src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_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_ElecCSTab_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwLJ_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_elec->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->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             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 214             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 215             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 216
 217             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 218
 219             /* Load parameters for j particles */
 220             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 221             vdwjidx0A        = 2*vdwtype[jnrA+0];
 222             vdwjidx0B        = 2*vdwtype[jnrB+0];
 223
 224             fjx0             = _fjsp_setzero_v2r8();
 225             fjy0             = _fjsp_setzero_v2r8();
 226             fjz0             = _fjsp_setzero_v2r8();
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             /* Compute parameters for interactions between i and j atoms */
 233             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 234                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 235
 236             /* LENNARD-JONES DISPERSION/REPULSION */
 237
 238             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 239             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 240             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 241             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 242             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 243
 244             /* Update potential sum for this i atom from the interaction with this j atom. */
 245             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 246
 247             fscal            = fvdw;
 248
 249             /* Update vectorial force */
 250             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 251             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 252             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 253
 254             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 255             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 256             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 257
 258             /**************************
 259              * CALCULATE INTERACTIONS *
 260              **************************/
 261
 262             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 263
 264             /* Compute parameters for interactions between i and j atoms */
 265             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 266
 267             /* Calculate table index by multiplying r with table scale and truncate to integer */
 268             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 269             itab_tmp         = _fjsp_dtox_v2r8(rt);
 270             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 271             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 272             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 273
 274             vfconv.i[0]     *= 4;
 275             vfconv.i[1]     *= 4;
 276
 277             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 278             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 279             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 280             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 281             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 282             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 283             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 284             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 285             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 286             velec            = _fjsp_mul_v2r8(qq10,VV);
 287             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 288             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 289
 290             /* Update potential sum for this i atom from the interaction with this j atom. */
 291             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 292
 293             fscal            = felec;
 294
 295             /* Update vectorial force */
 296             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 297             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 298             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 299
 300             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 301             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 302             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 303
 304             /**************************
 305              * CALCULATE INTERACTIONS *
 306              **************************/
 307
 308             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 309
 310             /* Compute parameters for interactions between i and j atoms */
 311             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 312
 313             /* Calculate table index by multiplying r with table scale and truncate to integer */
 314             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 315             itab_tmp         = _fjsp_dtox_v2r8(rt);
 316             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 317             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 318             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 319
 320             vfconv.i[0]     *= 4;
 321             vfconv.i[1]     *= 4;
 322
 323             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 324             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 325             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 326             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 327             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 328             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 329             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 330             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 331             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 332             velec            = _fjsp_mul_v2r8(qq20,VV);
 333             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 334             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 338
 339             fscal            = felec;
 340
 341             /* Update vectorial force */
 342             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 343             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 344             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 345
 346             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 347             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 348             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 349
 350             /**************************
 351              * CALCULATE INTERACTIONS *
 352              **************************/
 353
 354             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 355
 356             /* Compute parameters for interactions between i and j atoms */
 357             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 358
 359             /* Calculate table index by multiplying r with table scale and truncate to integer */
 360             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 361             itab_tmp         = _fjsp_dtox_v2r8(rt);
 362             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 363             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 364             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 365
 366             vfconv.i[0]     *= 4;
 367             vfconv.i[1]     *= 4;
 368
 369             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 370             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 371             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 372             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 373             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 374             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 375             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 376             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 377             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 378             velec            = _fjsp_mul_v2r8(qq30,VV);
 379             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 380             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Update vectorial force */
 388             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 389             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 390             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 391
 392             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 393             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 394             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 395
 396             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 397
 398             /* Inner loop uses 176 flops */
 399         }
 400
 401         if(jidx<j_index_end)
 402         {
 403
 404             jnrA             = jjnr[jidx];
 405             j_coord_offsetA  = DIM*jnrA;
 406
 407             /* load j atom coordinates */
 408             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 409                                               &jx0,&jy0,&jz0);
 410
 411             /* Calculate displacement vector */
 412             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 413             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 414             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 415             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 416             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 417             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 418             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 419             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 420             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 421             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 422             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 423             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 424
 425             /* Calculate squared distance and things based on it */
 426             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 427             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 428             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 429             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 430
 431             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 432             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 433             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 434
 435             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 436
 437             /* Load parameters for j particles */
