src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, 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_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_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              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  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       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[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     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* Avoid stupid compiler warnings */
 137     jnrA = jnrB = 0;
 138     j_coord_offsetA = 0;
 139     j_coord_offsetB = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     /* Start outer loop over neighborlists */
 145     for(iidx=0; iidx<nri; iidx++)
 146     {
 147         /* Load shift vector for this list */
 148         i_shift_offset   = DIM*shiftidx[iidx];
 149
 150         /* Load limits for loop over neighbors */
 151         j_index_start    = jindex[iidx];
 152         j_index_end      = jindex[iidx+1];
 153
 154         /* Get outer coordinate index */
 155         inr              = iinr[iidx];
 156         i_coord_offset   = DIM*inr;
 157
 158         /* Load i particle coords and add shift vector */
 159         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 160                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165         fix1             = _fjsp_setzero_v2r8();
 166         fiy1             = _fjsp_setzero_v2r8();
 167         fiz1             = _fjsp_setzero_v2r8();
 168         fix2             = _fjsp_setzero_v2r8();
 169         fiy2             = _fjsp_setzero_v2r8();
 170         fiz2             = _fjsp_setzero_v2r8();
 171         fix3             = _fjsp_setzero_v2r8();
 172         fiy3             = _fjsp_setzero_v2r8();
 173         fiz3             = _fjsp_setzero_v2r8();
 174
 175         /* Reset potential sums */
 176         velecsum         = _fjsp_setzero_v2r8();
 177         vvdwsum          = _fjsp_setzero_v2r8();
 178
 179         /* Start inner kernel loop */
 180         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 181         {
 182
 183             /* Get j neighbor index, and coordinate index */
 184             jnrA             = jjnr[jidx];
 185             jnrB             = jjnr[jidx+1];
 186             j_coord_offsetA  = DIM*jnrA;
 187             j_coord_offsetB  = DIM*jnrB;
 188
 189             /* load j atom coordinates */
 190             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 191                                               &jx0,&jy0,&jz0);
 192
 193             /* Calculate displacement vector */
 194             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 195             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 196             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 197             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 198             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 199             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 200             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 201             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 202             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 203             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 204             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 205             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 217
 218             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 219             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 220             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 221
 222             /* Load parameters for j particles */
 223             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 224             vdwjidx0A        = 2*vdwtype[jnrA+0];
 225             vdwjidx0B        = 2*vdwtype[jnrB+0];
 226
 227             fjx0             = _fjsp_setzero_v2r8();
 228             fjy0             = _fjsp_setzero_v2r8();
 229             fjz0             = _fjsp_setzero_v2r8();
 230
 231             /**************************
 232              * CALCULATE INTERACTIONS *
 233              **************************/
 234
 235             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 239                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 240
 241             /* Calculate table index by multiplying r with table scale and truncate to integer */
 242             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 243             itab_tmp         = _fjsp_dtox_v2r8(rt);
 244             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 245             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 246             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 247
 248             vfconv.i[0]     *= 8;
 249             vfconv.i[1]     *= 8;
 250
 251             /* CUBIC SPLINE TABLE DISPERSION */
 252             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 253             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 254             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 255             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 256             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 257             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 258             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 259             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 260             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 261             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 262             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 263
 264             /* CUBIC SPLINE TABLE REPULSION */
 265             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 266             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 267             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 268             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 269             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 271             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 272             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 273             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 274             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 275             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 276             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 277             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 278
 279             /* Update potential sum for this i atom from the interaction with this j atom. */
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = fvdw;
 283
 284             /* Update vectorial force */
 285             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 286             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 287             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 288
 289             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 290             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 291             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             /* Compute parameters for interactions between i and j atoms */
 298             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 299
 300             /* COULOMB ELECTROSTATICS */
 301             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 302             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 303
 304             /* Update potential sum for this i atom from the interaction with this j atom. */
 305             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 306
 307             fscal            = felec;
 308
 309             /* Update vectorial force */
 310             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 311             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 312             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 313
 314             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 315             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 316             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             /* Compute parameters for interactions between i and j atoms */
 323             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 324
 325             /* COULOMB ELECTROSTATICS */
 326             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 327             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 336             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 337             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 342
 343             /**************************
 344              * CALCULATE INTERACTIONS *
 345              **************************/
 346
 347             /* Compute parameters for interactions between i and j atoms */
 348             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 349
 350             /* COULOMB ELECTROSTATICS */
 351             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 352             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 356
 357             fscal            = felec;
 358
 359             /* Update vectorial force */
 360             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 361             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 362             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 363
 364             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 365             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 366             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 367
 368             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 369
 370             /* Inner loop uses 155 flops */
 371         }
 372
 373         if(jidx<j_index_end)
 374         {
 375
 376             jnrA             = jjnr[jidx];
 377             j_coord_offsetA  = DIM*jnrA;
 378
 379             /* load j atom coordinates */
 380             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 381                                               &jx0,&jy0,&jz0);
 382
 383             /* Calculate displacement vector */
