src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwCSTab_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_ElecCoul_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwCSTab_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     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_vdw->data;
 124     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     /* Start outer loop over neighborlists */
 142     for(iidx=0; iidx<nri; iidx++)
 143     {
 144         /* Load shift vector for this list */
 145         i_shift_offset   = DIM*shiftidx[iidx];
 146
 147         /* Load limits for loop over neighbors */
 148         j_index_start    = jindex[iidx];
 149         j_index_end      = jindex[iidx+1];
 150
 151         /* Get outer coordinate index */
 152         inr              = iinr[iidx];
 153         i_coord_offset   = DIM*inr;
 154
 155         /* Load i particle coords and add shift vector */
 156         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 157                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 158
 159         fix0             = _fjsp_setzero_v2r8();
 160         fiy0             = _fjsp_setzero_v2r8();
 161         fiz0             = _fjsp_setzero_v2r8();
 162         fix1             = _fjsp_setzero_v2r8();
 163         fiy1             = _fjsp_setzero_v2r8();
 164         fiz1             = _fjsp_setzero_v2r8();
 165         fix2             = _fjsp_setzero_v2r8();
 166         fiy2             = _fjsp_setzero_v2r8();
 167         fiz2             = _fjsp_setzero_v2r8();
 168
 169         /* Reset potential sums */
 170         velecsum         = _fjsp_setzero_v2r8();
 171         vvdwsum          = _fjsp_setzero_v2r8();
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 175         {
 176
 177             /* Get j neighbor index, and coordinate index */
 178             jnrA             = jjnr[jidx];
 179             jnrB             = jjnr[jidx+1];
 180             j_coord_offsetA  = DIM*jnrA;
 181             j_coord_offsetB  = DIM*jnrB;
 182
 183             /* load j atom coordinates */
 184             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                               &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 189             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 190             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 191             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 192             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 193             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 194             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 195             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 196             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 197
 198             /* Calculate squared distance and things based on it */
 199             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 200             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 201             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 202
 203             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 204             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 205             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 206
 207             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 208             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 209             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 210
 211             /* Load parameters for j particles */
 212             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 213             vdwjidx0A        = 2*vdwtype[jnrA+0];
 214             vdwjidx0B        = 2*vdwtype[jnrB+0];
 215
 216             fjx0             = _fjsp_setzero_v2r8();
 217             fjy0             = _fjsp_setzero_v2r8();
 218             fjz0             = _fjsp_setzero_v2r8();
 219
 220             /**************************
 221              * CALCULATE INTERACTIONS *
 222              **************************/
 223
 224             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 225
 226             /* Compute parameters for interactions between i and j atoms */
 227             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 228             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 229                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 230
 231             /* Calculate table index by multiplying r with table scale and truncate to integer */
 232             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 233             itab_tmp         = _fjsp_dtox_v2r8(rt);
 234             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 235             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 236             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 237
 238             vfconv.i[0]     *= 8;
 239             vfconv.i[1]     *= 8;
 240
 241             /* COULOMB ELECTROSTATICS */
 242             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 243             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 244
 245             /* CUBIC SPLINE TABLE DISPERSION */
 246             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 247             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 248             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 249             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 250             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 251             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 252             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 253             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 254             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 255             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 256             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 257
 258             /* CUBIC SPLINE TABLE REPULSION */
 259             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 260             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 261             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 262             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 263             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 264             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 265             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 266             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 267             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 268             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 269             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 270             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 271             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 272
 273             /* Update potential sum for this i atom from the interaction with this j atom. */
 274             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 275             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 276
 277             fscal            = _fjsp_add_v2r8(felec,fvdw);
 278
 279             /* Update vectorial force */
 280             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 281             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 282             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 283
 284             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 285             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 286             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 287
 288             /**************************
 289              * CALCULATE INTERACTIONS *
 290              **************************/
 291
 292             /* Compute parameters for interactions between i and j atoms */
 293             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 294
 295             /* COULOMB ELECTROSTATICS */
 296             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 297             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 298
 299             /* Update potential sum for this i atom from the interaction with this j atom. */
 300             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 301
 302             fscal            = felec;
 303
 304             /* Update vectorial force */
 305             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 306             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 307             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 308
 309             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 310             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 311             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 312
 313             /**************************
 314              * CALCULATE INTERACTIONS *
 315              **************************/
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 319
 320             /* COULOMB ELECTROSTATICS */
 321             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 322             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 323
 324             /* Update potential sum for this i atom from the interaction with this j atom. */
 325             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 326
 327             fscal            = felec;
 328
 329             /* Update vectorial force */
 330             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 331             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 332             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 333
 334             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 335             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 336             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 337
 338             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 339
