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, 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 "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_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 396
 397             /* Calculate table index by multiplying r with table scale and truncate to integer */
 398             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 399             itab_tmp         = _fjsp_dtox_v2r8(rt);
 400             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 401             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 402             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 403
 404             vfconv.i[0]     *= 8;
 405             vfconv.i[1]     *= 8;
 406
 407             /* COULOMB ELECTROSTATICS */
 408             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 409             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 410
 411             /* CUBIC SPLINE TABLE DISPERSION */
 412             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 413             F                = _fjsp_setzero_v2r8();
 414             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 415             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 416             H                = _fjsp_setzero_v2r8();
 417             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 418             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 419             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 420             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 421             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 422             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 423
 424             /* CUBIC SPLINE TABLE REPULSION */
 425             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 426             F                = _fjsp_setzero_v2r8();
 427             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 428             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 429             H                = _fjsp_setzero_v2r8();
 430             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 431             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 432             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 433             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 434             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 435             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 436             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 437             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 438
 439             /* Update potential sum for this i atom from the interaction with this j atom. */
 440             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 441             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 442             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 443             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 444
 445             fscal            = _fjsp_add_v2r8(felec,fvdw);
 446
 447             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 448
 449             /* Update vectorial force */
 450             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 451             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 452             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 453
 454             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 455             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 456             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             /* Compute parameters for interactions between i and j atoms */
 463             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 464
 465             /* COULOMB ELECTROSTATICS */
 466             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 467             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 468
 469             /* Update potential sum for this i atom from the interaction with this j atom. */
 470             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 471             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 472
 473             fscal            = felec;
 474
 475             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 476
 477             /* Update vectorial force */
 478             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 479             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 480             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 481
 482             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 483             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 484             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             /* Compute parameters for interactions between i and j atoms */
 491             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 492
 493             /* COULOMB ELECTROSTATICS */
 494             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 495             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 496
 497             /* Update potential sum for this i atom from the interaction with this j atom. */
 498             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 499             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 504
 505             /* Update vectorial force */
 506             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 507             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 508             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 509
 510             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 511             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 512             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 513
 514             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 515
 516             /* Inner loop uses 131 flops */
 517         }
 518
 519         /* End of innermost loop */
 520
 521         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 522                                               f+i_coord_offset,fshift+i_shift_offset);
 523
 524         ggid                        = gid[iidx];
 525         /* Update potential energies */
 526         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 527         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 528
 529         /* Increment number of inner iterations */
 530         inneriter                  += j_index_end - j_index_start;
 531
 532         /* Outer loop uses 20 flops */
 533     }
 534
 535     /* Increment number of outer iterations */
 536     outeriter        += nri;
 537
 538     /* Update outer/inner flops */
 539
 540     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*131);
 541 }
 542 /*
 543  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 544  * Electrostatics interaction: Coulomb
 545  * VdW interaction:            CubicSplineTable
 546  * Geometry:                   Water3-Particle
 547  * Calculate force/pot:        Force
 548  */
 549 void
 550 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 551                     (t_nblist                    * gmx_restrict       nlist,
 552                      rvec                        * gmx_restrict          xx,
 553                      rvec                        * gmx_restrict          ff,
 554                      t_forcerec                  * gmx_restrict          fr,
 555                      t_mdatoms                   * gmx_restrict     mdatoms,
 556                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 557                      t_nrnb                      * gmx_restrict        nrnb)
 558 {
 559     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 560      * just 0 for non-waters.
 561      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 562      * jnr indices corresponding to data put in the four positions in the SIMD register.
