src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_vdw->data;
 124     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     /* Start outer loop over neighborlists */
 142     for(iidx=0; iidx<nri; iidx++)
 143     {
 144         /* Load shift vector for this list */
 145         i_shift_offset   = DIM*shiftidx[iidx];
 146
 147         /* Load limits for loop over neighbors */
 148         j_index_start    = jindex[iidx];
 149         j_index_end      = jindex[iidx+1];
 150
 151         /* Get outer coordinate index */
 152         inr              = iinr[iidx];
 153         i_coord_offset   = DIM*inr;
 154
 155         /* Load i particle coords and add shift vector */
 156         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 157                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 158
 159         fix0             = _fjsp_setzero_v2r8();
 160         fiy0             = _fjsp_setzero_v2r8();
 161         fiz0             = _fjsp_setzero_v2r8();
 162         fix1             = _fjsp_setzero_v2r8();
 163         fiy1             = _fjsp_setzero_v2r8();
 164         fiz1             = _fjsp_setzero_v2r8();
 165         fix2             = _fjsp_setzero_v2r8();
 166         fiy2             = _fjsp_setzero_v2r8();
 167         fiz2             = _fjsp_setzero_v2r8();
 168
 169         /* Reset potential sums */
 170         velecsum         = _fjsp_setzero_v2r8();
 171         vvdwsum          = _fjsp_setzero_v2r8();
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 175         {
 176
 177             /* Get j neighbor index, and coordinate index */
 178             jnrA             = jjnr[jidx];
 179             jnrB             = jjnr[jidx+1];
 180             j_coord_offsetA  = DIM*jnrA;
 181             j_coord_offsetB  = DIM*jnrB;
 182
 183             /* load j atom coordinates */
 184             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                               &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 189             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 190             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 191             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 192             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 193             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 194             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 195             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 196             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 197
 198             /* Calculate squared distance and things based on it */
 199             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 200             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 201             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 202
 203             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 204             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 205             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 206
 207             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 208             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 209             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 210
 211             /* Load parameters for j particles */
 212             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 213             vdwjidx0A        = 2*vdwtype[jnrA+0];
 214             vdwjidx0B        = 2*vdwtype[jnrB+0];
 215
 216             fjx0             = _fjsp_setzero_v2r8();
 217             fjy0             = _fjsp_setzero_v2r8();
 218             fjz0             = _fjsp_setzero_v2r8();
 219
 220             /**************************
 221              * CALCULATE INTERACTIONS *
 222              **************************/
 223
 224             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 225
 226             /* Compute parameters for interactions between i and j atoms */
 227             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 228             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 229                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 230
 231             /* Calculate table index by multiplying r with table scale and truncate to integer */
 232             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 233             itab_tmp         = _fjsp_dtox_v2r8(rt);
 234             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 235             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 236             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 237
 238             vfconv.i[0]     *= 8;
 239             vfconv.i[1]     *= 8;
 240
 241             /* COULOMB ELECTROSTATICS */
 242             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 243             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 244
 245             /* CUBIC SPLINE TABLE DISPERSION */
 246             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 247             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 248             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 249             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 250             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 251             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 252             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 253             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 254             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 255             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 256             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 257
 258             /* CUBIC SPLINE TABLE REPULSION */
 259             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 260             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 261             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 262             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 263             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 264             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 265             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 266             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 267             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 268             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 269             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 270             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 271             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 272
 273             /* Update potential sum for this i atom from the interaction with this j atom. */
 274             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 275             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 276
 277             fscal            = _fjsp_add_v2r8(felec,fvdw);
 278
 279             /* Update vectorial force */
 280             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 281             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 282             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 283
 284             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 285             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 286             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 287
 288             /**************************
 289              * CALCULATE INTERACTIONS *
 290              **************************/
 291
 292             /* Compute parameters for interactions between i and j atoms */
 293             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 294
 295             /* COULOMB ELECTROSTATICS */
 296             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 297             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 298
 299             /* Update potential sum for this i atom from the interaction with this j atom. */
 300             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 301
 302             fscal            = felec;
 303
 304             /* Update vectorial force */
 305             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 306             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 307             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 308
 309             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 310             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 311             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 312
 313             /**************************
 314              * CALCULATE INTERACTIONS *
 315              **************************/
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 319
 320             /* COULOMB ELECTROSTATICS */
 321             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 322             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 323
 324             /* Update potential sum for this i atom from the interaction with this j atom. */
 325             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 326
 327             fscal            = felec;
 328
 329             /* Update vectorial force */
 330             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 331             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 332             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 333
 334             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 335             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 336             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 337
 338             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 339
 340             /* Inner loop uses 131 flops */
 341         }
 342
 343         if(jidx<j_index_end)
 344         {
 345
 346             jnrA             = jjnr[jidx];
 347             j_coord_offsetA  = DIM*jnrA;
 348
 349             /* load j atom coordinates */
