             /**************************
 255              * CALCULATE INTERACTIONS *
 256              **************************/
 257
 258             /* Compute parameters for interactions between i and j atoms */
 259             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 260
 261             /* COULOMB ELECTROSTATICS */
 262             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 263             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 264
 265             /* Update potential sum for this i atom from the interaction with this j atom. */
 266             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 267
 268             fscal            = felec;
 269
 270             /* Update vectorial force */
 271             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 272             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 273             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 274
 275             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 276             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 277             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 278
 279             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 280
 281             /* Inner loop uses 96 flops */
 282         }
 283
 284         if(jidx<j_index_end)
 285         {
 286
 287             jnrA             = jjnr[jidx];
 288             j_coord_offsetA  = DIM*jnrA;
 289
 290             /* load j atom coordinates */
 291             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 292                                               &jx0,&jy0,&jz0);
 293
 294             /* Calculate displacement vector */
 295             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 296             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 297             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 298             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 299             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 300             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 301             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 302             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 303             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 304
 305             /* Calculate squared distance and things based on it */
 306             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 307             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 308             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 309
 310             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 311             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 312             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 313
 314             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 315             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 316             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 317
 318             /* Load parameters for j particles */
 319             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 320
 321             fjx0             = _fjsp_setzero_v2r8();
 322             fjy0             = _fjsp_setzero_v2r8();
 323             fjz0             = _fjsp_setzero_v2r8();
 324
 325             /**************************
 326              * CALCULATE INTERACTIONS *
 327              **************************/
 328
 329             /* Compute parameters for interactions between i and j atoms */
 330             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 331
 332             /* COULOMB ELECTROSTATICS */
 333             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 334             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 338             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 339
 340             fscal            = felec;
 341
 342             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 343
 344             /* Update vectorial force */
 345             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 346             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 347             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 348
 349             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 350             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 351             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 352
 353             /**************************
 354              * CALCULATE INTERACTIONS *
 355              **************************/
 356
 357             /* Compute parameters for interactions between i and j atoms */
 358             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 359
 360             /* COULOMB ELECTROSTATICS */
 361             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 362             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 363
 364             /* Update potential sum for this i atom from the interaction with this j atom. */
 365             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 366             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 371
 372             /* Update vectorial force */
 373             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 374             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 375             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 376
 377             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 378             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 379             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 380
 381             /**************************
 382              * CALCULATE INTERACTIONS *
 383              **************************/
 384
 385             /* Compute parameters for interactions between i and j atoms */
 386             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 387
 388             /* COULOMB ELECTROSTATICS */
 389             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 390             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 391
 392             /* Update potential sum for this i atom from the interaction with this j atom. */
 393             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 394             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 395
 396             fscal            = felec;
 397
 398             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 399
 400             /* Update vectorial force */
 401             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 402             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 403             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 404
 405             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 406             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 407             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 408
 409             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 410
 411             /* Inner loop uses 96 flops */
 412         }
 413
 414         /* End of innermost loop */
 415
 416         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 417                                               f+i_coord_offset,fshift+i_shift_offset);
 418
 419         ggid                        = gid[iidx];
 420         /* Update potential energies */
 421         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 422
 423         /* Increment number of inner iterations */
 424         inneriter                  += j_index_end - j_index_start;
 425
 426         /* Outer loop uses 19 flops */
 427     }
 428
 429     /* Increment number of outer iterations */
 430     outeriter        += nri;
 431
 432     /* Update outer/inner flops */
 433
 434     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*96);
 435 }
 436 /*
 437  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 438  * Electrostatics interaction: Coulomb
 439  * VdW interaction:            None
 440  * Geometry:                   Water3-Particle
 441  * Calculate force/pot:        Force
 442  */
 443 void
 444 nb_kernel_ElecCoul_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 445                     (t_nblist                    * gmx_restrict       nlist,
 446                      rvec                        * gmx_restrict          xx,
 447                      rvec                        * gmx_restrict          ff,
 448                      t_forcerec                  * gmx_restrict          fr,
 449                      t_mdatoms                   * gmx_restrict     mdatoms,
 450                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 451                      t_nrnb                      * gmx_restrict        nrnb)
 452 {
 453     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 454      * just 0 for non-waters.
