src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_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/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            None
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_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       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  95     real             *vftab;
  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     vftab            = kernel_data->table_elec->data;
 117     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 118
 119     /* Setup water-specific parameters */
 120     inr              = nlist->iinr[0];
 121     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 122     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 123     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 124
 125     /* Avoid stupid compiler warnings */
 126     jnrA = jnrB = 0;
 127     j_coord_offsetA = 0;
 128     j_coord_offsetB = 0;
 129
 130     outeriter        = 0;
 131     inneriter        = 0;
 132
 133     /* Start outer loop over neighborlists */
 134     for(iidx=0; iidx<nri; iidx++)
 135     {
 136         /* Load shift vector for this list */
 137         i_shift_offset   = DIM*shiftidx[iidx];
 138
 139         /* Load limits for loop over neighbors */
 140         j_index_start    = jindex[iidx];
 141         j_index_end      = jindex[iidx+1];
 142
 143         /* Get outer coordinate index */
 144         inr              = iinr[iidx];
 145         i_coord_offset   = DIM*inr;
 146
 147         /* Load i particle coords and add shift vector */
 148         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 149                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 150
 151         fix1             = _fjsp_setzero_v2r8();
 152         fiy1             = _fjsp_setzero_v2r8();
 153         fiz1             = _fjsp_setzero_v2r8();
 154         fix2             = _fjsp_setzero_v2r8();
 155         fiy2             = _fjsp_setzero_v2r8();
 156         fiz2             = _fjsp_setzero_v2r8();
 157         fix3             = _fjsp_setzero_v2r8();
 158         fiy3             = _fjsp_setzero_v2r8();
 159         fiz3             = _fjsp_setzero_v2r8();
 160
 161         /* Reset potential sums */
 162         velecsum         = _fjsp_setzero_v2r8();
 163
 164         /* Start inner kernel loop */
 165         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 166         {
 167
 168             /* Get j neighbor index, and coordinate index */
 169             jnrA             = jjnr[jidx];
 170             jnrB             = jjnr[jidx+1];
 171             j_coord_offsetA  = DIM*jnrA;
 172             j_coord_offsetB  = DIM*jnrB;
 173
 174             /* load j atom coordinates */
 175             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 176                                               &jx0,&jy0,&jz0);
 177
 178             /* Calculate displacement vector */
 179             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 180             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 181             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 182             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 183             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 184             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 185             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 186             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 187             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 188
 189             /* Calculate squared distance and things based on it */
 190             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 191             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 192             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 193
 194             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 195             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 196             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 197
 198             /* Load parameters for j particles */
 199             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 200
 201             fjx0             = _fjsp_setzero_v2r8();
 202             fjy0             = _fjsp_setzero_v2r8();
 203             fjz0             = _fjsp_setzero_v2r8();
 204
 205             /**************************
 206              * CALCULATE INTERACTIONS *
 207              **************************/
 208
 209             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 210
 211             /* Compute parameters for interactions between i and j atoms */
 212             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 213
 214             /* Calculate table index by multiplying r with table scale and truncate to integer */
 215             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 216             itab_tmp         = _fjsp_dtox_v2r8(rt);
 217             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 218             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 219             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 220
 221             vfconv.i[0]     *= 4;
 222             vfconv.i[1]     *= 4;
 223
 224             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 225             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 226             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 227             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 228             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 229             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 230             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 231             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 232             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 233             velec            = _fjsp_mul_v2r8(qq10,VV);
 234             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 235             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 236
 237             /* Update potential sum for this i atom from the interaction with this j atom. */
 238             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 239
 240             fscal            = felec;
 241
 242             /* Update vectorial force */
 243             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 244             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 245             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 246
 247             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 248             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 249             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 250
 251             /**************************
 252              * CALCULATE INTERACTIONS *
 253              **************************/
 254
 255             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 256
 257             /* Compute parameters for interactions between i and j atoms */
 258             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 259
 260             /* Calculate table index by multiplying r with table scale and truncate to integer */
 261             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 262             itab_tmp         = _fjsp_dtox_v2r8(rt);
 263             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 264             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 265             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 266
 267             vfconv.i[0]     *= 4;
 268             vfconv.i[1]     *= 4;
 269
 270             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 271             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 272             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 273             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 274             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 275             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 276             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 277             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 278             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 279             velec            = _fjsp_mul_v2r8(qq20,VV);
 280             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 281             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 282
 283             /* Update potential sum for this i atom from the interaction with this j atom. */
 284             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 285