 350             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 351                                               &jx0,&jy0,&jz0);
 352
 353             /* Calculate displacement vector */
 354             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 355             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 356             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 357             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 358             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 359             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 360             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 361             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 362             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 363
 364             /* Calculate squared distance and things based on it */
 365             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 366             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 367             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 368
 369             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 370             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 371             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 372
 373             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 374             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 375             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 376
 377             /* Load parameters for j particles */
 378             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 379             vdwjidx0A        = 2*vdwtype[jnrA+0];
 380
 381             fjx0             = _fjsp_setzero_v2r8();
 382             fjy0             = _fjsp_setzero_v2r8();
 383             fjz0             = _fjsp_setzero_v2r8();
 384
 385             /**************************
 386              * CALCULATE INTERACTIONS *
 387              **************************/
 388
 389             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 390
 391             /* Compute parameters for interactions between i and j atoms */
 392             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 393             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 394                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 395
 396             /* Calculate table index by multiplying r with table scale and truncate to integer */
 397             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 398             itab_tmp         = _fjsp_dtox_v2r8(rt);
 399             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 400             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 401             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 402
 403             vfconv.i[0]     *= 8;
 404             vfconv.i[1]     *= 8;
 405
 406             /* COULOMB ELECTROSTATICS */
 407             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 408             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 409
 410             /* CUBIC SPLINE TABLE DISPERSION */
 411             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 412             F                = _fjsp_setzero_v2r8();
 413             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 414             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 415             H                = _fjsp_setzero_v2r8();
 416             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 417             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 418             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 419             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 420             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 421             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 422
 423             /* CUBIC SPLINE TABLE REPULSION */
 424             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 425             F                = _fjsp_setzero_v2r8();
 426             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 427             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 428             H                = _fjsp_setzero_v2r8();
 429             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 430             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 431             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 432             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 433             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 434             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 435             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 436             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 437
 438             /* Update potential sum for this i atom from the interaction with this j atom. */
 439             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 440             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 441             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 442             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 443
 444             fscal            = _fjsp_add_v2r8(felec,fvdw);
 445
 446             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 447
 448             /* Update vectorial force */
 449             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 450             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 451             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 452
 453             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 454             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 455             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             /* Compute parameters for interactions between i and j atoms */
 462             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 463
 464             /* COULOMB ELECTROSTATICS */
 465             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 466             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 467
 468             /* Update potential sum for this i atom from the interaction with this j atom. */
 469             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 470             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 471
 472             fscal            = felec;
 473
 474             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 475
 476             /* Update vectorial force */
 477             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 478             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 479             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 480
 481             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 482             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 483             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             /* Compute parameters for interactions between i and j atoms */
 490             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 491
 492             /* COULOMB ELECTROSTATICS */
 493             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 494             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 498             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 499
 500             fscal            = felec;
 501
 502             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 503
 504             /* Update vectorial force */
 505             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 506             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 507             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 508
 509             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 510             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 511             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 512
 513             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 514
 515             /* Inner loop uses 131 flops */
 516         }
 517
 518         /* End of innermost loop */
 519
 520         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 521                                               f+i_coord_offset,fshift+i_shift_offset);
 522
 523         ggid                        = gid[iidx];
 524         /* Update potential energies */
 525         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 526         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 527
 528         /* Increment number of inner iterations */
 529         inneriter                  += j_index_end - j_index_start;
 530
 531         /* Outer loop uses 20 flops */
 532     }
 533
 534     /* Increment number of outer iterations */
 535     outeriter        += nri;
 536
 537     /* Update outer/inner flops */
 538
 539     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*131);
 540 }
 541 /*
 542  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 543  * Electrostatics interaction: Coulomb
 544  * VdW interaction:            CubicSplineTable
 545  * Geometry:                   Water3-Particle
 546  * Calculate force/pot:        Force
 547  */
 548 void
 549 nb_kernel_ElecCoul_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 550                     (t_nblist                    * gmx_restrict       nlist,
 551                      rvec                        * gmx_restrict          xx,
 552                      rvec                        * gmx_restrict          ff,
 553                      struct t_forcerec           * gmx_restrict          fr,
 554                      t_mdatoms                   * gmx_restrict     mdatoms,
 555                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 556                      t_nrnb                      * gmx_restrict        nrnb)
 557 {
 558     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 559      * just 0 for non-waters.
