src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwNone_GeomW4P1_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 "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            None
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwNone_GeomW4P1_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              vdwioffset1;
  80     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  81     int              vdwioffset2;
  82     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  83     int              vdwioffset3;
  84     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  88     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  89     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     _fjsp_v2r8       itab_tmp;
  93     _fjsp_v2r8       dummy_mask,cutoff_mask;
  94     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  95     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  96     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  97
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 110     charge           = mdatoms->chargeA;
 111
 112     /* Setup water-specific parameters */
 113     inr              = nlist->iinr[0];
 114     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 115     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 116     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 117
 118     /* Avoid stupid compiler warnings */
 119     jnrA = jnrB = 0;
 120     j_coord_offsetA = 0;
 121     j_coord_offsetB = 0;
 122
 123     outeriter        = 0;
 124     inneriter        = 0;
 125
 126     /* Start outer loop over neighborlists */
 127     for(iidx=0; iidx<nri; iidx++)
 128     {
 129         /* Load shift vector for this list */
 130         i_shift_offset   = DIM*shiftidx[iidx];
 131
 132         /* Load limits for loop over neighbors */
 133         j_index_start    = jindex[iidx];
 134         j_index_end      = jindex[iidx+1];
 135
 136         /* Get outer coordinate index */
 137         inr              = iinr[iidx];
 138         i_coord_offset   = DIM*inr;
 139
 140         /* Load i particle coords and add shift vector */
 141         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 142                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 143
 144         fix1             = _fjsp_setzero_v2r8();
 145         fiy1             = _fjsp_setzero_v2r8();
 146         fiz1             = _fjsp_setzero_v2r8();
 147         fix2             = _fjsp_setzero_v2r8();
 148         fiy2             = _fjsp_setzero_v2r8();
 149         fiz2             = _fjsp_setzero_v2r8();
 150         fix3             = _fjsp_setzero_v2r8();
 151         fiy3             = _fjsp_setzero_v2r8();
 152         fiz3             = _fjsp_setzero_v2r8();
 153
 154         /* Reset potential sums */
 155         velecsum         = _fjsp_setzero_v2r8();
 156
 157         /* Start inner kernel loop */
 158         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 159         {
 160
 161             /* Get j neighbor index, and coordinate index */
 162             jnrA             = jjnr[jidx];
 163             jnrB             = jjnr[jidx+1];
 164             j_coord_offsetA  = DIM*jnrA;
 165             j_coord_offsetB  = DIM*jnrB;
 166
 167             /* load j atom coordinates */
 168             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 169                                               &jx0,&jy0,&jz0);
 170
 171             /* Calculate displacement vector */
 172             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 173             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 174             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 175             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 176             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 177             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 178             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 179             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 180             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 181
 182             /* Calculate squared distance and things based on it */
 183             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 184             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 185             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 186
 187             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 188             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 189             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 190
 191             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 192             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 193             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 194
 195             /* Load parameters for j particles */
 196             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 197
 198             fjx0             = _fjsp_setzero_v2r8();
 199             fjy0             = _fjsp_setzero_v2r8();
 200             fjz0             = _fjsp_setzero_v2r8();
 201
 202             /**************************
 203              * CALCULATE INTERACTIONS *
 204              **************************/
 205
 206             /* Compute parameters for interactions between i and j atoms */
 207             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 208
 209             /* COULOMB ELECTROSTATICS */
 210             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 211             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 212
 213             /* Update potential sum for this i atom from the interaction with this j atom. */
 214             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 215
 216             fscal            = felec;
 217
 218             /* Update vectorial force */
 219             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 220             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 221             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 222
 223             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 224             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 225             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 226
 227             /**************************
 228              * CALCULATE INTERACTIONS *
 229              **************************/
 230
 231             /* Compute parameters for interactions between i and j atoms */
 232             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 233
 234             /* COULOMB ELECTROSTATICS */
 235             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 236             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 237
 238             /* Update potential sum for this i atom from the interaction with this j atom. */
 239             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 240
 241             fscal            = felec;
 242
 243             /* Update vectorial force */
 244             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 245             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 246             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 247
 248             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 249             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 250             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 251
 252             /**************************
 253              * CALCULATE INTERACTIONS *
 254              **************************/
 255
 256             /* Compute parameters for interactions between i and j atoms */
 257             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 258
 259             /* COULOMB ELECTROSTATICS */
 260             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 261             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 262
 263             /* Update potential sum for this i atom from the interaction with this j atom. */
