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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Coulomb
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_vdw->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* Avoid stupid compiler warnings */
 136     jnrA = jnrB = 0;
 137     j_coord_offsetA = 0;
 138     j_coord_offsetB = 0;
 139
 140     outeriter        = 0;
 141     inneriter        = 0;
 142
 143     /* Start outer loop over neighborlists */
 144     for(iidx=0; iidx<nri; iidx++)
 145     {
 146         /* Load shift vector for this list */
 147         i_shift_offset   = DIM*shiftidx[iidx];
 148
 149         /* Load limits for loop over neighbors */
 150         j_index_start    = jindex[iidx];
 151         j_index_end      = jindex[iidx+1];
 152
 153         /* Get outer coordinate index */
 154         inr              = iinr[iidx];
 155         i_coord_offset   = DIM*inr;
 156
 157         /* Load i particle coords and add shift vector */
 158         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 159                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 160
 161         fix0             = _fjsp_setzero_v2r8();
 162         fiy0             = _fjsp_setzero_v2r8();
 163         fiz0             = _fjsp_setzero_v2r8();
 164         fix1             = _fjsp_setzero_v2r8();
 165         fiy1             = _fjsp_setzero_v2r8();
 166         fiz1             = _fjsp_setzero_v2r8();
 167         fix2             = _fjsp_setzero_v2r8();
 168         fiy2             = _fjsp_setzero_v2r8();
 169         fiz2             = _fjsp_setzero_v2r8();
 170
 171         /* Reset potential sums */
 172         velecsum         = _fjsp_setzero_v2r8();
 173         vvdwsum          = _fjsp_setzero_v2r8();
 174
 175         /* Start inner kernel loop */
 176         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 177         {
 178
 179             /* Get j neighbor index, and coordinate index */
 180             jnrA             = jjnr[jidx];
 181             jnrB             = jjnr[jidx+1];
 182             j_coord_offsetA  = DIM*jnrA;
 183             j_coord_offsetB  = DIM*jnrB;
 184
 185             /* load j atom coordinates */
 186             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 187                                               &jx0,&jy0,&jz0);
 188
 189             /* Calculate displacement vector */
 190             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 191             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 192             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 193             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 194             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 195             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 196             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 197             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 198             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 199
 200             /* Calculate squared distance and things based on it */
 201             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 202             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 203             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 204
 205             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 206             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 207             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 208
 209             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 210             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 211             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 212
 213             /* Load parameters for j particles */
 214             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 215             vdwjidx0A        = 2*vdwtype[jnrA+0];
 216             vdwjidx0B        = 2*vdwtype[jnrB+0];
 217
 218             fjx0             = _fjsp_setzero_v2r8();
 219             fjy0             = _fjsp_setzero_v2r8();
 220             fjz0             = _fjsp_setzero_v2r8();
 221
 222             /**************************
 223              * CALCULATE INTERACTIONS *
 224              **************************/
 225
 226             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 227
 228             /* Compute parameters for interactions between i and j atoms */
 229             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 230             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 231                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 232
 233             /* Calculate table index by multiplying r with table scale and truncate to integer */
 234             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 235             itab_tmp         = _fjsp_dtox_v2r8(rt);
 236             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 237             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 238             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 239
 240             vfconv.i[0]     *= 8;
 241             vfconv.i[1]     *= 8;
 242
 243             /* COULOMB ELECTROSTATICS */
 244             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 245             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 246
 247             /* CUBIC SPLINE TABLE DISPERSION */
 248             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 249             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 250             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 251             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 252             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 253             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 254             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 255             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 256             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 257             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 258             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 259
 260             /* CUBIC SPLINE TABLE REPULSION */
 261             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 262             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 263             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 264             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 265             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 266             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 267             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 268             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 269             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 270             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 271             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 272             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 273             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 274
 275             /* Update potential sum for this i atom from the interaction with this j atom. */
 276             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 277             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 278
 279             fscal            = _fjsp_add_v2r8(felec,fvdw);
 280
 281             /* Update vectorial force */
 282             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 283             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 284             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 285
 286             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 287             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 288             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 289
 290             /**************************
 291              * CALCULATE INTERACTIONS *
 292              **************************/
 293
 294             /* Compute parameters for interactions between i and j atoms */
 295             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 296
 297             /* COULOMB ELECTROSTATICS */
 298             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 299             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 300
 301             /* Update potential sum for this i atom from the interaction with this j atom. */
 302             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 303
 304             fscal            = felec;
 305
 306             /* Update vectorial force */
 307             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 308             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 309             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 310
 311             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 312             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 313             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 314
 315             /**************************
 316              * CALCULATE INTERACTIONS *
 317              **************************/
 318
 319             /* Compute parameters for interactions between i and j atoms */
 320             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 321
 322             /* COULOMB ELECTROSTATICS */
 323             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 324             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 325
 326             /* Update potential sum for this i atom from the interaction with this j atom. */
 327             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 328
 329             fscal            = felec;
 330
 331             /* Update vectorial force */
 332             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 333             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 334             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 335