 560      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 561      * jnr indices corresponding to data put in the four positions in the SIMD register.
 562      */
 563     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 564     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 565     int              jnrA,jnrB;
 566     int              j_coord_offsetA,j_coord_offsetB;
 567     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 568     real             rcutoff_scalar;
 569     real             *shiftvec,*fshift,*x,*f;
 570     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 571     int              vdwioffset0;
 572     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 573     int              vdwioffset1;
 574     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 575     int              vdwioffset2;
 576     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 577     int              vdwjidx0A,vdwjidx0B;
 578     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 579     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 580     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 581     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 582     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 583     real             *charge;
 584     int              nvdwtype;
 585     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 586     int              *vdwtype;
 587     real             *vdwparam;
 588     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 589     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 590     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 591     real             *vftab;
 592     _fjsp_v2r8       itab_tmp;
 593     _fjsp_v2r8       dummy_mask,cutoff_mask;
 594     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 595     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 596     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 597
 598     x                = xx[0];
 599     f                = ff[0];
 600
 601     nri              = nlist->nri;
 602     iinr             = nlist->iinr;
 603     jindex           = nlist->jindex;
 604     jjnr             = nlist->jjnr;
 605     shiftidx         = nlist->shift;
 606     gid              = nlist->gid;
 607     shiftvec         = fr->shift_vec[0];
 608     fshift           = fr->fshift[0];
 609     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 610     charge           = mdatoms->chargeA;
 611     nvdwtype         = fr->ntype;
 612     vdwparam         = fr->nbfp;
 613     vdwtype          = mdatoms->typeA;
 614
 615     vftab            = kernel_data->table_vdw->data;
 616     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 617
 618     /* Setup water-specific parameters */
 619     inr              = nlist->iinr[0];
 620     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 621     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 622     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 623     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 624
 625     /* Avoid stupid compiler warnings */
 626     jnrA = jnrB = 0;
 627     j_coord_offsetA = 0;
 628     j_coord_offsetB = 0;
 629
 630     outeriter        = 0;
 631     inneriter        = 0;
 632
 633     /* Start outer loop over neighborlists */
 634     for(iidx=0; iidx<nri; iidx++)
 635     {
 636         /* Load shift vector for this list */
 637         i_shift_offset   = DIM*shiftidx[iidx];
 638
 639         /* Load limits for loop over neighbors */
 640         j_index_start    = jindex[iidx];
 641         j_index_end      = jindex[iidx+1];
 642
 643         /* Get outer coordinate index */
 644         inr              = iinr[iidx];
 645         i_coord_offset   = DIM*inr;
 646
 647         /* Load i particle coords and add shift vector */
 648         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 649                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 650
 651         fix0             = _fjsp_setzero_v2r8();
 652         fiy0             = _fjsp_setzero_v2r8();
 653         fiz0             = _fjsp_setzero_v2r8();
 654         fix1             = _fjsp_setzero_v2r8();
 655         fiy1             = _fjsp_setzero_v2r8();
 656         fiz1             = _fjsp_setzero_v2r8();
 657         fix2             = _fjsp_setzero_v2r8();
 658         fiy2             = _fjsp_setzero_v2r8();
 659         fiz2             = _fjsp_setzero_v2r8();
 660
 661         /* Start inner kernel loop */
 662         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 663         {
 664
 665             /* Get j neighbor index, and coordinate index */
 666             jnrA             = jjnr[jidx];
 667             jnrB             = jjnr[jidx+1];
 668             j_coord_offsetA  = DIM*jnrA;
 669             j_coord_offsetB  = DIM*jnrB;
 670
 671             /* load j atom coordinates */
 672             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 673                                               &jx0,&jy0,&jz0);
 674
 675             /* Calculate displacement vector */
 676             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 677             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 678             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 679             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 680             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 681             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 682             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 683             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 684             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 685
 686             /* Calculate squared distance and things based on it */
 687             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 688             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 689             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 690
 691             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 692             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 693             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 694
 695             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 696             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 697             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 698
 699             /* Load parameters for j particles */
 700             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 701             vdwjidx0A        = 2*vdwtype[jnrA+0];
 702             vdwjidx0B        = 2*vdwtype[jnrB+0];
 703
 704             fjx0             = _fjsp_setzero_v2r8();
 705             fjy0             = _fjsp_setzero_v2r8();
 706             fjz0             = _fjsp_setzero_v2r8();
 707
 708             /**************************
 709              * CALCULATE INTERACTIONS *
 710              **************************/
 711
 712             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 713
 714             /* Compute parameters for interactions between i and j atoms */
 715             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 716             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 717                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 718
 719             /* Calculate table index by multiplying r with table scale and truncate to integer */