 264             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 265
 266             fscal            = felec;
 267
 268             /* Update vectorial force */
 269             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 270             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 271             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 272
 273             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 274             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 275             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 276
 277             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 278
 279             /* Inner loop uses 96 flops */
 280         }
 281
 282         if(jidx<j_index_end)
 283         {
 284
 285             jnrA             = jjnr[jidx];
 286             j_coord_offsetA  = DIM*jnrA;
 287
 288             /* load j atom coordinates */
 289             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 290                                               &jx0,&jy0,&jz0);
 291
 292             /* Calculate displacement vector */
 293             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 294             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 295             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 296             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 297             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 298             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 299             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 300             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 301             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 302
 303             /* Calculate squared distance and things based on it */
 304             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 305             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 306             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 307
 308             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 309             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 310             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 311
 312             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 313             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 314             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 315
 316             /* Load parameters for j particles */
 317             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 318
 319             fjx0             = _fjsp_setzero_v2r8();
 320             fjy0             = _fjsp_setzero_v2r8();
 321             fjz0             = _fjsp_setzero_v2r8();
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 329
 330             /* COULOMB ELECTROSTATICS */
 331             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 332             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 333
 334             /* Update potential sum for this i atom from the interaction with this j atom. */
 335             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 336             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 337
 338             fscal            = felec;
 339
 340             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 341
 342             /* Update vectorial force */
 343             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 344             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 345             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 346
 347             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 348             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 349             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 350
 351             /**************************
 352              * CALCULATE INTERACTIONS *
 353              **************************/
 354
 355             /* Compute parameters for interactions between i and j atoms */
 356             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 357
 358             /* COULOMB ELECTROSTATICS */
 359             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 360             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 361
 362             /* Update potential sum for this i atom from the interaction with this j atom. */
 363             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 364             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 369
 370             /* Update vectorial force */
 371             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 372             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 373             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 374
 375             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 376             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 377             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 378
 379             /**************************
 380              * CALCULATE INTERACTIONS *
 381              **************************/
 382
 383             /* Compute parameters for interactions between i and j atoms */
 384             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 385
 386             /* COULOMB ELECTROSTATICS */
 387             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 388             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 392             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 393
 394             fscal            = felec;
 395
 396             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 397
 398             /* Update vectorial force */
 399             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 400             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 401             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 402
 403             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 404             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 405             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 406
 407             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 408
 409             /* Inner loop uses 96 flops */
 410         }
 411
 412         /* End of innermost loop */
 413
 414         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 415                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 416
 417         ggid                        = gid[iidx];
 418         /* Update potential energies */
 419         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 420
 421         /* Increment number of inner iterations */
 422         inneriter                  += j_index_end - j_index_start;
 423
 424         /* Outer loop uses 19 flops */
 425     }
 426
 427     /* Increment number of outer iterations */
 428     outeriter        += nri;
 429
 430     /* Update outer/inner flops */
 431
 432     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*96);
 433 }
 434 /*
 435  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 436  * Electrostatics interaction: Coulomb
 437  * VdW interaction:            None
 438  * Geometry:                   Water4-Particle
 439  * Calculate force/pot:        Force
 440  */
 441 void
 442 nb_kernel_ElecCoul_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 443                     (t_nblist                    * gmx_restrict       nlist,
 444                      rvec                        * gmx_restrict          xx,
 445                      rvec                        * gmx_restrict          ff,
 446                      t_forcerec                  * gmx_restrict          fr,
 447                      t_mdatoms                   * gmx_restrict     mdatoms,
 448                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 449                      t_nrnb                      * gmx_restrict        nrnb)
 450 {
 451     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 452      * just 0 for non-waters.