 336             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 337             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 338             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 339
 340             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 341
 342             /* Inner loop uses 131 flops */
 343         }
 344
 345         if(jidx<j_index_end)
 346         {
 347
 348             jnrA             = jjnr[jidx];
 349             j_coord_offsetA  = DIM*jnrA;
 350
 351             /* load j atom coordinates */
 352             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 353                                               &jx0,&jy0,&jz0);
 354
 355             /* Calculate displacement vector */
 356             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 357             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 358             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 359             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 360             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 361             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 362             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 363             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 364             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 365
 366             /* Calculate squared distance and things based on it */
 367             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 368             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 369             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 370
 371             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 372             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 373             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 374
 375             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 376             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 377             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 378
 379             /* Load parameters for j particles */
 380             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 381             vdwjidx0A        = 2*vdwtype[jnrA+0];
 382
 383             fjx0             = _fjsp_setzero_v2r8();
 384             fjy0             = _fjsp_setzero_v2r8();
 385             fjz0             = _fjsp_setzero_v2r8();
 386
 387             /**************************
 388              * CALCULATE INTERACTIONS *
 389              **************************/
 390
 391             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 392
 393             /* Compute parameters for interactions between i and j atoms */
 394             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 395             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 396                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 397
 398             /* Calculate table index by multiplying r with table scale and truncate to integer */
 399             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 400             itab_tmp         = _fjsp_dtox_v2r8(rt);
 401             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 402             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 403             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 404
 405             vfconv.i[0]     *= 8;
 406             vfconv.i[1]     *= 8;
 407
 408             /* COULOMB ELECTROSTATICS */
 409             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 410             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 411
 412             /* CUBIC SPLINE TABLE DISPERSION */
 413             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 414             F                = _fjsp_setzero_v2r8();
 415             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 416             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 417             H                = _fjsp_setzero_v2r8();
 418             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 419             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 420             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 421             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 422             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 423             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 424
 425             /* CUBIC SPLINE TABLE REPULSION */
 426             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 427             F                = _fjsp_setzero_v2r8();
 428             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 429             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 430             H                = _fjsp_setzero_v2r8();
 431             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 432             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 433             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 434             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 435             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 436             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 437             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 438             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 442             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 443             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 444             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 445
 446             fscal            = _fjsp_add_v2r8(felec,fvdw);
 447
 448             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 449
 450             /* Update vectorial force */
 451             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 452             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 453             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 454
 455             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 456             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 457             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* Compute parameters for interactions between i and j atoms */
 464             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 465
 466             /* COULOMB ELECTROSTATICS */
 467             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 468             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 472             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 477
 478             /* Update vectorial force */
 479             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 480             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 481             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 482
 483             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 484             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 485             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             /* Compute parameters for interactions between i and j atoms */
 492             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 493
 494             /* COULOMB ELECTROSTATICS */
 495             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 496             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 500             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 501
 502             fscal            = felec;
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 508             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 509             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 514
 515             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 516
 517             /* Inner loop uses 131 flops */
 518         }
 519
 520         /* End of innermost loop */
 521
 522         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 523                                               f+i_coord_offset,fshift+i_shift_offset);
 524
 525         ggid                        = gid[iidx];
 526         /* Update potential energies */
 527         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 528         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 529
 530         /* Increment number of inner iterations */
 531         inneriter                  += j_index_end - j_index_start;
 532
 533         /* Outer loop uses 20 flops */
 534     }
 535
 536     /* Increment number of outer iterations */
 537     outeriter        += nri;
 538
 539     /* Update outer/inner flops */
 540
 541     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*131);
 542 }
 543 /*
 544  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 545  * Electrostatics interaction: Coulomb
 546  * VdW interaction:            CubicSplineTable
 547  * Geometry:                   Water3-Particle
 548  * Calculate force/pot:        Force
 549  */
 550 void
 551 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 552                     (t_nblist                    * gmx_restrict       nlist,
 553                      rvec                        * gmx_restrict          xx,
 554                      rvec                        * gmx_restrict          ff,
 555                      t_forcerec                  * gmx_restrict          fr,
 556                      t_mdatoms                   * gmx_restrict     mdatoms,
 557                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 558                      t_nrnb                      * gmx_restrict        nrnb)
 559 {
 560     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 561      * just 0 for non-waters.
 562      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 563      * jnr indices corresponding to data put in the four positions in the SIMD register.
