src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_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 #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/math/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            None
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEw_VdwNone_GeomW4P1_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              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              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  90     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  91     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  95     real             *ewtab;
  96     _fjsp_v2r8       itab_tmp;
  97     _fjsp_v2r8       dummy_mask,cutoff_mask;
  98     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  99     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 100     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 101
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 114     charge           = mdatoms->chargeA;
 115
 116     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 117     ewtab            = fr->ic->tabq_coul_FDV0;
 118     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 119     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 120
 121     /* Setup water-specific parameters */
 122     inr              = nlist->iinr[0];
 123     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 124     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 125     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 126
 127     /* Avoid stupid compiler warnings */
 128     jnrA = jnrB = 0;
 129     j_coord_offsetA = 0;
 130     j_coord_offsetB = 0;
 131
 132     outeriter        = 0;
 133     inneriter        = 0;
 134
 135     /* Start outer loop over neighborlists */
 136     for(iidx=0; iidx<nri; iidx++)
 137     {
 138         /* Load shift vector for this list */
 139         i_shift_offset   = DIM*shiftidx[iidx];
 140
 141         /* Load limits for loop over neighbors */
 142         j_index_start    = jindex[iidx];
 143         j_index_end      = jindex[iidx+1];
 144
 145         /* Get outer coordinate index */
 146         inr              = iinr[iidx];
 147         i_coord_offset   = DIM*inr;
 148
 149         /* Load i particle coords and add shift vector */
 150         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 151                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 152
 153         fix1             = _fjsp_setzero_v2r8();
 154         fiy1             = _fjsp_setzero_v2r8();
 155         fiz1             = _fjsp_setzero_v2r8();
 156         fix2             = _fjsp_setzero_v2r8();
 157         fiy2             = _fjsp_setzero_v2r8();
 158         fiz2             = _fjsp_setzero_v2r8();
 159         fix3             = _fjsp_setzero_v2r8();
 160         fiy3             = _fjsp_setzero_v2r8();
 161         fiz3             = _fjsp_setzero_v2r8();
 162
 163         /* Reset potential sums */
 164         velecsum         = _fjsp_setzero_v2r8();
 165
 166         /* Start inner kernel loop */
 167         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 168         {
 169
 170             /* Get j neighbor index, and coordinate index */
 171             jnrA             = jjnr[jidx];
 172             jnrB             = jjnr[jidx+1];
 173             j_coord_offsetA  = DIM*jnrA;
 174             j_coord_offsetB  = DIM*jnrB;
 175
 176             /* load j atom coordinates */
 177             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 178                                               &jx0,&jy0,&jz0);
 179
 180             /* Calculate displacement vector */
 181             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 182             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 183             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 184             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 185             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 186             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 187             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 188             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 189             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 190
 191             /* Calculate squared distance and things based on it */
 192             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 193             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 194             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 195
 196             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 197             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 198             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 199
 200             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 201             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 202             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 203
 204             /* Load parameters for j particles */
 205             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 206
 207             fjx0             = _fjsp_setzero_v2r8();
 208             fjy0             = _fjsp_setzero_v2r8();
 209             fjz0             = _fjsp_setzero_v2r8();
 210
 211             /**************************
 212              * CALCULATE INTERACTIONS *
 213              **************************/
 214
 215             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 216
 217             /* Compute parameters for interactions between i and j atoms */
 218             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 219
 220             /* EWALD ELECTROSTATICS */
 221
 222             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 223             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 224             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 225             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 226             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 227
 228             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 229             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 230             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 231             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 232             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 233             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 234             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 235             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 236             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 237             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 238
 239             /* Update potential sum for this i atom from the interaction with this j atom. */
 240             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 241
 242             fscal            = felec;
 243
 244             /* Update vectorial force */
 245             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 246             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 247             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 248
 249             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 250             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 251             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 252
 253             /**************************
 254              * CALCULATE INTERACTIONS *
 255              **************************/
 256
 257             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 258
 259             /* Compute parameters for interactions between i and j atoms */
 260             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 261
 262             /* EWALD ELECTROSTATICS */
 263
 264             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 265             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 266             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 267             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 268             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 269
 270             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 271             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 272             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 273             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 274             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 275             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 276             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 277             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 278             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 279             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 280
 281             /* Update potential sum for this i atom from the interaction with this j atom. */
 282             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 283
 284             fscal            = felec;
 285
 286             /* Update vectorial force */
 287             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 288             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 289             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 300
 301             /* Compute parameters for interactions between i and j atoms */
 302             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 303
 304             /* EWALD ELECTROSTATICS */
 305
 306             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 307             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 308             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 309             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 310             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 311
 312             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 313             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 314             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 315             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 316             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 317             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 318             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 319             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 320             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 321             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 322
 323             /* Update potential sum for this i atom from the interaction with this j atom. */
 324             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 325
 326             fscal            = felec;
 327
 328             /* Update vectorial force */