 563      */
 564     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 565     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 566     int              jnrA,jnrB;
 567     int              j_coord_offsetA,j_coord_offsetB;
 568     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 569     real             rcutoff_scalar;
 570     real             *shiftvec,*fshift,*x,*f;
 571     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 572     int              vdwioffset0;
 573     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 574     int              vdwioffset1;
 575     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 576     int              vdwioffset2;
 577     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 578     int              vdwjidx0A,vdwjidx0B;
 579     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 580     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 581     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 582     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 583     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 584     real             *charge;
 585     int              nvdwtype;
 586     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 587     int              *vdwtype;
 588     real             *vdwparam;
 589     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 590     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 591     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 592     real             *vftab;
 593     _fjsp_v2r8       itab_tmp;
 594     _fjsp_v2r8       dummy_mask,cutoff_mask;
 595     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 596     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 597     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 598
 599     x                = xx[0];
 600     f                = ff[0];
 601
 602     nri              = nlist->nri;
 603     iinr             = nlist->iinr;
 604     jindex           = nlist->jindex;
 605     jjnr             = nlist->jjnr;
 606     shiftidx         = nlist->shift;
 607     gid              = nlist->gid;
 608     shiftvec         = fr->shift_vec[0];
 609     fshift           = fr->fshift[0];
 610     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 611     charge           = mdatoms->chargeA;
 612     nvdwtype         = fr->ntype;
 613     vdwparam         = fr->nbfp;
 614     vdwtype          = mdatoms->typeA;
 615
 616     vftab            = kernel_data->table_vdw->data;
 617     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 618
 619     /* Setup water-specific parameters */
 620     inr              = nlist->iinr[0];
 621     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 622     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 623     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 624     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 625
 626     /* Avoid stupid compiler warnings */
 627     jnrA = jnrB = 0;
 628     j_coord_offsetA = 0;
 629     j_coord_offsetB = 0;
 630
 631     outeriter        = 0;
 632     inneriter        = 0;
 633
 634     /* Start outer loop over neighborlists */
 635     for(iidx=0; iidx<nri; iidx++)
 636     {
 637         /* Load shift vector for this list */
 638         i_shift_offset   = DIM*shiftidx[iidx];
 639
 640         /* Load limits for loop over neighbors */
 641         j_index_start    = jindex[iidx];
 642         j_index_end      = jindex[iidx+1];
 643
 644         /* Get outer coordinate index */
 645         inr              = iinr[iidx];
 646         i_coord_offset   = DIM*inr;
 647
 648         /* Load i particle coords and add shift vector */
 649         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 650                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 651
 652         fix0             = _fjsp_setzero_v2r8();
 653         fiy0             = _fjsp_setzero_v2r8();
 654         fiz0             = _fjsp_setzero_v2r8();
 655         fix1             = _fjsp_setzero_v2r8();
 656         fiy1             = _fjsp_setzero_v2r8();
 657         fiz1             = _fjsp_setzero_v2r8();
 658         fix2             = _fjsp_setzero_v2r8();
 659         fiy2             = _fjsp_setzero_v2r8();
 660         fiz2             = _fjsp_setzero_v2r8();
 661
 662         /* Start inner kernel loop */
 663         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 664         {
 665
 666             /* Get j neighbor index, and coordinate index */
 667             jnrA             = jjnr[jidx];
 668             jnrB             = jjnr[jidx+1];
 669             j_coord_offsetA  = DIM*jnrA;
 670             j_coord_offsetB  = DIM*jnrB;
 671
 672             /* load j atom coordinates */
 673             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 674                                               &jx0,&jy0,&jz0);
 675
 676             /* Calculate displacement vector */
 677             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 678             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 679             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 680             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 681             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 682             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 683             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 684             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 685             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 686
 687             /* Calculate squared distance and things based on it */
 688             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 689             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 690             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 691
 692             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 693             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 694             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 695
 696             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 697             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 698             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 699
 700             /* Load parameters for j particles */
 701             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 702             vdwjidx0A        = 2*vdwtype[jnrA+0];
 703             vdwjidx0B        = 2*vdwtype[jnrB+0];
 704
 705             fjx0             = _fjsp_setzero_v2r8();
 706             fjy0             = _fjsp_setzero_v2r8();
 707             fjz0             = _fjsp_setzero_v2r8();
 708
 709             /**************************
 710              * CALCULATE INTERACTIONS *
 711              **************************/
 712
 713             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 714
 715             /* Compute parameters for interactions between i and j atoms */
 716             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 717             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 718                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 719
 720             /* Calculate table index by multiplying r with table scale and truncate to integer */
 721             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 722             itab_tmp         = _fjsp_dtox_v2r8(rt);
 723             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 724             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 725             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 726
 727             vfconv.i[0]     *= 8;
 728             vfconv.i[1]     *= 8;
 729
 730             /* COULOMB ELECTROSTATICS */
 731             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 732             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 733
 734             /* CUBIC SPLINE TABLE DISPERSION */
 735             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 736             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 737             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 738             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 739             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 740             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 741             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 742             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 743             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 744
 745             /* CUBIC SPLINE TABLE REPULSION */
 746             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 747             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 748             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 749             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 750             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 751             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 752             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 753             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 754             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 755             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 756
 757             fscal            = _fjsp_add_v2r8(felec,fvdw);
 758
 759             /* Update vectorial force */
 760             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 761             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 762             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 763
 764             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 765             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 766             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 767
 768             /**************************
 769              * CALCULATE INTERACTIONS *
 770              **************************/
 771
 772             /* Compute parameters for interactions between i and j atoms */