 438             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 439             vdwjidx0A        = 2*vdwtype[jnrA+0];
 440
 441             fjx0             = _fjsp_setzero_v2r8();
 442             fjy0             = _fjsp_setzero_v2r8();
 443             fjz0             = _fjsp_setzero_v2r8();
 444
 445             /**************************
 446              * CALCULATE INTERACTIONS *
 447              **************************/
 448
 449             /* Compute parameters for interactions between i and j atoms */
 450             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 451                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 452
 453             /* LENNARD-JONES DISPERSION/REPULSION */
 454
 455             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 456             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 457             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 458             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 459             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 460
 461             /* Update potential sum for this i atom from the interaction with this j atom. */
 462             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 463             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 464
 465             fscal            = fvdw;
 466
 467             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 468
 469             /* Update vectorial force */
 470             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 471             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 472             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 473
 474             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 475             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 476             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 477
 478             /**************************
 479              * CALCULATE INTERACTIONS *
 480              **************************/
 481
 482             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 483
 484             /* Compute parameters for interactions between i and j atoms */
 485             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 486
 487             /* Calculate table index by multiplying r with table scale and truncate to integer */
 488             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 489             itab_tmp         = _fjsp_dtox_v2r8(rt);
 490             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 491             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 492             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 493
 494             vfconv.i[0]     *= 4;
 495             vfconv.i[1]     *= 4;
 496
 497             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 498             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 499             F                = _fjsp_setzero_v2r8();
 500             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 501             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 502             H                = _fjsp_setzero_v2r8();
 503             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 504             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 505             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 506             velec            = _fjsp_mul_v2r8(qq10,VV);
 507             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 508             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 512             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 513
 514             fscal            = felec;
 515
 516             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 517
 518             /* Update vectorial force */
 519             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 520             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 521             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 522
 523             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 524             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 525             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 526
 527             /**************************
 528              * CALCULATE INTERACTIONS *
 529              **************************/
 530
 531             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 532
 533             /* Compute parameters for interactions between i and j atoms */
 534             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 535
 536             /* Calculate table index by multiplying r with table scale and truncate to integer */
 537             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 538             itab_tmp         = _fjsp_dtox_v2r8(rt);
 539             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 540             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 541             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 542
 543             vfconv.i[0]     *= 4;
 544             vfconv.i[1]     *= 4;
 545
 546             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 547             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 548             F                = _fjsp_setzero_v2r8();
 549             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 550             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 551             H                = _fjsp_setzero_v2r8();
 552             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 553             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 554             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 555             velec            = _fjsp_mul_v2r8(qq20,VV);
 556             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 557             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 558
 559             /* Update potential sum for this i atom from the interaction with this j atom. */
 560             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 561             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 562
 563             fscal            = felec;
 564
 565             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 566
 567             /* Update vectorial force */
 568             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 569             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 570             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 571
 572             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 573             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 574             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 575
 576             /**************************
 577              * CALCULATE INTERACTIONS *
 578              **************************/
 579
 580             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 581
 582             /* Compute parameters for interactions between i and j atoms */
 583             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 584
 585             /* Calculate table index by multiplying r with table scale and truncate to integer */
 586             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 587             itab_tmp         = _fjsp_dtox_v2r8(rt);
 588             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 589             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 590             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 591
 592             vfconv.i[0]     *= 4;
 593             vfconv.i[1]     *= 4;
 594
 595             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 596             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 597             F                = _fjsp_setzero_v2r8();
 598             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 599             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 600             H                = _fjsp_setzero_v2r8();
 601             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 602             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 603             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 604             velec            = _fjsp_mul_v2r8(qq30,VV);
 605             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 606             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 607
 608             /* Update potential sum for this i atom from the interaction with this j atom. */
 609             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 610             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 611
 612             fscal            = felec;
 613
 614             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 615
 616             /* Update vectorial force */
 617             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 618             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 619             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 620
 621             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 622             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 623             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 624
 625             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 626
 627             /* Inner loop uses 176 flops */
 628         }
 629
 630         /* End of innermost loop */
 631
 632         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 633                                               f+i_coord_offset,fshift+i_shift_offset);
 634
 635         ggid                        = gid[iidx];
 636         /* Update potential energies */
 637         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 638         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 639
 640         /* Increment number of inner iterations */
 641         inneriter                  += j_index_end - j_index_start;
 642
 643         /* Outer loop uses 26 flops */
 644     }
 645
 646     /* Increment number of outer iterations */
 647     outeriter        += nri;
 648
 649     /* Update outer/inner flops */
 650
 651     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*176);
 652 }
 653 /*
 654  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 655  * Electrostatics interaction: CubicSplineTable
 656  * VdW interaction:            LennardJones
 657  * Geometry:                   Water4-Particle
 658  * Calculate force/pot:        Force
 659  */
 660 void
 661 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 662                     (t_nblist                    * gmx_restrict       nlist,
 663                      rvec                        * gmx_restrict          xx,
 664                      rvec                        * gmx_restrict          ff,
 665                      struct t_forcerec           * gmx_restrict          fr,
 666                      t_mdatoms                   * gmx_restrict     mdatoms,
 667                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 668                      t_nrnb                      * gmx_restrict        nrnb)
 669 {
 670     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 671      * just 0 for non-waters.