 329             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 330             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 331             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 332
 333             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 334             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 335             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 336
 337             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 338
 339             /* Inner loop uses 135 flops */
 340         }
 341
 342         if(jidx<j_index_end)
 343         {
 344
 345             jnrA             = jjnr[jidx];
 346             j_coord_offsetA  = DIM*jnrA;
 347
 348             /* load j atom coordinates */
 349             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 350                                               &jx0,&jy0,&jz0);
 351
 352             /* Calculate displacement vector */
 353             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 354             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 355             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 356             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 357             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 358             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 359             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 360             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 361             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 362
 363             /* Calculate squared distance and things based on it */
 364             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 365             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 366             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 367
 368             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 369             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 370             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 371
 372             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 373             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 374             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 375
 376             /* Load parameters for j particles */
 377             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 378
 379             fjx0             = _fjsp_setzero_v2r8();
 380             fjy0             = _fjsp_setzero_v2r8();
 381             fjz0             = _fjsp_setzero_v2r8();
 382
 383             /**************************
 384              * CALCULATE INTERACTIONS *
 385              **************************/
 386
 387             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 388
 389             /* Compute parameters for interactions between i and j atoms */
 390             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 391
 392             /* EWALD ELECTROSTATICS */
 393
 394             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 395             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 396             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 397             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 398             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 399
 400             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 401             ewtabD           = _fjsp_setzero_v2r8();
 402             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 403             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 404             ewtabFn          = _fjsp_setzero_v2r8();
 405             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 406             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 407             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 408             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 409             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 410
 411             /* Update potential sum for this i atom from the interaction with this j atom. */
 412             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 413             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 414
 415             fscal            = felec;
 416
 417             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 418
 419             /* Update vectorial force */
 420             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 421             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 422             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 423
 424             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 425             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 426             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 433
 434             /* Compute parameters for interactions between i and j atoms */
 435             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 436
 437             /* EWALD ELECTROSTATICS */
 438
 439             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 440             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 441             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 442             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 443             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 444
 445             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 446             ewtabD           = _fjsp_setzero_v2r8();
 447             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 448             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 449             ewtabFn          = _fjsp_setzero_v2r8();
 450             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 451             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 452             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 453             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 454             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 458             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 459
 460             fscal            = felec;
 461
 462             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 463
 464             /* Update vectorial force */
 465             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 466             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 467             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 468
 469             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 470             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 471             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 472
 473             /**************************
 474              * CALCULATE INTERACTIONS *
 475              **************************/
 476
 477             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 478
 479             /* Compute parameters for interactions between i and j atoms */
 480             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 481
 482             /* EWALD ELECTROSTATICS */
 483
 484             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 485             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 486             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 487             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 488             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 489
 490             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 491             ewtabD           = _fjsp_setzero_v2r8();
 492             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 493             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 494             ewtabFn          = _fjsp_setzero_v2r8();
 495             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 496             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 497             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 498             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 499             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 500
 501             /* Update potential sum for this i atom from the interaction with this j atom. */
 502             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 503             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 504
 505             fscal            = felec;
 506
 507             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 508
 509             /* Update vectorial force */
 510             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 511             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 512             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 513
 514             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 515             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 516             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 517
 518             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 519
 520             /* Inner loop uses 135 flops */
 521         }
 522
 523         /* End of innermost loop */
 524
 525         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 526                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 527
 528         ggid                        = gid[iidx];
 529         /* Update potential energies */
 530         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 531
 532         /* Increment number of inner iterations */
 533         inneriter                  += j_index_end - j_index_start;
 534
 535         /* Outer loop uses 19 flops */
 536     }
 537
 538     /* Increment number of outer iterations */
 539     outeriter        += nri;
 540
 541     /* Update outer/inner flops */
 542
 543     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*135);
 544 }
 545 /*
 546  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 547  * Electrostatics interaction: Ewald
 548  * VdW interaction:            None
 549  * Geometry:                   Water4-Particle
 550  * Calculate force/pot:        Force
 551  */
 552 void
 553 nb_kernel_ElecEw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 554                     (t_nblist                    * gmx_restrict       nlist,
 555                      rvec                        * gmx_restrict          xx,
 556                      rvec                        * gmx_restrict          ff,
 557                      t_forcerec                  * gmx_restrict          fr,
 558                      t_mdatoms                   * gmx_restrict     mdatoms,
 559                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 560                      t_nrnb                      * gmx_restrict        nrnb)
 561 {
 562     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 563      * just 0 for non-waters.