 340             /* Inner loop uses 131 flops */
 341         }
 342
 343         if(jidx<j_index_end)
 344         {
 345
 346             jnrA             = jjnr[jidx];
 347             j_coord_offsetA  = DIM*jnrA;
 348
 349             /* load j atom coordinates */
 350             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 351                                               &jx0,&jy0,&jz0);
 352
 353             /* Calculate displacement vector */
 354             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 355             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 356             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 357             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 358             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 359             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 360             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 361             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 362             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 363
 364             /* Calculate squared distance and things based on it */
 365             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 366             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 367             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 368
 369             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 370             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 371             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 372
 373             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 374             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 375             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 376
 377             /* Load parameters for j particles */
 378             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 379             vdwjidx0A        = 2*vdwtype[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             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 390
 391             /* Compute parameters for interactions between i and j atoms */
 392             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 393             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 394
 395             /* Calculate table index by multiplying r with table scale and truncate to integer */
 396             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 397             itab_tmp         = _fjsp_dtox_v2r8(rt);
 398             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 399             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 400             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 401
 402             vfconv.i[0]     *= 8;
 403             vfconv.i[1]     *= 8;
 404
 405             /* COULOMB ELECTROSTATICS */
 406             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 407             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 408
 409             /* CUBIC SPLINE TABLE DISPERSION */
 410             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 411             F                = _fjsp_setzero_v2r8();
 412             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 413             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 414             H                = _fjsp_setzero_v2r8();
 415             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 416             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 417             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 418             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 419             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 420             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 421
 422             /* CUBIC SPLINE TABLE REPULSION */
 423             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 424             F                = _fjsp_setzero_v2r8();
 425             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 426             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 427             H                = _fjsp_setzero_v2r8();
 428             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 429             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 430             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 431             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 432             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 433             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 434             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 435             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 436
 437             /* Update potential sum for this i atom from the interaction with this j atom. */
 438             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 439             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 440             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 441             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 442
 443             fscal            = _fjsp_add_v2r8(felec,fvdw);
 444
 445             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 446
 447             /* Update vectorial force */
 448             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 449             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 450             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 451
 452             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 453             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 454             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 455
 456             /**************************
 457              * CALCULATE INTERACTIONS *
 458              **************************/
 459
 460             /* Compute parameters for interactions between i and j atoms */
 461             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 462
 463             /* COULOMB ELECTROSTATICS */
 464             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 465             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 469             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 470
 471             fscal            = felec;
 472
 473             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 474
 475             /* Update vectorial force */
 476             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 477             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 478             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 479
 480             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 481             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 482             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* Compute parameters for interactions between i and j atoms */
 489             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 490
 491             /* COULOMB ELECTROSTATICS */
 492             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 493             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 497             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 498
 499             fscal            = felec;
 500
 501             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 502
 503             /* Update vectorial force */
 504             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 505             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 506             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 507
 508             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 509             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 510             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 511
 512             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 513
 514             /* Inner loop uses 131 flops */
 515         }
 516
 517         /* End of innermost loop */
 518
 519         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 520                                               f+i_coord_offset,fshift+i_shift_offset);
 521
 522         ggid                        = gid[iidx];
 523         /* Update potential energies */
 524         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 525         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 526
 527         /* Increment number of inner iterations */
 528         inneriter                  += j_index_end - j_index_start;
 529
 530         /* Outer loop uses 20 flops */
 531     }
 532
 533     /* Increment number of outer iterations */
 534     outeriter        += nri;
 535
 536     /* Update outer/inner flops */
 537
 538     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*131);
 539 }
 540 /*
 541  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 542  * Electrostatics interaction: Coulomb
 543  * VdW interaction:            CubicSplineTable
 544  * Geometry:                   Water3-Particle
 545  * Calculate force/pot:        Force
 546  */
 547 void
 548 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 549                     (t_nblist                    * gmx_restrict       nlist,
 550                      rvec                        * gmx_restrict          xx,
 551                      rvec                        * gmx_restrict          ff,
 552                      t_forcerec                  * gmx_restrict          fr,
 553                      t_mdatoms                   * gmx_restrict     mdatoms,
 554                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 555                      t_nrnb                      * gmx_restrict        nrnb)
 556 {
 557     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 558      * just 0 for non-waters.
 559      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 560      * jnr indices corresponding to data put in the four positions in the SIMD register.