 453      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 454      * jnr indices corresponding to data put in the four positions in the SIMD register.
 455      */
 456     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 457     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 458     int              jnrA,jnrB;
 459     int              j_coord_offsetA,j_coord_offsetB;
 460     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 461     real             rcutoff_scalar;
 462     real             *shiftvec,*fshift,*x,*f;
 463     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 464     int              vdwioffset1;
 465     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 466     int              vdwioffset2;
 467     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 468     int              vdwioffset3;
 469     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 470     int              vdwjidx0A,vdwjidx0B;
 471     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 472     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 473     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 474     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 475     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 476     real             *charge;
 477     _fjsp_v2r8       itab_tmp;
 478     _fjsp_v2r8       dummy_mask,cutoff_mask;
 479     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 480     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 481     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 482
 483     x                = xx[0];
 484     f                = ff[0];
 485
 486     nri              = nlist->nri;
 487     iinr             = nlist->iinr;
 488     jindex           = nlist->jindex;
 489     jjnr             = nlist->jjnr;
 490     shiftidx         = nlist->shift;
 491     gid              = nlist->gid;
 492     shiftvec         = fr->shift_vec[0];
 493     fshift           = fr->fshift[0];
 494     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 495     charge           = mdatoms->chargeA;
 496
 497     /* Setup water-specific parameters */
 498     inr              = nlist->iinr[0];
 499     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 500     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 501     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 502
 503     /* Avoid stupid compiler warnings */
 504     jnrA = jnrB = 0;
 505     j_coord_offsetA = 0;
 506     j_coord_offsetB = 0;
 507
 508     outeriter        = 0;
 509     inneriter        = 0;
 510
 511     /* Start outer loop over neighborlists */
 512     for(iidx=0; iidx<nri; iidx++)
 513     {
 514         /* Load shift vector for this list */
 515         i_shift_offset   = DIM*shiftidx[iidx];
 516
 517         /* Load limits for loop over neighbors */
 518         j_index_start    = jindex[iidx];
 519         j_index_end      = jindex[iidx+1];
 520
 521         /* Get outer coordinate index */
 522         inr              = iinr[iidx];
 523         i_coord_offset   = DIM*inr;
 524
 525         /* Load i particle coords and add shift vector */
 526         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 527                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 528
 529         fix1             = _fjsp_setzero_v2r8();
 530         fiy1             = _fjsp_setzero_v2r8();
 531         fiz1             = _fjsp_setzero_v2r8();
 532         fix2             = _fjsp_setzero_v2r8();
 533         fiy2             = _fjsp_setzero_v2r8();
 534         fiz2             = _fjsp_setzero_v2r8();
 535         fix3             = _fjsp_setzero_v2r8();
 536         fiy3             = _fjsp_setzero_v2r8();
 537         fiz3             = _fjsp_setzero_v2r8();
 538
 539         /* Start inner kernel loop */
 540         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 541         {
 542
 543             /* Get j neighbor index, and coordinate index */
 544             jnrA             = jjnr[jidx];
 545             jnrB             = jjnr[jidx+1];
 546             j_coord_offsetA  = DIM*jnrA;
 547             j_coord_offsetB  = DIM*jnrB;
 548
 549             /* load j atom coordinates */
 550             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 551                                               &jx0,&jy0,&jz0);
 552
 553             /* Calculate displacement vector */
 554             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 555             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 556             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 557             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 558             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 559             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 560             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 561             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 562             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 563
 564             /* Calculate squared distance and things based on it */
 565             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 566             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 567             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 568
 569             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 570             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 571             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 572
 573             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 574             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 575             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 576
 577             /* Load parameters for j particles */
 578             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 579
 580             fjx0             = _fjsp_setzero_v2r8();
 581             fjy0             = _fjsp_setzero_v2r8();
 582             fjz0             = _fjsp_setzero_v2r8();
 583
 584             /**************************
 585              * CALCULATE INTERACTIONS *
 586              **************************/
 587
 588             /* Compute parameters for interactions between i and j atoms */
 589             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 590
 591             /* COULOMB ELECTROSTATICS */
 592             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 593             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 594
 595             fscal            = felec;
 596
 597             /* Update vectorial force */
 598             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 599             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 600             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 601
 602             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 603             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 604             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 605
 606             /**************************
 607              * CALCULATE INTERACTIONS *
 608              **************************/
 609
 610             /* Compute parameters for interactions between i and j atoms */
 611             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 612
 613             /* COULOMB ELECTROSTATICS */
 614             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 615             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 616
 617             fscal            = felec;
 618
 619             /* Update vectorial force */
