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, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Coulomb
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 104     real             *vftab;
 105     _fjsp_v2r8       itab_tmp;
 106     _fjsp_v2r8       dummy_mask,cutoff_mask;
 107     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 108     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 109     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 110
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_vdw->data;
 129     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     /* Avoid stupid compiler warnings */
 139     jnrA = jnrB = 0;
 140     j_coord_offsetA = 0;
 141     j_coord_offsetB = 0;
 142
 143     outeriter        = 0;
 144     inneriter        = 0;
 145
 146     /* Start outer loop over neighborlists */
 147     for(iidx=0; iidx<nri; iidx++)
 148     {
 149         /* Load shift vector for this list */
 150         i_shift_offset   = DIM*shiftidx[iidx];
 151
 152         /* Load limits for loop over neighbors */
 153         j_index_start    = jindex[iidx];
 154         j_index_end      = jindex[iidx+1];
 155
 156         /* Get outer coordinate index */
 157         inr              = iinr[iidx];
 158         i_coord_offset   = DIM*inr;
 159
 160         /* Load i particle coords and add shift vector */
 161         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 162                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 163
 164         fix0             = _fjsp_setzero_v2r8();
 165         fiy0             = _fjsp_setzero_v2r8();
 166         fiz0             = _fjsp_setzero_v2r8();
 167         fix1             = _fjsp_setzero_v2r8();
 168         fiy1             = _fjsp_setzero_v2r8();
 169         fiz1             = _fjsp_setzero_v2r8();
 170         fix2             = _fjsp_setzero_v2r8();
 171         fiy2             = _fjsp_setzero_v2r8();
 172         fiz2             = _fjsp_setzero_v2r8();
 173         fix3             = _fjsp_setzero_v2r8();
 174         fiy3             = _fjsp_setzero_v2r8();
 175         fiz3             = _fjsp_setzero_v2r8();
 176
 177         /* Reset potential sums */
 178         velecsum         = _fjsp_setzero_v2r8();
 179         vvdwsum          = _fjsp_setzero_v2r8();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 193                                               &jx0,&jy0,&jz0);
 194
 195             /* Calculate displacement vector */
 196             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 197             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 198             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 199             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 200             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 201             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 202             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 203             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 204             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 205             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 206             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 207             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 208
 209             /* Calculate squared distance and things based on it */
 210             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 211             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 212             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 213             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 214
 215             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 216             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 217             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 218             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 219
 220             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 221             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 222             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 223
 224             /* Load parameters for j particles */
 225             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 226             vdwjidx0A        = 2*vdwtype[jnrA+0];
 227             vdwjidx0B        = 2*vdwtype[jnrB+0];
 228
 229             fjx0             = _fjsp_setzero_v2r8();
 230             fjy0             = _fjsp_setzero_v2r8();
 231             fjz0             = _fjsp_setzero_v2r8();
 232
 233             /**************************
 234              * CALCULATE INTERACTIONS *
 235              **************************/
 236
 237             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 238
 239             /* Compute parameters for interactions between i and j atoms */
 240             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 241                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 242
 243             /* Calculate table index by multiplying r with table scale and truncate to integer */
 244             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 245             itab_tmp         = _fjsp_dtox_v2r8(rt);
 246             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 247             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 248             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 249
 250             vfconv.i[0]     *= 8;
 251             vfconv.i[1]     *= 8;
 252
 253             /* CUBIC SPLINE TABLE DISPERSION */
 254             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 255             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 256             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 257             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 258             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 259             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 260             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 261             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 262             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 263             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 264             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 265
 266             /* CUBIC SPLINE TABLE REPULSION */
 267             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 268             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 269             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 270             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 271             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 272             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 273             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 274             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 275             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 276             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 277             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 278             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 279             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 280
 281             /* Update potential sum for this i atom from the interaction with this j atom. */
 282             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 283
 284             fscal            = fvdw;
 285
 286             /* Update vectorial force */
 287             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 288             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 289             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             /* Compute parameters for interactions between i and j atoms */
 300             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 301
 302             /* COULOMB ELECTROSTATICS */
 303             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 304             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 305
 306             /* Update potential sum for this i atom from the interaction with this j atom. */
 307             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 308
 309             fscal            = felec;
 310
 311             /* Update vectorial force */
 312             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 313             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 314             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 315
 316             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 317             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 318             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 319
 320             /**************************
 321              * CALCULATE INTERACTIONS *
 322              **************************/
 323
 324             /* Compute parameters for interactions between i and j atoms */
 325             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 326
 327             /* COULOMB ELECTROSTATICS */
 328             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 329             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 330
 331             /* Update potential sum for this i atom from the interaction with this j atom. */