 286             fscal            = felec;
 287
 288             /* Update vectorial force */
 289             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 290             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 291             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 292
 293             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 294             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 295             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 302
 303             /* Compute parameters for interactions between i and j atoms */
 304             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 305
 306             /* Calculate table index by multiplying r with table scale and truncate to integer */
 307             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 308             itab_tmp         = _fjsp_dtox_v2r8(rt);
 309             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 310             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 311             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 312
 313             vfconv.i[0]     *= 4;
 314             vfconv.i[1]     *= 4;
 315
 316             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 317             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 318             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 319             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 320             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 321             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 322             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 323             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 324             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 325             velec            = _fjsp_mul_v2r8(qq30,VV);
 326             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 327             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 336             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 337             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 342
 343             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 344
 345             /* Inner loop uses 141 flops */
 346         }
 347
 348         if(jidx<j_index_end)
 349         {
 350
 351             jnrA             = jjnr[jidx];
 352             j_coord_offsetA  = DIM*jnrA;
 353
 354             /* load j atom coordinates */
 355             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 356                                               &jx0,&jy0,&jz0);
 357
 358             /* Calculate displacement vector */
 359             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 360             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 361             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 362             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 363             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 364             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 365             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 366             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 367             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 368
 369             /* Calculate squared distance and things based on it */
 370             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 371             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 372             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 373
 374             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 375             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 376             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 377
 378             /* Load parameters for j particles */
 379             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 380
 381             fjx0             = _fjsp_setzero_v2r8();
 382             fjy0             = _fjsp_setzero_v2r8();
 383             fjz0             = _fjsp_setzero_v2r8();
 384
 385             /**************************
 386              * CALCULATE INTERACTIONS *
 387              **************************/
 388
 389             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 390
 391             /* Compute parameters for interactions between i and j atoms */
 392             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 393
 394             /* Calculate table index by multiplying r with table scale and truncate to integer */
 395             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 396             itab_tmp         = _fjsp_dtox_v2r8(rt);
 397             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 398             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 399             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 400
 401             vfconv.i[0]     *= 4;
 402             vfconv.i[1]     *= 4;
 403
 404             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 405             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 406             F                = _fjsp_setzero_v2r8();
 407             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 408             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 409             H                = _fjsp_setzero_v2r8();
 410             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 411             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 412             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 413             velec            = _fjsp_mul_v2r8(qq10,VV);
 414             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 415             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 419             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 420
 421             fscal            = felec;
 422
 423             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 424
 425             /* Update vectorial force */
 426             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 427             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 428             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 429
 430             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 431             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 432             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 433
 434             /**************************
 435              * CALCULATE INTERACTIONS *
 436              **************************/
 437
 438             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 439
 440             /* Compute parameters for interactions between i and j atoms */
 441             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 442
 443             /* Calculate table index by multiplying r with table scale and truncate to integer */
 444             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 445             itab_tmp         = _fjsp_dtox_v2r8(rt);
 446             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 447             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 448             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 449
 450             vfconv.i[0]     *= 4;
 451             vfconv.i[1]     *= 4;
 452
 453             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 454             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 455             F                = _fjsp_setzero_v2r8();
 456             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 457             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 458             H                = _fjsp_setzero_v2r8();
 459             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 460             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 461             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 462             velec            = _fjsp_mul_v2r8(qq20,VV);
 463             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 464             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 468             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 473
 474             /* Update vectorial force */
 475             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 476             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 477             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 478
 479             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 480             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 481             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 482