 384             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 385             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 386             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 387             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 388             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 389             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 390             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 391             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 392             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 393             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 394             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 395             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 396
 397             /* Calculate squared distance and things based on it */
 398             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 399             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 400             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 401             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 402
 403             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 404             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 405             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 406             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 407
 408             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 409             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 410             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 411
 412             /* Load parameters for j particles */
 413             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 414             vdwjidx0A        = 2*vdwtype[jnrA+0];
 415
 416             fjx0             = _fjsp_setzero_v2r8();
 417             fjy0             = _fjsp_setzero_v2r8();
 418             fjz0             = _fjsp_setzero_v2r8();
 419
 420             /**************************
 421              * CALCULATE INTERACTIONS *
 422              **************************/
 423
 424             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 425
 426             /* Compute parameters for interactions between i and j atoms */
 427             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 428                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 429
 430             /* Calculate table index by multiplying r with table scale and truncate to integer */
 431             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 432             itab_tmp         = _fjsp_dtox_v2r8(rt);
 433             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 434             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 435             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 436
 437             vfconv.i[0]     *= 8;
 438             vfconv.i[1]     *= 8;
 439
 440             /* CUBIC SPLINE TABLE DISPERSION */
 441             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 442             F                = _fjsp_setzero_v2r8();
 443             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 444             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 445             H                = _fjsp_setzero_v2r8();
 446             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 447             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 448             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 449             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 450             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 451             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 452
 453             /* CUBIC SPLINE TABLE REPULSION */
 454             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 455             F                = _fjsp_setzero_v2r8();
 456             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 457             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 458             H                = _fjsp_setzero_v2r8();
 459             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 460             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 461             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 462             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 463             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 464             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 465             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 466             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 467
 468             /* Update potential sum for this i atom from the interaction with this j atom. */
 469             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 470             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 471
 472             fscal            = fvdw;
 473
 474             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 475
 476             /* Update vectorial force */
 477             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 478             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 479             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 480
 481             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 482             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 483             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             /* Compute parameters for interactions between i and j atoms */
 490             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 491
 492             /* COULOMB ELECTROSTATICS */
 493             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 494             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 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             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 506             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 507             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 508
 509             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 510             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 511             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             /* Compute parameters for interactions between i and j atoms */
 518             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 519
 520             /* COULOMB ELECTROSTATICS */
 521             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 522             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 526             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 531
 532             /* Update vectorial force */
 533             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 534             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 535             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 536
 537             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 538             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 539             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 540
 541             /**************************
 542              * CALCULATE INTERACTIONS *
 543              **************************/
 544
 545             /* Compute parameters for interactions between i and j atoms */
 546             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 547
 548             /* COULOMB ELECTROSTATICS */
 549             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 550             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 551
 552             /* Update potential sum for this i atom from the interaction with this j atom. */
 553             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 554             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 555
 556             fscal            = felec;
 557
 558             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 559
 560             /* Update vectorial force */
 561             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 562             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 563             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 564
 565             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 566             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 567             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 568
 569             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 570
 571             /* Inner loop uses 155 flops */
 572         }
 573
 574         /* End of innermost loop */
 575
 576         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 577                                               f+i_coord_offset,fshift+i_shift_offset);
 578
 579         ggid                        = gid[iidx];
 580         /* Update potential energies */
 581         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 582         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 583
 584         /* Increment number of inner iterations */
 585         inneriter                  += j_index_end - j_index_start;
 586
 587         /* Outer loop uses 26 flops */
 588     }
 589
 590     /* Increment number of outer iterations */
 591     outeriter        += nri;
 592
 593     /* Update outer/inner flops */
 594
 595     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*155);
 596 }
 597 /*
 598  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 599  * Electrostatics interaction: Coulomb
 600  * VdW interaction:            CubicSplineTable
 601  * Geometry:                   Water4-Particle
 602  * Calculate force/pot:        Force
 603  */
 604 void
 605 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 606                     (t_nblist                    * gmx_restrict       nlist,
 607                      rvec                        * gmx_restrict          xx,
 608                      rvec                        * gmx_restrict          ff,
 609                      struct t_forcerec           * gmx_restrict          fr,
 610                      t_mdatoms                   * gmx_restrict     mdatoms,
 611                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 612                      t_nrnb                      * gmx_restrict        nrnb)
 613 {
 614     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 615      * just 0 for non-waters.
