src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwNone_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            None
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  93     real             *vftab;
  94     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113
 114     vftab            = kernel_data->table_elec->data;
 115     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 116
 117     /* Setup water-specific parameters */
 118     inr              = nlist->iinr[0];
 119     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 120     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 121     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 122
 123     /* Avoid stupid compiler warnings */
 124     jnrA = jnrB = 0;
 125     j_coord_offsetA = 0;
 126     j_coord_offsetB = 0;
 127
 128     outeriter        = 0;
 129     inneriter        = 0;
 130
 131     /* Start outer loop over neighborlists */
 132     for(iidx=0; iidx<nri; iidx++)
 133     {
 134         /* Load shift vector for this list */
 135         i_shift_offset   = DIM*shiftidx[iidx];
 136
 137         /* Load limits for loop over neighbors */
 138         j_index_start    = jindex[iidx];
 139         j_index_end      = jindex[iidx+1];
 140
 141         /* Get outer coordinate index */
 142         inr              = iinr[iidx];
 143         i_coord_offset   = DIM*inr;
 144
 145         /* Load i particle coords and add shift vector */
 146         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 147                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 148
 149         fix0             = _fjsp_setzero_v2r8();
 150         fiy0             = _fjsp_setzero_v2r8();
 151         fiz0             = _fjsp_setzero_v2r8();
 152         fix1             = _fjsp_setzero_v2r8();
 153         fiy1             = _fjsp_setzero_v2r8();
 154         fiz1             = _fjsp_setzero_v2r8();
 155         fix2             = _fjsp_setzero_v2r8();
 156         fiy2             = _fjsp_setzero_v2r8();
 157         fiz2             = _fjsp_setzero_v2r8();
 158
 159         /* Reset potential sums */
 160         velecsum         = _fjsp_setzero_v2r8();
 161
 162         /* Start inner kernel loop */
 163         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 164         {
 165
 166             /* Get j neighbor index, and coordinate index */
 167             jnrA             = jjnr[jidx];
 168             jnrB             = jjnr[jidx+1];
 169             j_coord_offsetA  = DIM*jnrA;
 170             j_coord_offsetB  = DIM*jnrB;
 171
 172             /* load j atom coordinates */
 173             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 174                                               &jx0,&jy0,&jz0);
 175
 176             /* Calculate displacement vector */
 177             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 178             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 179             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 180             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 181             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 182             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 183             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 184             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 185             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 186
 187             /* Calculate squared distance and things based on it */
 188             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 189             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 190             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 191
 192             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 193             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 194             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 195
 196             /* Load parameters for j particles */
 197             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 198
 199             fjx0             = _fjsp_setzero_v2r8();
 200             fjy0             = _fjsp_setzero_v2r8();
 201             fjz0             = _fjsp_setzero_v2r8();
 202
 203             /**************************
 204              * CALCULATE INTERACTIONS *
 205              **************************/
 206
 207             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 208
 209             /* Compute parameters for interactions between i and j atoms */
 210             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 211
 212             /* Calculate table index by multiplying r with table scale and truncate to integer */
 213             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 214             itab_tmp         = _fjsp_dtox_v2r8(rt);
 215             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 216             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 217             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 218
 219             vfconv.i[0]     *= 4;
 220             vfconv.i[1]     *= 4;
 221
 222             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 223             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 224             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 225             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 226             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 227             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 228             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 229             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 230             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 231             velec            = _fjsp_mul_v2r8(qq00,VV);
 232             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 233             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 234
 235             /* Update potential sum for this i atom from the interaction with this j atom. */
 236             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 237
 238             fscal            = felec;
 239
 240             /* Update vectorial force */
 241             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 242             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 243             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 244
 245             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 246             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 247             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 248
 249             /**************************
 250              * CALCULATE INTERACTIONS *
 251              **************************/
 252
 253             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 254
 255             /* Compute parameters for interactions between i and j atoms */
 256             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 257
 258             /* Calculate table index by multiplying r with table scale and truncate to integer */
 259             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 260             itab_tmp         = _fjsp_dtox_v2r8(rt);
 261             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 262             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 263             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 264
 265             vfconv.i[0]     *= 4;
 266             vfconv.i[1]     *= 4;
 267
 268             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 269             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 270             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 271             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 272             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 273             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 274             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 275             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 276             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 277             velec            = _fjsp_mul_v2r8(qq10,VV);
 278             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 279             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 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             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 288             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 289             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 300