 455      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 456      * jnr indices corresponding to data put in the four positions in the SIMD register.
 457      */
 458     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 459     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 460     int              jnrA,jnrB;
 461     int              j_coord_offsetA,j_coord_offsetB;
 462     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 463     real             rcutoff_scalar;
 464     real             *shiftvec,*fshift,*x,*f;
 465     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 466     int              vdwioffset0;
 467     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 468     int              vdwioffset1;
 469     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 470     int              vdwioffset2;
 471     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 472     int              vdwjidx0A,vdwjidx0B;
 473     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 474     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 475     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 476     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 477     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 478     real             *charge;
 479     _fjsp_v2r8       itab_tmp;
 480     _fjsp_v2r8       dummy_mask,cutoff_mask;
 481     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 482     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 483     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 484
 485     x                = xx[0];
 486     f                = ff[0];
 487
 488     nri              = nlist->nri;
 489     iinr             = nlist->iinr;
 490     jindex           = nlist->jindex;
 491     jjnr             = nlist->jjnr;
 492     shiftidx         = nlist->shift;
 493     gid              = nlist->gid;
 494     shiftvec         = fr->shift_vec[0];
 495     fshift           = fr->fshift[0];
 496     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 497     charge           = mdatoms->chargeA;
 498
 499     /* Setup water-specific parameters */
 500     inr              = nlist->iinr[0];
 501     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 502     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 503     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 504
 505     /* Avoid stupid compiler warnings */
 506     jnrA = jnrB = 0;
 507     j_coord_offsetA = 0;
 508     j_coord_offsetB = 0;
 509
 510     outeriter        = 0;
 511     inneriter        = 0;
 512
 513     /* Start outer loop over neighborlists */
 514     for(iidx=0; iidx<nri; iidx++)
 515     {
 516         /* Load shift vector for this list */
 517         i_shift_offset   = DIM*shiftidx[iidx];
 518
 519         /* Load limits for loop over neighbors */
 520         j_index_start    = jindex[iidx];
 521         j_index_end      = jindex[iidx+1];
 522
 523         /* Get outer coordinate index */
 524         inr              = iinr[iidx];
 525         i_coord_offset   = DIM*inr;
 526
 527         /* Load i particle coords and add shift vector */
 528         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 529                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 530
 531         fix0             = _fjsp_setzero_v2r8();
 532         fiy0             = _fjsp_setzero_v2r8();
 533         fiz0             = _fjsp_setzero_v2r8();
 534         fix1             = _fjsp_setzero_v2r8();
 535         fiy1             = _fjsp_setzero_v2r8();
 536         fiz1             = _fjsp_setzero_v2r8();
 537         fix2             = _fjsp_setzero_v2r8();
 538         fiy2             = _fjsp_setzero_v2r8();
 539         fiz2             = _fjsp_setzero_v2r8();
 540
 541         /* Start inner kernel loop */
 542         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 543         {
 544
 545             /* Get j neighbor index, and coordinate index */
 546             jnrA             = jjnr[jidx];
 547             jnrB             = jjnr[jidx+1];
 548             j_coord_offsetA  = DIM*jnrA;
 549             j_coord_offsetB  = DIM*jnrB;
 550
 551             /* load j atom coordinates */
 552             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 553                                               &jx0,&jy0,&jz0);
 554
 555             /* Calculate displacement vector */
 556             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 557             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 558             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 559             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 560             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 561             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 562             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 563             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 564             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 565
 566             /* Calculate squared distance and things based on it */
 567             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 568             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 569             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 570
 571             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 572             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 573             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 574
 575             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 576             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 577             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 578
 579             /* Load parameters for j particles */
 580             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 581
 582             fjx0             = _fjsp_setzero_v2r8();
 583             fjy0             = _fjsp_setzero_v2r8();
 584             fjz0             = _fjsp_setzero_v2r8();
 585
 586             /**************************
 587              * CALCULATE INTERACTIONS *
 588              **************************/
 589
 590             /* Compute parameters for interactions between i and j atoms */
 591             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 592
 593             /* COULOMB ELECTROSTATICS */
 594             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 595             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 596
 597             fscal            = felec;
 598
 599             /* Update vectorial force */
 600             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 601             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 602             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 603
 604             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 605             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 606             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 607
 608             /**************************
 609              * CALCULATE INTERACTIONS *
 610              **************************/
 611
 612             /* Compute parameters for interactions between i and j atoms */
 613             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 614
 615             /* COULOMB ELECTROSTATICS */
 616             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 617             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 618
 619             fscal            = felec;
 620
 621             /* Update vectorial force */