 483             /**************************
 484              * CALCULATE INTERACTIONS *
 485              **************************/
 486
 487             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 488
 489             /* Compute parameters for interactions between i and j atoms */
 490             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 491
 492             /* Calculate table index by multiplying r with table scale and truncate to integer */
 493             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 494             itab_tmp         = _fjsp_dtox_v2r8(rt);
 495             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 496             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 497             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 498
 499             vfconv.i[0]     *= 4;
 500             vfconv.i[1]     *= 4;
 501
 502             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 503             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 504             F                = _fjsp_setzero_v2r8();
 505             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 506             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 507             H                = _fjsp_setzero_v2r8();
 508             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 509             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 510             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 511             velec            = _fjsp_mul_v2r8(qq30,VV);
 512             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 513             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 514
 515             /* Update potential sum for this i atom from the interaction with this j atom. */
 516             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 517             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 518
 519             fscal            = felec;
 520
 521             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 522
 523             /* Update vectorial force */
 524             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 525             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 526             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 527
 528             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 529             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 530             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 531
 532             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 533
 534             /* Inner loop uses 141 flops */
 535         }
 536
 537         /* End of innermost loop */
 538
 539         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 540                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 541
 542         ggid                        = gid[iidx];
 543         /* Update potential energies */
 544         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 545
 546         /* Increment number of inner iterations */
 547         inneriter                  += j_index_end - j_index_start;
 548
 549         /* Outer loop uses 19 flops */
 550     }
 551
 552     /* Increment number of outer iterations */
 553     outeriter        += nri;
 554
 555     /* Update outer/inner flops */
 556
 557     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*141);
 558 }
 559 /*
 560  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 561  * Electrostatics interaction: CubicSplineTable
 562  * VdW interaction:            None
 563  * Geometry:                   Water4-Particle
 564  * Calculate force/pot:        Force
 565  */
 566 void
 567 nb_kernel_ElecCSTab_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 568                     (t_nblist                    * gmx_restrict       nlist,
 569                      rvec                        * gmx_restrict          xx,
 570                      rvec                        * gmx_restrict          ff,
 571                      t_forcerec                  * gmx_restrict          fr,
 572                      t_mdatoms                   * gmx_restrict     mdatoms,
 573                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 574                      t_nrnb                      * gmx_restrict        nrnb)
 575 {
 576     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 577      * just 0 for non-waters.
 578      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 579      * jnr indices corresponding to data put in the four positions in the SIMD register.
 580      */
 581     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 582     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 583     int              jnrA,jnrB;
 584     int              j_coord_offsetA,j_coord_offsetB;
 585     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 586     real             rcutoff_scalar;
 587     real             *shiftvec,*fshift,*x,*f;
 588     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 589     int              vdwioffset1;
 590     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 591     int              vdwioffset2;
 592     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 593     int              vdwioffset3;
 594     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 595     int              vdwjidx0A,vdwjidx0B;
 596     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 597     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 598     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 599     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 600     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 601     real             *charge;
 602     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 603     real             *vftab;
 604     _fjsp_v2r8       itab_tmp;
 605     _fjsp_v2r8       dummy_mask,cutoff_mask;
 606     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 607     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 608     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 609
 610     x                = xx[0];
 611     f                = ff[0];
 612
 613     nri              = nlist->nri;
 614     iinr             = nlist->iinr;
 615     jindex           = nlist->jindex;
 616     jjnr             = nlist->jjnr;
 617     shiftidx         = nlist->shift;
 618     gid              = nlist->gid;
 619     shiftvec         = fr->shift_vec[0];
 620     fshift           = fr->fshift[0];
 621     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 622     charge           = mdatoms->chargeA;
 623
 624     vftab            = kernel_data->table_elec->data;
 625     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 626
 627     /* Setup water-specific parameters */
 628     inr              = nlist->iinr[0];
 629     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 630     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 631     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 632
 633     /* Avoid stupid compiler warnings */
 634     jnrA = jnrB = 0;
 635     j_coord_offsetA = 0;
 636     j_coord_offsetB = 0;
 637
 638     outeriter        = 0;
 639     inneriter        = 0;
 640
 641     /* Start outer loop over neighborlists */
 642     for(iidx=0; iidx<nri; iidx++)
 643     {
 644         /* Load shift vector for this list */
 645         i_shift_offset   = DIM*shiftidx[iidx];
 646
 647         /* Load limits for loop over neighbors */
 648         j_index_start    = jindex[iidx];
 649         j_index_end      = jindex[iidx+1];
 650
 651         /* Get outer coordinate index */
 652         inr              = iinr[iidx];
 653         i_coord_offset   = DIM*inr;
 654
 655         /* Load i particle coords and add shift vector */
 656         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 657                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 658
 659         fix1             = _fjsp_setzero_v2r8();
 660         fiy1             = _fjsp_setzero_v2r8();
 661         fiz1             = _fjsp_setzero_v2r8();
 662         fix2             = _fjsp_setzero_v2r8();
 663         fiy2             = _fjsp_setzero_v2r8();
 664         fiz2             = _fjsp_setzero_v2r8();
 665         fix3             = _fjsp_setzero_v2r8();
 666         fiy3             = _fjsp_setzero_v2r8();
 667         fiz3             = _fjsp_setzero_v2r8();
 668
 669         /* Start inner kernel loop */
 670         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 671         {
 672
 673             /* Get j neighbor index, and coordinate index */