 301             /* Compute parameters for interactions between i and j atoms */
 302             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 303
 304             /* Calculate table index by multiplying r with table scale and truncate to integer */
 305             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 306             itab_tmp         = _fjsp_dtox_v2r8(rt);
 307             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 308             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 309             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 310
 311             vfconv.i[0]     *= 4;
 312             vfconv.i[1]     *= 4;
 313
 314             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 315             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 316             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 317             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 318             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 319             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 320             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 321             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 322             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 323             velec            = _fjsp_mul_v2r8(qq20,VV);
 324             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 325             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 326
 327             /* Update potential sum for this i atom from the interaction with this j atom. */
 328             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 329
 330             fscal            = felec;
 331
 332             /* Update vectorial force */
 333             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 334             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 335             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 336
 337             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 338             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 339             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 340
 341             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 342
 343             /* Inner loop uses 141 flops */
 344         }
 345
 346         if(jidx<j_index_end)
 347         {
 348
 349             jnrA             = jjnr[jidx];
 350             j_coord_offsetA  = DIM*jnrA;
 351
 352             /* load j atom coordinates */
 353             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 354                                               &jx0,&jy0,&jz0);
 355
 356             /* Calculate displacement vector */
 357             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 358             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 359             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 360             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 361             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 362             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 363             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 364             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 365             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 366
 367             /* Calculate squared distance and things based on it */
 368             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 369             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 370             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 371
 372             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 373             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 374             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 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             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 388
 389             /* Compute parameters for interactions between i and j atoms */
 390             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 391
 392             /* Calculate table index by multiplying r with table scale and truncate to integer */
 393             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 394             itab_tmp         = _fjsp_dtox_v2r8(rt);
 395             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 396             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 397             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 398
 399             vfconv.i[0]     *= 4;
 400             vfconv.i[1]     *= 4;
 401
 402             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 403             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 404             F                = _fjsp_setzero_v2r8();
 405             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 406             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 407             H                = _fjsp_setzero_v2r8();
 408             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 409             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 410             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 411             velec            = _fjsp_mul_v2r8(qq00,VV);
 412             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 413             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 414
 415             /* Update potential sum for this i atom from the interaction with this j atom. */
 416             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 417             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 418
 419             fscal            = felec;
 420
 421             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 422
 423             /* Update vectorial force */
 424             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 425             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 426             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 427
 428             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 429             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 430             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 437
 438             /* Compute parameters for interactions between i and j atoms */
 439             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 440
 441             /* Calculate table index by multiplying r with table scale and truncate to integer */
 442             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 443             itab_tmp         = _fjsp_dtox_v2r8(rt);
 444             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 445             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 446             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 447
 448             vfconv.i[0]     *= 4;
 449             vfconv.i[1]     *= 4;
 450
 451             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 452             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 453             F                = _fjsp_setzero_v2r8();
 454             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 455             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 456             H                = _fjsp_setzero_v2r8();
 457             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 458             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 459             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 460             velec            = _fjsp_mul_v2r8(qq10,VV);
 461             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 462             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 463
 464             /* Update potential sum for this i atom from the interaction with this j atom. */
 465             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 466             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 467
 468             fscal            = felec;
 469
 470             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 471
 472             /* Update vectorial force */
 473             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 474             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 475             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 476
 477             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 478             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 479             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 480