 622             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 623             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 624             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 625
 626             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 627             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 628             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 629
 630             /**************************
 631              * CALCULATE INTERACTIONS *
 632              **************************/
 633
 634             /* Compute parameters for interactions between i and j atoms */
 635             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 636
 637             /* COULOMB ELECTROSTATICS */
 638             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 639             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 640
 641             fscal            = felec;
 642
 643             /* Update vectorial force */
 644             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 645             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 646             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 647
 648             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 649             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 650             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 651
 652             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 653
 654             /* Inner loop uses 93 flops */
 655         }
 656
 657         if(jidx<j_index_end)
 658         {
 659
 660             jnrA             = jjnr[jidx];
 661             j_coord_offsetA  = DIM*jnrA;
 662
 663             /* load j atom coordinates */
 664             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 665                                               &jx0,&jy0,&jz0);
 666
 667             /* Calculate displacement vector */
 668             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 669             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 670             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 671             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 672             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 673             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 674             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 675             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 676             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 677
 678             /* Calculate squared distance and things based on it */
 679             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 680             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 681             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 682
 683             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 684             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 685             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 686
 687             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 688             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 689             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 690
 691             /* Load parameters for j particles */
 692             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 693
 694             fjx0             = _fjsp_setzero_v2r8();
 695             fjy0             = _fjsp_setzero_v2r8();
 696             fjz0             = _fjsp_setzero_v2r8();
 697
 698             /**************************
 699              * CALCULATE INTERACTIONS *
 700              **************************/
 701
 702             /* Compute parameters for interactions between i and j atoms */
 703             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 704
 705             /* COULOMB ELECTROSTATICS */
 706             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 707             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 708
 709             fscal            = felec;
 710
 711             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 712
 713             /* Update vectorial force */
 714             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 715             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 716             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 717
 718             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 719             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 720             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 721
 722             /**************************
 723              * CALCULATE INTERACTIONS *
 724              **************************/
 725
 726             /* Compute parameters for interactions between i and j atoms */
 727             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 728
 729             /* COULOMB ELECTROSTATICS */
 730             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 731             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 732
 733             fscal            = felec;
 734
 735             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 736
 737             /* Update vectorial force */
 738             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 739             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 740             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 741
 742             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 743             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 744             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 745
 746             /**************************
 747              * CALCULATE INTERACTIONS *
 748              **************************/
 749
 750             /* Compute parameters for interactions between i and j atoms */
 751             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 752
 753             /* COULOMB ELECTROSTATICS */
 754             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 755             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 756
 757             fscal            = felec;
 758
 759             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 760
 761             /* Update vectorial force */
 762             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 763             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 764             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 765
 766             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 767             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 768             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 769
 770             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 771
 772             /* Inner loop uses 93 flops */
 773         }
 774
 775         /* End of innermost loop */
 776
 777         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 778                                               f+i_coord_offset,fshift+i_shift_offset);
 779
 780         /* Increment number of inner iterations */
 781         inneriter                  += j_index_end - j_index_start;
 782
 783         /* Outer loop uses 18 flops */
 784     }
 785
 786     /* Increment number of outer iterations */
 787     outeriter        += nri;
 788
 789     /* Update outer/inner flops */
 790
 791     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*93);
 792 }