 616      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 617      * jnr indices corresponding to data put in the four positions in the SIMD register.
 618      */
 619     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 620     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 621     int              jnrA,jnrB;
 622     int              j_coord_offsetA,j_coord_offsetB;
 623     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 624     real             rcutoff_scalar;
 625     real             *shiftvec,*fshift,*x,*f;
 626     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 627     int              vdwioffset0;
 628     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 629     int              vdwioffset1;
 630     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 631     int              vdwioffset2;
 632     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 633     int              vdwioffset3;
 634     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 635     int              vdwjidx0A,vdwjidx0B;
 636     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 637     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 638     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 639     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 640     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 641     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 642     real             *charge;
 643     int              nvdwtype;
 644     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 645     int              *vdwtype;
 646     real             *vdwparam;
 647     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 648     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 649     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 650     real             *vftab;
 651     _fjsp_v2r8       itab_tmp;
 652     _fjsp_v2r8       dummy_mask,cutoff_mask;
 653     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 654     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 655     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 656
 657     x                = xx[0];
 658     f                = ff[0];
 659
 660     nri              = nlist->nri;
 661     iinr             = nlist->iinr;
 662     jindex           = nlist->jindex;
 663     jjnr             = nlist->jjnr;
 664     shiftidx         = nlist->shift;
 665     gid              = nlist->gid;
 666     shiftvec         = fr->shift_vec[0];
 667     fshift           = fr->fshift[0];
 668     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 669     charge           = mdatoms->chargeA;
 670     nvdwtype         = fr->ntype;
 671     vdwparam         = fr->nbfp;
 672     vdwtype          = mdatoms->typeA;
 673
 674     vftab            = kernel_data->table_vdw->data;
 675     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 676
 677     /* Setup water-specific parameters */
 678     inr              = nlist->iinr[0];
 679     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 680     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 681     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 682     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 683
 684     /* Avoid stupid compiler warnings */
 685     jnrA = jnrB = 0;
 686     j_coord_offsetA = 0;
 687     j_coord_offsetB = 0;
 688
 689     outeriter        = 0;
 690     inneriter        = 0;
 691
 692     /* Start outer loop over neighborlists */
 693     for(iidx=0; iidx<nri; iidx++)
 694     {
 695         /* Load shift vector for this list */
 696         i_shift_offset   = DIM*shiftidx[iidx];
 697
 698         /* Load limits for loop over neighbors */
 699         j_index_start    = jindex[iidx];
 700         j_index_end      = jindex[iidx+1];
 701
 702         /* Get outer coordinate index */
 703         inr              = iinr[iidx];
 704         i_coord_offset   = DIM*inr;
 705
 706         /* Load i particle coords and add shift vector */
 707         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 708                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 709
 710         fix0             = _fjsp_setzero_v2r8();
 711         fiy0             = _fjsp_setzero_v2r8();
 712         fiz0             = _fjsp_setzero_v2r8();
 713         fix1             = _fjsp_setzero_v2r8();
 714         fiy1             = _fjsp_setzero_v2r8();
 715         fiz1             = _fjsp_setzero_v2r8();
 716         fix2             = _fjsp_setzero_v2r8();
 717         fiy2             = _fjsp_setzero_v2r8();
 718         fiz2             = _fjsp_setzero_v2r8();
 719         fix3             = _fjsp_setzero_v2r8();
 720         fiy3             = _fjsp_setzero_v2r8();
 721         fiz3             = _fjsp_setzero_v2r8();
 722
 723         /* Start inner kernel loop */
 724         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 725         {
 726
 727             /* Get j neighbor index, and coordinate index */
 728             jnrA             = jjnr[jidx];
 729             jnrB             = jjnr[jidx+1];
 730             j_coord_offsetA  = DIM*jnrA;
 731             j_coord_offsetB  = DIM*jnrB;
 732
 733             /* load j atom coordinates */
 734             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 735                                               &jx0,&jy0,&jz0);
 736
 737             /* Calculate displacement vector */
 738             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 739             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 740             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 741             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 742             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 743             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 744             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 745             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 746             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 747             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 748             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 749             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 750
 751             /* Calculate squared distance and things based on it */
 752             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 753             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 754             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 755             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 756
 757             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 758             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 759             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 760             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 761
 762             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 763             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 764             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 765
 766             /* Load parameters for j particles */
 767             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 768             vdwjidx0A        = 2*vdwtype[jnrA+0];
 769             vdwjidx0B        = 2*vdwtype[jnrB+0];
 770
 771             fjx0             = _fjsp_setzero_v2r8();
 772             fjy0             = _fjsp_setzero_v2r8();
 773             fjz0             = _fjsp_setzero_v2r8();
 774
 775             /**************************
 776              * CALCULATE INTERACTIONS *