 481             /**************************
 482              * CALCULATE INTERACTIONS *
 483              **************************/
 484
 485             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 486
 487             /* Compute parameters for interactions between i and j atoms */
 488             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 489
 490             /* Calculate table index by multiplying r with table scale and truncate to integer */
 491             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 492             itab_tmp         = _fjsp_dtox_v2r8(rt);
 493             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 494             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 495             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 496
 497             vfconv.i[0]     *= 4;
 498             vfconv.i[1]     *= 4;
 499
 500             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 501             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 502             F                = _fjsp_setzero_v2r8();
 503             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 504             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 505             H                = _fjsp_setzero_v2r8();
 506             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 507             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 508             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 509             velec            = _fjsp_mul_v2r8(qq20,VV);
 510             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 511             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 512
 513             /* Update potential sum for this i atom from the interaction with this j atom. */
 514             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 515             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 520
 521             /* Update vectorial force */
 522             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 523             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 524             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 525
 526             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 527             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 528             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 529
 530             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 531
 532             /* Inner loop uses 141 flops */
 533         }
 534
 535         /* End of innermost loop */
 536
 537         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 538                                               f+i_coord_offset,fshift+i_shift_offset);
 539
 540         ggid                        = gid[iidx];
 541         /* Update potential energies */
 542         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 543
 544         /* Increment number of inner iterations */
 545         inneriter                  += j_index_end - j_index_start;
 546
 547         /* Outer loop uses 19 flops */
 548     }
 549
 550     /* Increment number of outer iterations */
 551     outeriter        += nri;
 552
 553     /* Update outer/inner flops */
 554
 555     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*141);
 556 }
 557 /*
 558  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 559  * Electrostatics interaction: CubicSplineTable
 560  * VdW interaction:            None
 561  * Geometry:                   Water3-Particle
 562  * Calculate force/pot:        Force
 563  */
 564 void
 565 nb_kernel_ElecCSTab_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 566                     (t_nblist                    * gmx_restrict       nlist,
 567                      rvec                        * gmx_restrict          xx,
 568                      rvec                        * gmx_restrict          ff,
 569                      t_forcerec                  * gmx_restrict          fr,
 570                      t_mdatoms                   * gmx_restrict     mdatoms,
 571                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 572                      t_nrnb                      * gmx_restrict        nrnb)
 573 {
 574     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 575      * just 0 for non-waters.
 576      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 577      * jnr indices corresponding to data put in the four positions in the SIMD register.
 578      */
 579     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 580     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 581     int              jnrA,jnrB;
 582     int              j_coord_offsetA,j_coord_offsetB;
 583     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 584     real             rcutoff_scalar;
 585     real             *shiftvec,*fshift,*x,*f;
 586     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 587     int              vdwioffset0;
 588     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 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              vdwjidx0A,vdwjidx0B;
 594     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 595     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 596     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 597     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 598     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 599     real             *charge;
 600     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 601     real             *vftab;
 602     _fjsp_v2r8       itab_tmp;
 603     _fjsp_v2r8       dummy_mask,cutoff_mask;
 604     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 605     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 606     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 607
 608     x                = xx[0];
 609     f                = ff[0];
 610
 611     nri              = nlist->nri;
 612     iinr             = nlist->iinr;
 613     jindex           = nlist->jindex;
 614     jjnr             = nlist->jjnr;
 615     shiftidx         = nlist->shift;
 616     gid              = nlist->gid;
 617     shiftvec         = fr->shift_vec[0];
 618     fshift           = fr->fshift[0];
 619     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 620     charge           = mdatoms->chargeA;
 621
 622     vftab            = kernel_data->table_elec->data;
 623     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 624
 625     /* Setup water-specific parameters */
 626     inr              = nlist->iinr[0];
 627     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 628     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 629     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 630
 631     /* Avoid stupid compiler warnings */
 632     jnrA = jnrB = 0;
 633     j_coord_offsetA = 0;
 634     j_coord_offsetB = 0;
 635
 636     outeriter        = 0;
 637     inneriter        = 0;
 638
 639     /* Start outer loop over neighborlists */
 640     for(iidx=0; iidx<nri; iidx++)
 641     {
 642         /* Load shift vector for this list */
 643         i_shift_offset   = DIM*shiftidx[iidx];
 644
 645         /* Load limits for loop over neighbors */
 646         j_index_start    = jindex[iidx];
 647         j_index_end      = jindex[iidx+1];
 648
 649         /* Get outer coordinate index */
 650         inr              = iinr[iidx];
 651         i_coord_offset   = DIM*inr;
 652
 653         /* Load i particle coords and add shift vector */
 654         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 655                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 656
 657         fix0             = _fjsp_setzero_v2r8();
 658         fiy0             = _fjsp_setzero_v2r8();
 659         fiz0             = _fjsp_setzero_v2r8();
 660         fix1             = _fjsp_setzero_v2r8();
 661         fiy1             = _fjsp_setzero_v2r8();
 662         fiz1             = _fjsp_setzero_v2r8();
 663         fix2             = _fjsp_setzero_v2r8();
 664         fiy2             = _fjsp_setzero_v2r8();
 665         fiz2             = _fjsp_setzero_v2r8();
 666
 667         /* Start inner kernel loop */
 668         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 669         {
 670
 671             /* Get j neighbor index, and coordinate index */
 672             jnrA             = jjnr[jidx];
 673             jnrB             = jjnr[jidx+1];
 674             j_coord_offsetA  = DIM*jnrA;
 675             j_coord_offsetB  = DIM*jnrB;
 676