 561      */
 562     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 563     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 564     int              jnrA,jnrB;
 565     int              j_coord_offsetA,j_coord_offsetB;
 566     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 567     real             rcutoff_scalar;
 568     real             *shiftvec,*fshift,*x,*f;
 569     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 570     int              vdwioffset0;
 571     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 572     int              vdwioffset1;
 573     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 574     int              vdwioffset2;
 575     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 576     int              vdwjidx0A,vdwjidx0B;
 577     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 578     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 579     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 580     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 581     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 582     real             *charge;
 583     int              nvdwtype;
 584     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 585     int              *vdwtype;
 586     real             *vdwparam;
 587     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 588     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 589     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 590     real             *vftab;
 591     _fjsp_v2r8       itab_tmp;
 592     _fjsp_v2r8       dummy_mask,cutoff_mask;
 593     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 594     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 595     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 596
 597     x                = xx[0];
 598     f                = ff[0];
 599
 600     nri              = nlist->nri;
 601     iinr             = nlist->iinr;
 602     jindex           = nlist->jindex;
 603     jjnr             = nlist->jjnr;
 604     shiftidx         = nlist->shift;
 605     gid              = nlist->gid;
 606     shiftvec         = fr->shift_vec[0];
 607     fshift           = fr->fshift[0];
 608     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 609     charge           = mdatoms->chargeA;
 610     nvdwtype         = fr->ntype;
 611     vdwparam         = fr->nbfp;
 612     vdwtype          = mdatoms->typeA;
 613
 614     vftab            = kernel_data->table_vdw->data;
 615     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 616
 617     /* Setup water-specific parameters */
 618     inr              = nlist->iinr[0];
 619     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 620     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 621     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 622     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 623
 624     /* Avoid stupid compiler warnings */
 625     jnrA = jnrB = 0;
 626     j_coord_offsetA = 0;
 627     j_coord_offsetB = 0;
 628
 629     outeriter        = 0;
 630     inneriter        = 0;
 631
 632     /* Start outer loop over neighborlists */
 633     for(iidx=0; iidx<nri; iidx++)
 634     {
 635         /* Load shift vector for this list */
 636         i_shift_offset   = DIM*shiftidx[iidx];
 637
 638         /* Load limits for loop over neighbors */
 639         j_index_start    = jindex[iidx];
 640         j_index_end      = jindex[iidx+1];
 641
 642         /* Get outer coordinate index */
 643         inr              = iinr[iidx];
 644         i_coord_offset   = DIM*inr;
 645
 646         /* Load i particle coords and add shift vector */
 647         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 648                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 649
 650         fix0             = _fjsp_setzero_v2r8();
 651         fiy0             = _fjsp_setzero_v2r8();
 652         fiz0             = _fjsp_setzero_v2r8();
 653         fix1             = _fjsp_setzero_v2r8();
 654         fiy1             = _fjsp_setzero_v2r8();
 655         fiz1             = _fjsp_setzero_v2r8();
 656         fix2             = _fjsp_setzero_v2r8();
 657         fiy2             = _fjsp_setzero_v2r8();
 658         fiz2             = _fjsp_setzero_v2r8();
 659
 660         /* Start inner kernel loop */
 661         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 662         {
 663
 664             /* Get j neighbor index, and coordinate index */
 665             jnrA             = jjnr[jidx];
 666             jnrB             = jjnr[jidx+1];
 667             j_coord_offsetA  = DIM*jnrA;
 668             j_coord_offsetB  = DIM*jnrB;
 669
 670             /* load j atom coordinates */
 671             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 672                                               &jx0,&jy0,&jz0);
 673
 674             /* Calculate displacement vector */
 675             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 676             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 677             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 678             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 679             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 680             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 681             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 682             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 683             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 684
 685             /* Calculate squared distance and things based on it */
 686             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 687             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 688             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 689
 690             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 691             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 692             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 693
 694             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 695             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 696             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 697
 698             /* Load parameters for j particles */
 699             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 700             vdwjidx0A        = 2*vdwtype[jnrA+0];
 701             vdwjidx0B        = 2*vdwtype[jnrB+0];
 702
 703             fjx0             = _fjsp_setzero_v2r8();
 704             fjy0             = _fjsp_setzero_v2r8();
 705             fjz0             = _fjsp_setzero_v2r8();
 706
 707             /**************************
 708              * CALCULATE INTERACTIONS *
 709              **************************/
 710
 711             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 712
 713             /* Compute parameters for interactions between i and j atoms */
 714             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 715             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 716                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 717
 718             /* Calculate table index by multiplying r with table scale and truncate to integer */
 719             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 720             itab_tmp         = _fjsp_dtox_v2r8(rt);
 721             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 722             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 723             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 724
 725             vfconv.i[0]     *= 8;
 726             vfconv.i[1]     *= 8;
 727
 728             /* COULOMB ELECTROSTATICS */
 729             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 730             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 731
 732             /* CUBIC SPLINE TABLE DISPERSION */
 733             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 734             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 735             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 736             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 737             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 738             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 739             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 740             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 741             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 742
 743             /* CUBIC SPLINE TABLE REPULSION */
 744             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 745             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 746             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 747             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 748             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 749             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 750             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 751             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 752             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 753             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 754
 755             fscal            = _fjsp_add_v2r8(felec,fvdw);
 756
 757             /* Update vectorial force */
 758             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 759             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 760             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 761
 762             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 763             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 764             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 765
 766             /**************************
 767              * CALCULATE INTERACTIONS *
 768              **************************/