 672      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 673      * jnr indices corresponding to data put in the four positions in the SIMD register.
 674      */
 675     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 676     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 677     int              jnrA,jnrB;
 678     int              j_coord_offsetA,j_coord_offsetB;
 679     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 680     real             rcutoff_scalar;
 681     real             *shiftvec,*fshift,*x,*f;
 682     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 683     int              vdwioffset0;
 684     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 685     int              vdwioffset1;
 686     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 687     int              vdwioffset2;
 688     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 689     int              vdwioffset3;
 690     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 691     int              vdwjidx0A,vdwjidx0B;
 692     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 693     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 694     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 695     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 696     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 697     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 698     real             *charge;
 699     int              nvdwtype;
 700     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 701     int              *vdwtype;
 702     real             *vdwparam;
 703     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 704     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 705     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 706     real             *vftab;
 707     _fjsp_v2r8       itab_tmp;
 708     _fjsp_v2r8       dummy_mask,cutoff_mask;
 709     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 710     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 711     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 712
 713     x                = xx[0];
 714     f                = ff[0];
 715
 716     nri              = nlist->nri;
 717     iinr             = nlist->iinr;
 718     jindex           = nlist->jindex;
 719     jjnr             = nlist->jjnr;
 720     shiftidx         = nlist->shift;
 721     gid              = nlist->gid;
 722     shiftvec         = fr->shift_vec[0];
 723     fshift           = fr->fshift[0];
 724     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 725     charge           = mdatoms->chargeA;
 726     nvdwtype         = fr->ntype;
 727     vdwparam         = fr->nbfp;
 728     vdwtype          = mdatoms->typeA;
 729
 730     vftab            = kernel_data->table_elec->data;
 731     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 732
 733     /* Setup water-specific parameters */
 734     inr              = nlist->iinr[0];
 735     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 736     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 737     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 738     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 739
 740     /* Avoid stupid compiler warnings */
 741     jnrA = jnrB = 0;
 742     j_coord_offsetA = 0;
 743     j_coord_offsetB = 0;
 744
 745     outeriter        = 0;
 746     inneriter        = 0;
 747
 748     /* Start outer loop over neighborlists */
 749     for(iidx=0; iidx<nri; iidx++)
 750     {
 751         /* Load shift vector for this list */
 752         i_shift_offset   = DIM*shiftidx[iidx];
 753
 754         /* Load limits for loop over neighbors */
 755         j_index_start    = jindex[iidx];
 756         j_index_end      = jindex[iidx+1];
 757
 758         /* Get outer coordinate index */
 759         inr              = iinr[iidx];
 760         i_coord_offset   = DIM*inr;
 761
 762         /* Load i particle coords and add shift vector */
 763         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 764                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 765
 766         fix0             = _fjsp_setzero_v2r8();
 767         fiy0             = _fjsp_setzero_v2r8();
 768         fiz0             = _fjsp_setzero_v2r8();
 769         fix1             = _fjsp_setzero_v2r8();
 770         fiy1             = _fjsp_setzero_v2r8();
 771         fiz1             = _fjsp_setzero_v2r8();
 772         fix2             = _fjsp_setzero_v2r8();
 773         fiy2             = _fjsp_setzero_v2r8();
 774         fiz2             = _fjsp_setzero_v2r8();
 775         fix3             = _fjsp_setzero_v2r8();
 776         fiy3             = _fjsp_setzero_v2r8();
 777         fiz3             = _fjsp_setzero_v2r8();
 778
 779         /* Start inner kernel loop */
 780         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 781         {
 782
 783             /* Get j neighbor index, and coordinate index */
 784             jnrA             = jjnr[jidx];
 785             jnrB             = jjnr[jidx+1];
 786             j_coord_offsetA  = DIM*jnrA;
 787             j_coord_offsetB  = DIM*jnrB;
 788
 789             /* load j atom coordinates */
 790             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 791                                               &jx0,&jy0,&jz0);
 792
 793             /* Calculate displacement vector */
 794             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 795             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 796             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 797             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 798             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 799             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 800             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 801             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 802             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 803             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 804             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 805             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 806
 807             /* Calculate squared distance and things based on it */
 808             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 809             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 810             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 811             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 812
 813             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 814             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 815             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 816
 817             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 818
 819             /* Load parameters for j particles */
 820             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 821             vdwjidx0A        = 2*vdwtype[jnrA+0];
 822             vdwjidx0B        = 2*vdwtype[jnrB+0];
 823
 824             fjx0             = _fjsp_setzero_v2r8();
 825             fjy0             = _fjsp_setzero_v2r8();
 826             fjz0             = _fjsp_setzero_v2r8();
 827
 828             /**************************
 829              * CALCULATE INTERACTIONS *
 830              **************************/
 831
 832             /* Compute parameters for interactions between i and j atoms */
 833             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 834                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 835
 836             /* LENNARD-JONES DISPERSION/REPULSION */
 837
 838             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 839             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 840
 841             fscal            = fvdw;
 842
 843             /* Update vectorial force */
 844             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 845             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 846             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 847
 848             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 849             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 850             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 857
 858             /* Compute parameters for interactions between i and j atoms */