 677             /* load j atom coordinates */
 678             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 679                                               &jx0,&jy0,&jz0);
 680
 681             /* Calculate displacement vector */
 682             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 683             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 684             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 685             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 686             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 687             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 688             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 689             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 690             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 691
 692             /* Calculate squared distance and things based on it */
 693             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 694             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 695             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 696
 697             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 698             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 699             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 700
 701             /* Load parameters for j particles */
 702             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 703
 704             fjx0             = _fjsp_setzero_v2r8();
 705             fjy0             = _fjsp_setzero_v2r8();
 706             fjz0             = _fjsp_setzero_v2r8();
 707
 708             /**************************
 709              * CALCULATE INTERACTIONS *
 710              **************************/
 711
 712             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 713
 714             /* Compute parameters for interactions between i and j atoms */
 715             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 716
 717             /* Calculate table index by multiplying r with table scale and truncate to integer */
 718             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 719             itab_tmp         = _fjsp_dtox_v2r8(rt);
 720             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 721             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 722             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 723
 724             vfconv.i[0]     *= 4;
 725             vfconv.i[1]     *= 4;
 726
 727             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 728             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 729             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 730             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 731             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 732             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 733             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 734             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 735             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 736             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 737
 738             fscal            = felec;
 739
 740             /* Update vectorial force */
 741             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 742             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 743             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 744
 745             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 746             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 747             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 748
 749             /**************************
 750              * CALCULATE INTERACTIONS *
 751              **************************/
 752
 753             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 754
 755             /* Compute parameters for interactions between i and j atoms */
 756             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 757
 758             /* Calculate table index by multiplying r with table scale and truncate to integer */
 759             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 760             itab_tmp         = _fjsp_dtox_v2r8(rt);
 761             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 762             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 763             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 764
 765             vfconv.i[0]     *= 4;
 766             vfconv.i[1]     *= 4;
 767
 768             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 769             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 770             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 771             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 772             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 773             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 774             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 775             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 776             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 777             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 778
 779             fscal            = felec;
 780
 781             /* Update vectorial force */
 782             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 783             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 784             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 785
 786             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 787             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 788             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 789
 790             /**************************
 791              * CALCULATE INTERACTIONS *
 792              **************************/
 793
 794             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 795
 796             /* Compute parameters for interactions between i and j atoms */
 797             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 798
 799             /* Calculate table index by multiplying r with table scale and truncate to integer */
 800             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 801             itab_tmp         = _fjsp_dtox_v2r8(rt);
 802             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 803             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 804             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 805
 806             vfconv.i[0]     *= 4;
 807             vfconv.i[1]     *= 4;
 808
 809             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 810             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 811             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 812             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 813             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 814             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 815             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 816             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 817             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 818             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 819
 820             fscal            = felec;
 821
 822             /* Update vectorial force */
 823             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 824             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 825             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 826
 827             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 828             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 829             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 830
 831             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 832
 833             /* Inner loop uses 129 flops */
 834         }
 835
 836         if(jidx<j_index_end)
 837         {
 838
 839             jnrA             = jjnr[jidx];
 840             j_coord_offsetA  = DIM*jnrA;
 841
 842             /* load j atom coordinates */
 843             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 844                                               &jx0,&jy0,&jz0);
 845
 846             /* Calculate displacement vector */