 564      */
 565     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 566     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 567     int              jnrA,jnrB;
 568     int              j_coord_offsetA,j_coord_offsetB;
 569     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 570     real             rcutoff_scalar;
 571     real             *shiftvec,*fshift,*x,*f;
 572     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 573     int              vdwioffset0;
 574     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 575     int              vdwioffset1;
 576     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 577     int              vdwioffset2;
 578     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 579     int              vdwjidx0A,vdwjidx0B;
 580     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 581     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 582     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 583     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 584     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 585     real             *charge;
 586     int              nvdwtype;
 587     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 588     int              *vdwtype;
 589     real             *vdwparam;
 590     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 591     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 592     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 593     real             *vftab;
 594     _fjsp_v2r8       itab_tmp;
 595     _fjsp_v2r8       dummy_mask,cutoff_mask;
 596     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 597     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 598     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 599
 600     x                = xx[0];
 601     f                = ff[0];
 602
 603     nri              = nlist->nri;
 604     iinr             = nlist->iinr;
 605     jindex           = nlist->jindex;
 606     jjnr             = nlist->jjnr;
 607     shiftidx         = nlist->shift;
 608     gid              = nlist->gid;
 609     shiftvec         = fr->shift_vec[0];
 610     fshift           = fr->fshift[0];
 611     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 612     charge           = mdatoms->chargeA;
 613     nvdwtype         = fr->ntype;
 614     vdwparam         = fr->nbfp;
 615     vdwtype          = mdatoms->typeA;
 616
 617     vftab            = kernel_data->table_vdw->data;
 618     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 619
 620     /* Setup water-specific parameters */
 621     inr              = nlist->iinr[0];
 622     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 623     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 624     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 625     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 626
 627     /* Avoid stupid compiler warnings */
 628     jnrA = jnrB = 0;
 629     j_coord_offsetA = 0;
 630     j_coord_offsetB = 0;
 631
 632     outeriter        = 0;
 633     inneriter        = 0;
 634
 635     /* Start outer loop over neighborlists */
 636     for(iidx=0; iidx<nri; iidx++)
 637     {
 638         /* Load shift vector for this list */
 639         i_shift_offset   = DIM*shiftidx[iidx];
 640
 641         /* Load limits for loop over neighbors */
 642         j_index_start    = jindex[iidx];
 643         j_index_end      = jindex[iidx+1];
 644
 645         /* Get outer coordinate index */
 646         inr              = iinr[iidx];
 647         i_coord_offset   = DIM*inr;
 648
 649         /* Load i particle coords and add shift vector */
 650         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 651                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 652
 653         fix0             = _fjsp_setzero_v2r8();
 654         fiy0             = _fjsp_setzero_v2r8();
 655         fiz0             = _fjsp_setzero_v2r8();
 656         fix1             = _fjsp_setzero_v2r8();
 657         fiy1             = _fjsp_setzero_v2r8();
 658         fiz1             = _fjsp_setzero_v2r8();
 659         fix2             = _fjsp_setzero_v2r8();
 660         fiy2             = _fjsp_setzero_v2r8();
 661         fiz2             = _fjsp_setzero_v2r8();
 662
 663         /* Start inner kernel loop */
 664         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 665         {
 666
 667             /* Get j neighbor index, and coordinate index */
 668             jnrA             = jjnr[jidx];
 669             jnrB             = jjnr[jidx+1];
 670             j_coord_offsetA  = DIM*jnrA;
 671             j_coord_offsetB  = DIM*jnrB;
 672
 673             /* load j atom coordinates */
 674             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 675                                               &jx0,&jy0,&jz0);
 676
 677             /* Calculate displacement vector */
 678             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 679             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 680             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 681             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 682             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 683             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 684             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 685             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 686             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 687
 688             /* Calculate squared distance and things based on it */
 689             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 690             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 691             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 692
 693             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 694             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 695             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 696
 697             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 698             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 699             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 700
 701             /* Load parameters for j particles */
 702             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 703             vdwjidx0A        = 2*vdwtype[jnrA+0];
 704             vdwjidx0B        = 2*vdwtype[jnrB+0];
 705
 706             fjx0             = _fjsp_setzero_v2r8();
 707             fjy0             = _fjsp_setzero_v2r8();
 708             fjz0             = _fjsp_setzero_v2r8();
 709
 710             /**************************
 711              * CALCULATE INTERACTIONS *
 712              **************************/
 713
 714             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 715
 716             /* Compute parameters for interactions between i and j atoms */
 717             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 718             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 719                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 720
 721             /* Calculate table index by multiplying r with table scale and truncate to integer */
 722             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 723             itab_tmp         = _fjsp_dtox_v2r8(rt);
 724             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 725             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 726             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 727
 728             vfconv.i[0]     *= 8;
 729             vfconv.i[1]     *= 8;
 730
 731             /* COULOMB ELECTROSTATICS */
 732             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 733             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 734
 735             /* CUBIC SPLINE TABLE DISPERSION */
 736             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 737             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 738             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 739             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 740             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 741             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 742             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 743             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 744             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 745
 746             /* CUBIC SPLINE TABLE REPULSION */
 747             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 748             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 749             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 750             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 751             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 752             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 753             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 754             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 755             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 756             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 757
 758             fscal            = _fjsp_add_v2r8(felec,fvdw);
 759
 760             /* Update vectorial force */
 761             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 762             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 763             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 764
 765             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 766             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 767             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 768
 769             /**************************
 770              * CALCULATE INTERACTIONS *