 720             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 721             itab_tmp         = _fjsp_dtox_v2r8(rt);
 722             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 723             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 724             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 725
 726             vfconv.i[0]     *= 8;
 727             vfconv.i[1]     *= 8;
 728
 729             /* COULOMB ELECTROSTATICS */
 730             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 731             felec            = _fjsp_mul_v2r8(velec,rinvsq00);
 732
 733             /* CUBIC SPLINE TABLE DISPERSION */
 734             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 735             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 736             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 737             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 738             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 739             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 740             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 741             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 742             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 743
 744             /* CUBIC SPLINE TABLE REPULSION */
 745             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 746             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 747             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 748             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 749             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 750             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 751             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 752             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 753             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 754             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 755
 756             fscal            = _fjsp_add_v2r8(felec,fvdw);
 757
 758             /* Update vectorial force */
 759             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 760             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 761             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 762
 763             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 764             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 765             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 766
 767             /**************************
 768              * CALCULATE INTERACTIONS *
 769              **************************/
 770
 771             /* Compute parameters for interactions between i and j atoms */
 772             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 773
 774             /* COULOMB ELECTROSTATICS */
 775             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 776             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 777
 778             fscal            = felec;
 779
 780             /* Update vectorial force */
 781             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 782             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 783             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 784
 785             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 786             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 787             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 788
 789             /**************************
 790              * CALCULATE INTERACTIONS *
 791              **************************/
 792
 793             /* Compute parameters for interactions between i and j atoms */
 794             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 795
 796             /* COULOMB ELECTROSTATICS */
 797             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 798             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 799
 800             fscal            = felec;
 801
 802             /* Update vectorial force */
 803             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 804             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 805             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 806
 807             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 808             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 809             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 810
 811             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 812
 813             /* Inner loop uses 120 flops */
 814         }
 815
 816         if(jidx<j_index_end)
 817         {
 818
 819             jnrA             = jjnr[jidx];
 820             j_coord_offsetA  = DIM*jnrA;
 821
 822             /* load j atom coordinates */
 823             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 824                                               &jx0,&jy0,&jz0);
 825
 826             /* Calculate displacement vector */
 827             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 828             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 829             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 830             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 831             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 832             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 833             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 834             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 835             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 836
 837             /* Calculate squared distance and things based on it */
 838             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 839             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 840             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 841
 842             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 843             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 844             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 845
 846             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 847             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 848             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 849
 850             /* Load parameters for j particles */
 851             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 852             vdwjidx0A        = 2*vdwtype[jnrA+0];
 853
 854             fjx0             = _fjsp_setzero_v2r8();
 855             fjy0             = _fjsp_setzero_v2r8();
 856             fjz0             = _fjsp_setzero_v2r8();
 857
 858             /**************************
 859              * CALCULATE INTERACTIONS *
 860              **************************/
 861
 862             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 863
 864             /* Compute parameters for interactions between i and j atoms */
 865             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 866             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 867                                          vdwparam+vdwioffset0+vdwjidx0B,&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 }