 777              **************************/
 778
 779             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 780
 781             /* Compute parameters for interactions between i and j atoms */
 782             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 783                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 784
 785             /* Calculate table index by multiplying r with table scale and truncate to integer */
 786             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 787             itab_tmp         = _fjsp_dtox_v2r8(rt);
 788             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 789             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 790             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 791
 792             vfconv.i[0]     *= 8;
 793             vfconv.i[1]     *= 8;
 794
 795             /* CUBIC SPLINE TABLE DISPERSION */
 796             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 797             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 798             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 799             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 800             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 801             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 802             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 803             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 804             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 805
 806             /* CUBIC SPLINE TABLE REPULSION */
 807             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 808             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 809             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 810             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 811             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 812             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 813             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 814             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 815             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 816             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 817
 818             fscal            = fvdw;
 819
 820             /* Update vectorial force */
 821             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 822             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 823             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 824
 825             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 826             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 827             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 828
 829             /**************************
 830              * CALCULATE INTERACTIONS *
 831              **************************/
 832
 833             /* Compute parameters for interactions between i and j atoms */
 834             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 835
 836             /* COULOMB ELECTROSTATICS */
 837             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 838             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 839
 840             fscal            = felec;
 841
 842             /* Update vectorial force */
 843             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 844             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 845             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 846
 847             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 848             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 849             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             /* Compute parameters for interactions between i and j atoms */
 856             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 857
 858             /* COULOMB ELECTROSTATICS */
 859             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 860             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 861
 862             fscal            = felec;
 863
 864             /* Update vectorial force */
 865             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 866             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 867             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 868
 869             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 870             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 871             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 872
 873             /**************************
 874              * CALCULATE INTERACTIONS *
 875              **************************/
 876
 877             /* Compute parameters for interactions between i and j atoms */
 878             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 879
 880             /* COULOMB ELECTROSTATICS */
 881             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 882             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 883
 884             fscal            = felec;
 885
 886             /* Update vectorial force */
 887             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 888             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 889             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 890
 891             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 892             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 893             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 894
 895             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 896
 897             /* Inner loop uses 144 flops */
 898         }
 899
 900         if(jidx<j_index_end)
 901         {
 902
 903             jnrA             = jjnr[jidx];
 904             j_coord_offsetA  = DIM*jnrA;
 905
 906             /* load j atom coordinates */
 907             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 908                                               &jx0,&jy0,&jz0);
 909
 910             /* Calculate displacement vector */
 911             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 912             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 913             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 914             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 915             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 916             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 917             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 918             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 919             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 920             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 921             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 922             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 923
 924             /* Calculate squared distance and things based on it */
 925             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 926             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 927             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 928             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 929
 930             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 931             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 932             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 933             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 934