 674             jnrA             = jjnr[jidx];
 675             jnrB             = jjnr[jidx+1];
 676             j_coord_offsetA  = DIM*jnrA;
 677             j_coord_offsetB  = DIM*jnrB;
 678
 679             /* load j atom coordinates */
 680             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 681                                               &jx0,&jy0,&jz0);
 682
 683             /* Calculate displacement vector */
 684             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 685             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 686             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 687             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 688             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 689             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 690             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 691             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 692             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 693
 694             /* Calculate squared distance and things based on it */
 695             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 696             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 697             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 698
 699             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 700             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 701             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 702
 703             /* Load parameters for j particles */
 704             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 705
 706             fjx0             = _fjsp_setzero_v2r8();
 707             fjy0             = _fjsp_setzero_v2r8();
 708             fjz0             = _fjsp_setzero_v2r8();
 709
 710             /**************************
 711              * CALCULATE INTERACTIONS *
 712              **************************/
 713
 714             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 715
 716             /* Compute parameters for interactions between i and j atoms */
 717             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 718
 719             /* Calculate table index by multiplying r with table scale and truncate to integer */
 720             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 721             itab_tmp         = _fjsp_dtox_v2r8(rt);
 722             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 723             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 724             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 725
 726             vfconv.i[0]     *= 4;
 727             vfconv.i[1]     *= 4;
 728
 729             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 730             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 731             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 732             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 733             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 734             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 735             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 736             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 737             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 738             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 739
 740             fscal            = felec;
 741
 742             /* Update vectorial force */
 743             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 744             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 745             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 746
 747             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 748             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 749             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 750
 751             /**************************
 752              * CALCULATE INTERACTIONS *
 753              **************************/
 754
 755             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 756
 757             /* Compute parameters for interactions between i and j atoms */
 758             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 759
 760             /* Calculate table index by multiplying r with table scale and truncate to integer */
 761             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 762             itab_tmp         = _fjsp_dtox_v2r8(rt);
 763             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 764             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 765             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 766
 767             vfconv.i[0]     *= 4;
 768             vfconv.i[1]     *= 4;
 769
 770             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 771             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 772             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 773             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 774             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 775             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 776             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 777             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 778             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 779             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 780
 781             fscal            = felec;
 782
 783             /* Update vectorial force */
 784             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 785             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 786             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 787
 788             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 789             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 790             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 791
 792             /**************************
 793              * CALCULATE INTERACTIONS *
 794              **************************/
 795
 796             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 797
 798             /* Compute parameters for interactions between i and j atoms */
 799             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 800
 801             /* Calculate table index by multiplying r with table scale and truncate to integer */
 802             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 803             itab_tmp         = _fjsp_dtox_v2r8(rt);
 804             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 805             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 806             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 807
 808             vfconv.i[0]     *= 4;
 809             vfconv.i[1]     *= 4;
 810
 811             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 812             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 813             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 814             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 815             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 816             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 817             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 818             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 819             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 820             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 821
 822             fscal            = felec;
 823
 824             /* Update vectorial force */
 825             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 826             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 827             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 828
 829             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 830             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 831             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 832
 833             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 834
 835             /* Inner loop uses 129 flops */
 836         }
 837
 838         if(jidx<j_index_end)
 839         {
 840
 841             jnrA             = jjnr[jidx];
 842             j_coord_offsetA  = DIM*jnrA;
 843
 844             /* load j atom coordinates */
 845             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 846                                               &jx0,&jy0,&jz0);
 847