 847             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 848             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 849             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 850             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 851             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 852             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 853             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 854             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 855             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 856
 857             /* Calculate squared distance and things based on it */
 858             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 859             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 860             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 861
 862             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 863             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 864             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 865
 866             /* Load parameters for j particles */
 867             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 868
 869             fjx0             = _fjsp_setzero_v2r8();
 870             fjy0             = _fjsp_setzero_v2r8();
 871             fjz0             = _fjsp_setzero_v2r8();
 872
 873             /**************************
 874              * CALCULATE INTERACTIONS *
 875              **************************/
 876
 877             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 878
 879             /* Compute parameters for interactions between i and j atoms */
 880             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 881
 882             /* Calculate table index by multiplying r with table scale and truncate to integer */
 883             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 884             itab_tmp         = _fjsp_dtox_v2r8(rt);
 885             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 886             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 887             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 888
 889             vfconv.i[0]     *= 4;
 890             vfconv.i[1]     *= 4;
 891
 892             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 893             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 894             F                = _fjsp_setzero_v2r8();
 895             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 896             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 897             H                = _fjsp_setzero_v2r8();
 898             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 899             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 900             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 901             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 902
 903             fscal            = felec;
 904
 905             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 906
 907             /* Update vectorial force */
 908             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 909             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 910             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 911
 912             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 913             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 914             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 915
 916             /**************************
 917              * CALCULATE INTERACTIONS *
 918              **************************/
 919
 920             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 921
 922             /* Compute parameters for interactions between i and j atoms */
 923             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 924
 925             /* Calculate table index by multiplying r with table scale and truncate to integer */
 926             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 927             itab_tmp         = _fjsp_dtox_v2r8(rt);
 928             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 929             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 930             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 931
 932             vfconv.i[0]     *= 4;
 933             vfconv.i[1]     *= 4;
 934
 935             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 936             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 937             F                = _fjsp_setzero_v2r8();
 938             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 939             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 940             H                = _fjsp_setzero_v2r8();
 941             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 942             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 943             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 944             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 945
 946             fscal            = felec;
 947
 948             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 949
 950             /* Update vectorial force */
 951             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 952             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 953             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 954
 955             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 956             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 957             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 958
 959             /**************************
 960              * CALCULATE INTERACTIONS *
 961              **************************/
 962
 963             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 964
 965             /* Compute parameters for interactions between i and j atoms */
 966             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 967
 968             /* Calculate table index by multiplying r with table scale and truncate to integer */
 969             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 970             itab_tmp         = _fjsp_dtox_v2r8(rt);
 971             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 972             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 973             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 974
 975             vfconv.i[0]     *= 4;
 976             vfconv.i[1]     *= 4;
 977
 978             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 979             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 980             F                = _fjsp_setzero_v2r8();
 981             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 982             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 983             H                = _fjsp_setzero_v2r8();
 984             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 985             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 986             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 987             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 988
 989             fscal            = felec;
 990
 991             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 992
 993             /* Update vectorial force */
 994             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 995             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 996             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 997
 998             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 999             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1000             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1001
1002             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1003
1004             /* Inner loop uses 129 flops */
1005         }
1006
1007         /* End of innermost loop */
1008
1009         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1010                                               f+i_coord_offset,fshift+i_shift_offset);
1011
1012         /* Increment number of inner iterations */
1013         inneriter                  += j_index_end - j_index_start;
1014
1015         /* Outer loop uses 18 flops */
1016     }
1017
1018     /* Increment number of outer iterations */
1019     outeriter        += nri;
1020
1021     /* Update outer/inner flops */
1022
1023     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*129);
1024 }