 769
 770             /* Compute parameters for interactions between i and j atoms */
 771             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 772
 773             /* COULOMB ELECTROSTATICS */
 774             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 775             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 776
 777             fscal            = felec;
 778
 779             /* Update vectorial force */
 780             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 781             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 782             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 783
 784             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 785             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 786             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* Compute parameters for interactions between i and j atoms */
 793             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 794
 795             /* COULOMB ELECTROSTATICS */
 796             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 797             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 798
 799             fscal            = felec;
 800
 801             /* Update vectorial force */
 802             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 803             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 804             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 805
 806             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 807             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 808             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 809
 810             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 811
 812             /* Inner loop uses 120 flops */
 813         }
 814
 815         if(jidx<j_index_end)
 816         {
 817
 818             jnrA             = jjnr[jidx];
 819             j_coord_offsetA  = DIM*jnrA;
 820
 821             /* load j atom coordinates */
 822             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 823                                               &jx0,&jy0,&jz0);
 824
 825             /* Calculate displacement vector */
 826             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 827             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 828             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 829             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 830             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 831             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 832             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 833             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 834             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 835
 836             /* Calculate squared distance and things based on it */
 837             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 838             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 839             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 840
 841             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 842             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 843             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 844
 845             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 846             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 847             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 848
 849             /* Load parameters for j particles */
 850             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 851             vdwjidx0A        = 2*vdwtype[jnrA+0];
 852
 853             fjx0             = _fjsp_setzero_v2r8();
 854             fjy0             = _fjsp_setzero_v2r8();
 855             fjz0             = _fjsp_setzero_v2r8();
 856
 857             /**************************
 858              * CALCULATE INTERACTIONS *
 859              **************************/
 860
 861             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 862
 863             /* Compute parameters for interactions between i and j atoms */
 864             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 865             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 866
 867             /* Calculate table index by multiplying r with table scale and truncate to integer */
 868             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 869             itab_tmp         = _fjsp_dtox_v2r8(rt);
 870             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 871             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 872             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 873
 874             vfconv.i[0]     *= 8;
 875             vfconv.i[1]     *= 8;
 876
 877             /* COULOMB ELECTROSTATICS */
 878             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 879             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 880
 881             /* CUBIC SPLINE TABLE DISPERSION */
 882             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 883             F                = _fjsp_setzero_v2r8();
 884             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 885             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 886             H                = _fjsp_setzero_v2r8();
 887             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 888             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 889             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 890             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 891
 892             /* CUBIC SPLINE TABLE REPULSION */
 893             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 894             F                = _fjsp_setzero_v2r8();
 895             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 896             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 897             H                = _fjsp_setzero_v2r8();
 898             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 899             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 900             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 901             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 902             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 903
 904             fscal            = _fjsp_add_v2r8(felec,fvdw);
 905
 906             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 907
 908             /* Update vectorial force */
 909             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 910             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 911             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 912
 913             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 914             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 915             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 916
 917             /**************************
 918              * CALCULATE INTERACTIONS *
 919              **************************/
 920
 921             /* Compute parameters for interactions between i and j atoms */
 922             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 923
 924             /* COULOMB ELECTROSTATICS */
 925             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 926             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 927
 928             fscal            = felec;
 929
 930             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 931
 932             /* Update vectorial force */
 933             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 934             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 935             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 936
 937             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 938             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 939             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 940
 941             /**************************
 942              * CALCULATE INTERACTIONS *
 943              **************************/
 944
 945             /* Compute parameters for interactions between i and j atoms */
 946             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 947
 948             /* COULOMB ELECTROSTATICS */
 949             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 950             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 951
 952             fscal            = felec;
 953
 954             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 955
 956             /* Update vectorial force */
 957             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 958             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 959             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 960
 961             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 962             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 963             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 964
 965             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 966
 967             /* Inner loop uses 120 flops */
 968         }
 969
 970         /* End of innermost loop */
 971
 972         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 973                                               f+i_coord_offset,fshift+i_shift_offset);
 974
 975         /* Increment number of inner iterations */
 976         inneriter                  += j_index_end - j_index_start;
 977
 978         /* Outer loop uses 18 flops */
 979     }
 980
 981     /* Increment number of outer iterations */
 982     outeriter        += nri;
 983
 984     /* Update outer/inner flops */
 985
 986     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*120);
 987 }