 848             /* Calculate displacement vector */
 849             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 850             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 851             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 852             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 853             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 854             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 855             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 856             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 857             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 858
 859             /* Calculate squared distance and things based on it */
 860             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 861             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 862             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 863
 864             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 865             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 866             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 867
 868             /* Load parameters for j particles */
 869             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 870
 871             fjx0             = _fjsp_setzero_v2r8();
 872             fjy0             = _fjsp_setzero_v2r8();
 873             fjz0             = _fjsp_setzero_v2r8();
 874
 875             /**************************
 876              * CALCULATE INTERACTIONS *
 877              **************************/
 878
 879             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 880
 881             /* Compute parameters for interactions between i and j atoms */
 882             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 883
 884             /* Calculate table index by multiplying r with table scale and truncate to integer */
 885             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 886             itab_tmp         = _fjsp_dtox_v2r8(rt);
 887             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 888             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 889             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 890
 891             vfconv.i[0]     *= 4;
 892             vfconv.i[1]     *= 4;
 893
 894             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 895             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 896             F                = _fjsp_setzero_v2r8();
 897             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 898             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 899             H                = _fjsp_setzero_v2r8();
 900             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 901             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 902             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 903             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 904
 905             fscal            = felec;
 906
 907             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 908
 909             /* Update vectorial force */
 910             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 911             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 912             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 913
 914             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 915             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 916             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 917
 918             /**************************
 919              * CALCULATE INTERACTIONS *
 920              **************************/
 921
 922             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 923
 924             /* Compute parameters for interactions between i and j atoms */
 925             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 926
 927             /* Calculate table index by multiplying r with table scale and truncate to integer */
 928             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 929             itab_tmp         = _fjsp_dtox_v2r8(rt);
 930             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 931             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 932             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 933
 934             vfconv.i[0]     *= 4;
 935             vfconv.i[1]     *= 4;
 936
 937             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 938             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 939             F                = _fjsp_setzero_v2r8();
 940             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 941             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 942             H                = _fjsp_setzero_v2r8();
 943             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 944             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 945             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 946             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 947
 948             fscal            = felec;
 949
 950             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 951
 952             /* Update vectorial force */
 953             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 954             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 955             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 956
 957             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 958             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 959             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 960
 961             /**************************
 962              * CALCULATE INTERACTIONS *
 963              **************************/
 964
 965             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 966
 967             /* Compute parameters for interactions between i and j atoms */
 968             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 969
 970             /* Calculate table index by multiplying r with table scale and truncate to integer */
 971             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 972             itab_tmp         = _fjsp_dtox_v2r8(rt);
 973             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 974             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 975             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 976
 977             vfconv.i[0]     *= 4;
 978             vfconv.i[1]     *= 4;
 979
 980             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 981             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 982             F                = _fjsp_setzero_v2r8();
 983             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 984             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 985             H                = _fjsp_setzero_v2r8();
 986             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 987             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 988             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 989             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 990
 991             fscal            = felec;
 992
 993             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 994
 995             /* Update vectorial force */
 996             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 997             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 998             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 999
1000             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1001             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1002             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1003
1004             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1005
1006             /* Inner loop uses 129 flops */
1007         }
1008
1009         /* End of innermost loop */
1010
1011         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1012                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1013
1014         /* Increment number of inner iterations */
1015         inneriter                  += j_index_end - j_index_start;
1016
1017         /* Outer loop uses 18 flops */
1018     }
1019
1020     /* Increment number of outer iterations */
1021     outeriter        += nri;
1022
1023     /* Update outer/inner flops */
1024
1025     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*129);
1026 }