 773             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 774
 775             /* COULOMB ELECTROSTATICS */
 776             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 777             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 778
 779             fscal            = felec;
 780
 781             /* Update vectorial force */
 782             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 783             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 784             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 785
 786             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 787             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 788             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 789
 790             /**************************
 791              * CALCULATE INTERACTIONS *
 792              **************************/
 793
 794             /* Compute parameters for interactions between i and j atoms */
 795             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 796
 797             /* COULOMB ELECTROSTATICS */
 798             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 799             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 800
 801             fscal            = felec;
 802
 803             /* Update vectorial force */
 804             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 805             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 806             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 807
 808             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 809             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 810             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 811
 812             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 813
 814             /* Inner loop uses 120 flops */
 815         }
 816
 817         if(jidx<j_index_end)
 818         {
 819
 820             jnrA             = jjnr[jidx];
 821             j_coord_offsetA  = DIM*jnrA;
 822
 823             /* load j atom coordinates */
 824             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 825                                               &jx0,&jy0,&jz0);
 826
 827             /* Calculate displacement vector */
 828             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 829             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 830             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 831             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 832             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 833             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 834             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 835             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 836             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 837
 838             /* Calculate squared distance and things based on it */
 839             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 840             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 841             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 842
 843             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 844             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 845             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 846
 847             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 848             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 849             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 850
 851             /* Load parameters for j particles */
 852             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 853             vdwjidx0A        = 2*vdwtype[jnrA+0];
 854
 855             fjx0             = _fjsp_setzero_v2r8();
 856             fjy0             = _fjsp_setzero_v2r8();
 857             fjz0             = _fjsp_setzero_v2r8();
 858
 859             /**************************
 860              * CALCULATE INTERACTIONS *
 861              **************************/
 862
 863             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 864
 865             /* Compute parameters for interactions between i and j atoms */
 866             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 867             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 868
 869             /* Calculate table index by multiplying r with table scale and truncate to integer */
 870             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 871             itab_tmp         = _fjsp_dtox_v2r8(rt);
 872             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 873             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 874             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 875
 876             vfconv.i[0]     *= 8;
 877             vfconv.i[1]     *= 8;
 878
 879             /* COULOMB ELECTROSTATICS */
 880             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 881             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 882
 883             /* CUBIC SPLINE TABLE DISPERSION */
 884             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 885             F                = _fjsp_setzero_v2r8();
 886             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 887             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 888             H                = _fjsp_setzero_v2r8();
 889             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 890             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 891             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 892             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 893
 894             /* CUBIC SPLINE TABLE REPULSION */
 895             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 896             F                = _fjsp_setzero_v2r8();
 897             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 898             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 899             H                = _fjsp_setzero_v2r8();
 900             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 901             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 902             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 903             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 904             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 905
 906             fscal            = _fjsp_add_v2r8(felec,fvdw);
 907
 908             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 909
 910             /* Update vectorial force */
 911             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 912             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 913             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 914
 915             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 916             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 917             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 918
 919             /**************************
 920              * CALCULATE INTERACTIONS *
 921              **************************/
 922
 923             /* Compute parameters for interactions between i and j atoms */
 924             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 925
 926             /* COULOMB ELECTROSTATICS */
 927             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 928             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 929
 930             fscal            = felec;
 931
 932             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 933
 934             /* Update vectorial force */
 935             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 936             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 937             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 938
 939             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 940             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 941             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 942
 943             /**************************
 944              * CALCULATE INTERACTIONS *
 945              **************************/
 946
 947             /* Compute parameters for interactions between i and j atoms */
 948             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 949
 950             /* COULOMB ELECTROSTATICS */
 951             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 952             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 953
 954             fscal            = felec;
 955
 956             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 957
 958             /* Update vectorial force */
 959             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 960             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 961             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 962
 963             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 964             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 965             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 966
 967             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 968
 969             /* Inner loop uses 120 flops */
 970         }
 971
 972         /* End of innermost loop */
 973
 974         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 975                                               f+i_coord_offset,fshift+i_shift_offset);
 976
 977         /* Increment number of inner iterations */
 978         inneriter                  += j_index_end - j_index_start;
 979
 980         /* Outer loop uses 18 flops */
 981     }
 982
 983     /* Increment number of outer iterations */
 984     outeriter        += nri;
 985
 986     /* Update outer/inner flops */
 987
 988     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*120);
 989 }