 564      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 565      * jnr indices corresponding to data put in the four positions in the SIMD register.
 566      */
 567     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 568     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 569     int              jnrA,jnrB;
 570     int              j_coord_offsetA,j_coord_offsetB;
 571     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 572     real             rcutoff_scalar;
 573     real             *shiftvec,*fshift,*x,*f;
 574     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 575     int              vdwioffset1;
 576     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 577     int              vdwioffset2;
 578     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 579     int              vdwioffset3;
 580     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 581     int              vdwjidx0A,vdwjidx0B;
 582     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 583     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 584     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 585     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 586     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 587     real             *charge;
 588     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 589     real             *ewtab;
 590     _fjsp_v2r8       itab_tmp;
 591     _fjsp_v2r8       dummy_mask,cutoff_mask;
 592     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 593     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 594     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 595
 596     x                = xx[0];
 597     f                = ff[0];
 598
 599     nri              = nlist->nri;
 600     iinr             = nlist->iinr;
 601     jindex           = nlist->jindex;
 602     jjnr             = nlist->jjnr;
 603     shiftidx         = nlist->shift;
 604     gid              = nlist->gid;
 605     shiftvec         = fr->shift_vec[0];
 606     fshift           = fr->fshift[0];
 607     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 608     charge           = mdatoms->chargeA;
 609
 610     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 611     ewtab            = fr->ic->tabq_coul_F;
 612     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 613     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 614
 615     /* Setup water-specific parameters */
 616     inr              = nlist->iinr[0];
 617     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 618     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 619     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 620
 621     /* Avoid stupid compiler warnings */
 622     jnrA = jnrB = 0;
 623     j_coord_offsetA = 0;
 624     j_coord_offsetB = 0;
 625
 626     outeriter        = 0;
 627     inneriter        = 0;
 628
 629     /* Start outer loop over neighborlists */
 630     for(iidx=0; iidx<nri; iidx++)
 631     {
 632         /* Load shift vector for this list */
 633         i_shift_offset   = DIM*shiftidx[iidx];
 634
 635         /* Load limits for loop over neighbors */
 636         j_index_start    = jindex[iidx];
 637         j_index_end      = jindex[iidx+1];
 638
 639         /* Get outer coordinate index */
 640         inr              = iinr[iidx];
 641         i_coord_offset   = DIM*inr;
 642
 643         /* Load i particle coords and add shift vector */
 644         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 645                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 646
 647         fix1             = _fjsp_setzero_v2r8();
 648         fiy1             = _fjsp_setzero_v2r8();
 649         fiz1             = _fjsp_setzero_v2r8();
 650         fix2             = _fjsp_setzero_v2r8();
 651         fiy2             = _fjsp_setzero_v2r8();
 652         fiz2             = _fjsp_setzero_v2r8();
 653         fix3             = _fjsp_setzero_v2r8();
 654         fiy3             = _fjsp_setzero_v2r8();
 655         fiz3             = _fjsp_setzero_v2r8();
 656
 657         /* Start inner kernel loop */
 658         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 659         {
 660
 661             /* Get j neighbor index, and coordinate index */
 662             jnrA             = jjnr[jidx];
 663             jnrB             = jjnr[jidx+1];
 664             j_coord_offsetA  = DIM*jnrA;
 665             j_coord_offsetB  = DIM*jnrB;
 666
 667             /* load j atom coordinates */
 668             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 669                                               &jx0,&jy0,&jz0);
 670
 671             /* Calculate displacement vector */
 672             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 673             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 674             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 675             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 676             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 677             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 678             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 679             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 680             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 681
 682             /* Calculate squared distance and things based on it */
 683             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 684             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 685             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 686
 687             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 688             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 689             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 690
 691             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 692             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 693             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 694
 695             /* Load parameters for j particles */
 696             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 697
 698             fjx0             = _fjsp_setzero_v2r8();
 699             fjy0             = _fjsp_setzero_v2r8();
 700             fjz0             = _fjsp_setzero_v2r8();
 701
 702             /**************************
 703              * CALCULATE INTERACTIONS *
 704              **************************/
 705
 706             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 707
 708             /* Compute parameters for interactions between i and j atoms */
 709             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 710
 711             /* EWALD ELECTROSTATICS */
 712
 713             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 714             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 715             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 716             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 717             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 718
 719             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 720                                          &ewtabF,&ewtabFn);
 721             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 722             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 723
 724             fscal            = felec;
 725
 726             /* Update vectorial force */
 727             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 728             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 729             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 730
 731             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 732             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 733             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 734
 735             /**************************
 736              * CALCULATE INTERACTIONS *
 737              **************************/
 738
 739             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 740
 741             /* Compute parameters for interactions between i and j atoms */
 742             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 743
 744             /* EWALD ELECTROSTATICS */
 745
 746             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 747             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 748             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 749             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 750             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 751
 752             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 753                                          &ewtabF,&ewtabFn);
 754             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 755             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 756
 757             fscal            = felec;
 758
 759             /* Update vectorial force */
 760             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 761             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 762             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 763