 859             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 860
 861             /* Calculate table index by multiplying r with table scale and truncate to integer */
 862             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 863             itab_tmp         = _fjsp_dtox_v2r8(rt);
 864             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 865             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 866             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 867
 868             vfconv.i[0]     *= 4;
 869             vfconv.i[1]     *= 4;
 870
 871             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 872             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 873             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 874             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 875             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 876             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 877             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 878             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 879             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 880             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 881
 882             fscal            = felec;
 883
 884             /* Update vectorial force */
 885             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 886             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 887             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 888
 889             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 890             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 891             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 892
 893             /**************************
 894              * CALCULATE INTERACTIONS *
 895              **************************/
 896
 897             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 898
 899             /* Compute parameters for interactions between i and j atoms */
 900             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 901
 902             /* Calculate table index by multiplying r with table scale and truncate to integer */
 903             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 904             itab_tmp         = _fjsp_dtox_v2r8(rt);
 905             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 906             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 907             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 908
 909             vfconv.i[0]     *= 4;
 910             vfconv.i[1]     *= 4;
 911
 912             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 913             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 914             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 915             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 916             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 917             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 918             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 919             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 920             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 921             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 922
 923             fscal            = felec;
 924
 925             /* Update vectorial force */
 926             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 927             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 928             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 929
 930             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 931             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 932             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 933
 934             /**************************
 935              * CALCULATE INTERACTIONS *
 936              **************************/
 937
 938             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 939
 940             /* Compute parameters for interactions between i and j atoms */
 941             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 942
 943             /* Calculate table index by multiplying r with table scale and truncate to integer */
 944             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 945             itab_tmp         = _fjsp_dtox_v2r8(rt);
 946             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 947             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 948             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 949
 950             vfconv.i[0]     *= 4;
 951             vfconv.i[1]     *= 4;
 952
 953             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 954             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 955             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 956             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 957             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 958             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 959             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 960             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 961             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 962             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 963
 964             fscal            = felec;
 965
 966             /* Update vectorial force */
 967             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 968             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 969             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 970
 971             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 972             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 973             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 974
 975             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 976
 977             /* Inner loop uses 159 flops */
 978         }
 979
 980         if(jidx<j_index_end)
 981         {
 982
 983             jnrA             = jjnr[jidx];
 984             j_coord_offsetA  = DIM*jnrA;
 985
 986             /* load j atom coordinates */
 987             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 988                                               &jx0,&jy0,&jz0);
 989
 990             /* Calculate displacement vector */
 991             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 992             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 993             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 994             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 995             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 996             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 997             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 998             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 999             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1000             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1001             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1002             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1003
1004             /* Calculate squared distance and things based on it */
1005             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1006             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1007             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1008             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1009
1010             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1011             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1012             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1013
1014             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
1015
1016             /* Load parameters for j particles */
1017             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1018             vdwjidx0A        = 2*vdwtype[jnrA+0];
1019
1020             fjx0             = _fjsp_setzero_v2r8();
1021             fjy0             = _fjsp_setzero_v2r8();
1022             fjz0             = _fjsp_setzero_v2r8();