 620             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 621             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 622             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 623
 624             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 625             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 626             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 627
 628             /**************************
 629              * CALCULATE INTERACTIONS *
 630              **************************/
 631
 632             /* Compute parameters for interactions between i and j atoms */
 633             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 634
 635             /* COULOMB ELECTROSTATICS */
 636             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 637             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 638
 639             fscal            = felec;
 640
 641             /* Update vectorial force */
 642             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 643             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 644             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 645
 646             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 647             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 648             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 649
 650             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 651
 652             /* Inner loop uses 93 flops */
 653         }
 654
 655         if(jidx<j_index_end)
 656         {
 657
 658             jnrA             = jjnr[jidx];
 659             j_coord_offsetA  = DIM*jnrA;
 660
 661             /* load j atom coordinates */
 662             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 663                                               &jx0,&jy0,&jz0);
 664
 665             /* Calculate displacement vector */
 666             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 667             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 668             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 669             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 670             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 671             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 672             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 673             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 674             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 675
 676             /* Calculate squared distance and things based on it */
 677             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 678             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 679             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 680
 681             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 682             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 683             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 684
 685             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 686             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 687             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 688
 689             /* Load parameters for j particles */
 690             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 691
 692             fjx0             = _fjsp_setzero_v2r8();
 693             fjy0             = _fjsp_setzero_v2r8();
 694             fjz0             = _fjsp_setzero_v2r8();
 695
 696             /**************************
 697              * CALCULATE INTERACTIONS *
 698              **************************/
 699
 700             /* Compute parameters for interactions between i and j atoms */
 701             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 702
 703             /* COULOMB ELECTROSTATICS */
 704             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 705             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 706
 707             fscal            = felec;
 708
 709             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 710
 711             /* Update vectorial force */
 712             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 713             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 714             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 715
 716             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 717             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 718             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 719
 720             /**************************
 721              * CALCULATE INTERACTIONS *
 722              **************************/
 723
 724             /* Compute parameters for interactions between i and j atoms */
 725             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 726
 727             /* COULOMB ELECTROSTATICS */
 728             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 729             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 730
 731             fscal            = felec;
 732
 733             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 734
 735             /* Update vectorial force */
 736             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 737             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 738             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 739
 740             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 741             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 742             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 743
 744             /**************************
 745              * CALCULATE INTERACTIONS *
 746              **************************/
 747
 748             /* Compute parameters for interactions between i and j atoms */
 749             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 750
 751             /* COULOMB ELECTROSTATICS */
 752             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 753             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 754
 755             fscal            = felec;
 756
 757             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 758
 759             /* Update vectorial force */
 760             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 761             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 762             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 763
 764             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 765             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 766             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 767
 768             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 769
 770             /* Inner loop uses 93 flops */
 771         }
 772
 773         /* End of innermost loop */
 774
 775         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 776                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 777
 778         /* Increment number of inner iterations */
 779         inneriter                  += j_index_end - j_index_start;
 780
 781         /* Outer loop uses 18 flops */
 782     }
 783
 784     /* Increment number of outer iterations */
 785     outeriter        += nri;
 786
 787     /* Update outer/inner flops */
 788
 789     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*93);
 790 }