 771              **************************/
 772
 773             /* Compute parameters for interactions between i and j atoms */
 774             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 775
 776             /* COULOMB ELECTROSTATICS */
 777             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 778             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 779
 780             fscal            = felec;
 781
 782             /* Update vectorial force */
 783             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 784             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 785             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 786
 787             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 788             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 789             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             /* Compute parameters for interactions between i and j atoms */
 796             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 797
 798             /* COULOMB ELECTROSTATICS */
 799             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 800             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 801
 802             fscal            = felec;
 803
 804             /* Update vectorial force */
 805             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 806             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 807             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 808
 809             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 810             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 811             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 812
 813             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 814
 815             /* Inner loop uses 120 flops */
 816         }
 817
 818         if(jidx<j_index_end)
 819         {
 820
 821             jnrA             = jjnr[jidx];
 822             j_coord_offsetA  = DIM*jnrA;
 823
 824             /* load j atom coordinates */
 825             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 826                                               &jx0,&jy0,&jz0);
 827
 828             /* Calculate displacement vector */
 829             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 830             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 831             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 832             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 833             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 834             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 835             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 836             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 837             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 838
 839             /* Calculate squared distance and things based on it */
 840             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 841             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 842             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 843
 844             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 845             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 846             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 847
 848             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 849             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 850             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 851
 852             /* Load parameters for j particles */
 853             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 854             vdwjidx0A        = 2*vdwtype[jnrA+0];
 855
 856             fjx0             = _fjsp_setzero_v2r8();
 857             fjy0             = _fjsp_setzero_v2r8();
 858             fjz0             = _fjsp_setzero_v2r8();
 859
 860             /**************************
 861              * CALCULATE INTERACTIONS *
 862              **************************/
 863
 864             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 865
 866             /* Compute parameters for interactions between i and j atoms */
 867             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 868             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 869                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 870
 871             /* Calculate table index by multiplying r with table scale and truncate to integer */
 872             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 873             itab_tmp         = _fjsp_dtox_v2r8(rt);
 874             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 875             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 876             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 877
 878             vfconv.i[0]     *= 8;
 879             vfconv.i[1]     *= 8;
 880
 881             /* COULOMB ELECTROSTATICS */
 882             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 883             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 884
 885             /* CUBIC SPLINE TABLE DISPERSION */
 886             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 887             F                = _fjsp_setzero_v2r8();
 888             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 889             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 890             H                = _fjsp_setzero_v2r8();
 891             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 892             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 893             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 894             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 895
 896             /* CUBIC SPLINE TABLE REPULSION */
 897             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 898             F                = _fjsp_setzero_v2r8();
 899             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 900             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 901             H                = _fjsp_setzero_v2r8();
 902             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 903             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 904             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 905             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 906             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 907
 908             fscal            = _fjsp_add_v2r8(felec,fvdw);
 909
 910             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 911
 912             /* Update vectorial force */
 913             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 914             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 915             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 916
 917             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 918             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 919             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             /* Compute parameters for interactions between i and j atoms */
 926             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 927
 928             /* COULOMB ELECTROSTATICS */
 929             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 930             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 931
 932             fscal            = felec;
 933
 934             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 935
 936             /* Update vectorial force */
 937             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 938             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 939             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 940
 941             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 942             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 943             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 944
 945             /**************************
 946              * CALCULATE INTERACTIONS *
 947              **************************/
 948
 949             /* Compute parameters for interactions between i and j atoms */
 950             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 951
 952             /* COULOMB ELECTROSTATICS */
 953             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 954             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 955
 956             fscal            = felec;
 957
 958             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 959
 960             /* Update vectorial force */
 961             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 962             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 963             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 964
 965             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 966             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 967             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 968
 969             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 970
 971             /* Inner loop uses 120 flops */
 972         }
 973
 974         /* End of innermost loop */
 975
 976         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 977                                               f+i_coord_offset,fshift+i_shift_offset);
 978
 979         /* Increment number of inner iterations */
 980         inneriter                  += j_index_end - j_index_start;
 981
 982         /* Outer loop uses 18 flops */
 983     }
 984
 985     /* Increment number of outer iterations */
 986     outeriter        += nri;
 987
 988     /* Update outer/inner flops */
 989
 990     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*120);
 991 }