1023
1024             /**************************
1025              * CALCULATE INTERACTIONS *
1026              **************************/
1027
1028             /* Compute parameters for interactions between i and j atoms */
1029             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1030                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1031
1032             /* LENNARD-JONES DISPERSION/REPULSION */
1033
1034             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1035             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1036
1037             fscal            = fvdw;
1038
1039             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1040
1041             /* Update vectorial force */
1042             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1043             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1044             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1045
1046             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1047             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1048             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1049
1050             /**************************
1051              * CALCULATE INTERACTIONS *
1052              **************************/
1053
1054             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1055
1056             /* Compute parameters for interactions between i and j atoms */
1057             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1058
1059             /* Calculate table index by multiplying r with table scale and truncate to integer */
1060             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1061             itab_tmp         = _fjsp_dtox_v2r8(rt);
1062             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1063             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1064             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1065
1066             vfconv.i[0]     *= 4;
1067             vfconv.i[1]     *= 4;
1068
1069             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1070             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1071             F                = _fjsp_setzero_v2r8();
1072             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1073             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1074             H                = _fjsp_setzero_v2r8();
1075             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1076             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1077             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1078             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1079
1080             fscal            = felec;
1081
1082             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1083
1084             /* Update vectorial force */
1085             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1086             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1087             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1088
1089             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1090             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1091             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1092
1093             /**************************
1094              * CALCULATE INTERACTIONS *
1095              **************************/
1096
1097             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1098
1099             /* Compute parameters for interactions between i and j atoms */
1100             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1101
1102             /* Calculate table index by multiplying r with table scale and truncate to integer */
1103             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1104             itab_tmp         = _fjsp_dtox_v2r8(rt);
1105             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1106             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1107             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1108
1109             vfconv.i[0]     *= 4;
1110             vfconv.i[1]     *= 4;
1111
1112             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1113             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1114             F                = _fjsp_setzero_v2r8();
1115             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1116             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1117             H                = _fjsp_setzero_v2r8();
1118             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1119             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1120             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1121             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1122
1123             fscal            = felec;
1124
1125             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1126
1127             /* Update vectorial force */
1128             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1129             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1130             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1131
1132             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1133             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1134             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1135
1136             /**************************
1137              * CALCULATE INTERACTIONS *
1138              **************************/
1139
1140             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1141
1142             /* Compute parameters for interactions between i and j atoms */
1143             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1144
1145             /* Calculate table index by multiplying r with table scale and truncate to integer */
1146             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1147             itab_tmp         = _fjsp_dtox_v2r8(rt);
1148             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1149             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1150             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1151
1152             vfconv.i[0]     *= 4;
1153             vfconv.i[1]     *= 4;
1154
1155             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1156             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1157             F                = _fjsp_setzero_v2r8();
1158             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1159             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1160             H                = _fjsp_setzero_v2r8();
1161             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1162             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1163             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1164             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1165
1166             fscal            = felec;
1167
1168             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1169
1170             /* Update vectorial force */
1171             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1172             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1173             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1174
1175             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1176             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1177             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1178
1179             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1180
1181             /* Inner loop uses 159 flops */
1182         }
1183
1184         /* End of innermost loop */
1185
1186         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1187                                               f+i_coord_offset,fshift+i_shift_offset);
1188
1189         /* Increment number of inner iterations */
1190         inneriter                  += j_index_end - j_index_start;
1191
1192         /* Outer loop uses 24 flops */
1193     }
1194
1195     /* Increment number of outer iterations */
1196     outeriter        += nri;
1197
1198     /* Update outer/inner flops */
1199
1200     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);
1201 }