 935             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 936             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 937             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 938
 939             /* Load parameters for j particles */
 940             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 941             vdwjidx0A        = 2*vdwtype[jnrA+0];
 942
 943             fjx0             = _fjsp_setzero_v2r8();
 944             fjy0             = _fjsp_setzero_v2r8();
 945             fjz0             = _fjsp_setzero_v2r8();
 946
 947             /**************************
 948              * CALCULATE INTERACTIONS *
 949              **************************/
 950
 951             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 952
 953             /* Compute parameters for interactions between i and j atoms */
 954             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 955                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 956
 957             /* Calculate table index by multiplying r with table scale and truncate to integer */
 958             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 959             itab_tmp         = _fjsp_dtox_v2r8(rt);
 960             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 961             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 962             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 963
 964             vfconv.i[0]     *= 8;
 965             vfconv.i[1]     *= 8;
 966
 967             /* CUBIC SPLINE TABLE DISPERSION */
 968             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 969             F                = _fjsp_setzero_v2r8();
 970             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 971             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 972             H                = _fjsp_setzero_v2r8();
 973             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 974             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 975             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 976             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 977
 978             /* CUBIC SPLINE TABLE REPULSION */
 979             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 980             F                = _fjsp_setzero_v2r8();
 981             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 982             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 983             H                = _fjsp_setzero_v2r8();
 984             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 985             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 986             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 987             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 988             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 989
 990             fscal            = fvdw;
 991
 992             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 993
 994             /* Update vectorial force */
 995             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 996             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 997             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 998
 999             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1000             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1001             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1002
1003             /**************************
1004              * CALCULATE INTERACTIONS *
1005              **************************/
1006
1007             /* Compute parameters for interactions between i and j atoms */
1008             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1009
1010             /* COULOMB ELECTROSTATICS */
1011             velec            = _fjsp_mul_v2r8(qq10,rinv10);
1012             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
1013
1014             fscal            = felec;
1015
1016             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1017
1018             /* Update vectorial force */
1019             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1020             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1021             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1022
1023             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1024             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1025             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1026
1027             /**************************
1028              * CALCULATE INTERACTIONS *
1029              **************************/
1030
1031             /* Compute parameters for interactions between i and j atoms */
1032             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1033
1034             /* COULOMB ELECTROSTATICS */
1035             velec            = _fjsp_mul_v2r8(qq20,rinv20);
1036             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
1037
1038             fscal            = felec;
1039
1040             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1041
1042             /* Update vectorial force */
1043             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1044             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1045             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1046
1047             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1048             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1049             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1050
1051             /**************************
1052              * CALCULATE INTERACTIONS *
1053              **************************/
1054
1055             /* Compute parameters for interactions between i and j atoms */
1056             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1057
1058             /* COULOMB ELECTROSTATICS */
1059             velec            = _fjsp_mul_v2r8(qq30,rinv30);
1060             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
1061
1062             fscal            = felec;
1063
1064             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1065
1066             /* Update vectorial force */
1067             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1068             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1069             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1070
1071             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1072             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1073             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1074
1075             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1076
1077             /* Inner loop uses 144 flops */
1078         }
1079
1080         /* End of innermost loop */
1081
1082         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1083                                               f+i_coord_offset,fshift+i_shift_offset);
1084
1085         /* Increment number of inner iterations */
1086         inneriter                  += j_index_end - j_index_start;
1087
1088         /* Outer loop uses 24 flops */
1089     }
1090
1091     /* Increment number of outer iterations */
1092     outeriter        += nri;
1093
1094     /* Update outer/inner flops */
1095
1096     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1097 }