 764             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 765             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 766             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 767
 768             /**************************
 769              * CALCULATE INTERACTIONS *
 770              **************************/
 771
 772             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 773
 774             /* Compute parameters for interactions between i and j atoms */
 775             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 776
 777             /* EWALD ELECTROSTATICS */
 778
 779             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 780             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 781             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 782             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 783             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 784
 785             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 786                                          &ewtabF,&ewtabFn);
 787             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 788             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 789
 790             fscal            = felec;
 791
 792             /* Update vectorial force */
 793             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 794             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 795             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 796
 797             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 798             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 799             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 800
 801             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 802
 803             /* Inner loop uses 120 flops */
 804         }
 805
 806         if(jidx<j_index_end)
 807         {
 808
 809             jnrA             = jjnr[jidx];
 810             j_coord_offsetA  = DIM*jnrA;
 811
 812             /* load j atom coordinates */
 813             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 814                                               &jx0,&jy0,&jz0);
 815
 816             /* Calculate displacement vector */
 817             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 818             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 819             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 820             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 821             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 822             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 823             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 824             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 825             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 826
 827             /* Calculate squared distance and things based on it */
 828             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 829             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 830             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 831
 832             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 833             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 834             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 835
 836             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 837             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 838             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 839
 840             /* Load parameters for j particles */
 841             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 842
 843             fjx0             = _fjsp_setzero_v2r8();
 844             fjy0             = _fjsp_setzero_v2r8();
 845             fjz0             = _fjsp_setzero_v2r8();
 846
 847             /**************************
 848              * CALCULATE INTERACTIONS *
 849              **************************/
 850
 851             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 852
 853             /* Compute parameters for interactions between i and j atoms */
 854             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 855
 856             /* EWALD ELECTROSTATICS */
 857
 858             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 859             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 860             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 861             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 862             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 863
 864             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 865             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 866             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 867
 868             fscal            = felec;
 869
 870             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 871
 872             /* Update vectorial force */
 873             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 874             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 875             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 876
 877             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 878             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 879             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 880
 881             /**************************
 882              * CALCULATE INTERACTIONS *
 883              **************************/
 884
 885             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 886
 887             /* Compute parameters for interactions between i and j atoms */
 888             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 889
 890             /* EWALD ELECTROSTATICS */
 891
 892             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 893             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 894             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 895             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 896             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 897
 898             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 899             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 900             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 901
 902             fscal            = felec;
 903
 904             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 905
 906             /* Update vectorial force */
 907             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 908             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 909             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 910
 911             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 912             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 913             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 914
 915             /**************************
 916              * CALCULATE INTERACTIONS *
 917              **************************/
 918
 919             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 920
 921             /* Compute parameters for interactions between i and j atoms */
 922             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 923
 924             /* EWALD ELECTROSTATICS */
 925
 926             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 927             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 928             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 929             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 930             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 931
 932             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 933             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 934             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 935
 936             fscal            = felec;
 937
 938             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 939
 940             /* Update vectorial force */
 941             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 942             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 943             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 944
 945             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 946             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 947             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 948
 949             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 950
 951             /* Inner loop uses 120 flops */
 952         }
 953
 954         /* End of innermost loop */
 955
 956         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 957                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 958
 959         /* Increment number of inner iterations */
 960         inneriter                  += j_index_end - j_index_start;
 961
 962         /* Outer loop uses 18 flops */
 963     }
 964
 965     /* Increment number of outer iterations */
 966     outeriter        += nri;
 967
 968     /* Update outer/inner flops */
 969
 970     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*120);
 971 }