 332             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 333
 334             fscal            = felec;
 335
 336             /* Update vectorial force */
 337             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 338             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 339             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 340
 341             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 342             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 343             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 344
 345             /**************************
 346              * CALCULATE INTERACTIONS *
 347              **************************/
 348
 349             /* Compute parameters for interactions between i and j atoms */
 350             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 351
 352             /* COULOMB ELECTROSTATICS */
 353             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 354             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 355
 356             /* Update potential sum for this i atom from the interaction with this j atom. */
 357             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 358
 359             fscal            = felec;
 360
 361             /* Update vectorial force */
 362             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 363             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 364             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 365
 366             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 367             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 368             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 369
 370             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 371
 372             /* Inner loop uses 155 flops */
 373         }
 374
 375         if(jidx<j_index_end)
 376         {
 377
 378             jnrA             = jjnr[jidx];
 379             j_coord_offsetA  = DIM*jnrA;
 380
 381             /* load j atom coordinates */
 382             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 383                                               &jx0,&jy0,&jz0);
 384
 385             /* Calculate displacement vector */
 386             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 387             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 388             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 389             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 390             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 391             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 392             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 393             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 394             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 395             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 396             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 397             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 398
 399             /* Calculate squared distance and things based on it */
 400             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 401             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 402             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 403             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 404
 405             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 406             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 407             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 408             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 409
 410             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 411             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 412             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 413
 414             /* Load parameters for j particles */
 415             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 416             vdwjidx0A        = 2*vdwtype[jnrA+0];
 417
 418             fjx0             = _fjsp_setzero_v2r8();
 419             fjy0             = _fjsp_setzero_v2r8();
 420             fjz0             = _fjsp_setzero_v2r8();
 421
 422             /**************************
 423              * CALCULATE INTERACTIONS *
 424              **************************/
 425
 426             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 427
 428             /* Compute parameters for interactions between i and j atoms */
 429             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 430                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 431
 432             /* Calculate table index by multiplying r with table scale and truncate to integer */
 433             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 434             itab_tmp         = _fjsp_dtox_v2r8(rt);
 435             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 436             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 437             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 438
 439             vfconv.i[0]     *= 8;
 440             vfconv.i[1]     *= 8;
 441
 442             /* CUBIC SPLINE TABLE DISPERSION */
 443             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 444             F                = _fjsp_setzero_v2r8();
 445             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 446             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 447             H                = _fjsp_setzero_v2r8();
 448             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 449             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 450             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 451             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 452             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 453             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 454
 455             /* CUBIC SPLINE TABLE REPULSION */
 456             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 457             F                = _fjsp_setzero_v2r8();
 458             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 459             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 460             H                = _fjsp_setzero_v2r8();
 461             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 462             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 463             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 464             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 465             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 466             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 467             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 468             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 472             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 473
 474             fscal            = fvdw;
 475
 476             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 477
 478             /* Update vectorial force */
 479             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 480             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 481             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 482
 483             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 484             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 485             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             /* Compute parameters for interactions between i and j atoms */
 492             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 493
 494             /* COULOMB ELECTROSTATICS */
 495             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 496             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 500             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 501
 502             fscal            = felec;
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 508             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 509             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             /* Compute parameters for interactions between i and j atoms */
 520             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 521
 522             /* COULOMB ELECTROSTATICS */
 523             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 524             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 525
 526             /* Update potential sum for this i atom from the interaction with this j atom. */
 527             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 528             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 533
 534             /* Update vectorial force */
 535             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 536             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 537             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 538
 539             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 540             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 541             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 542
 543             /**************************
 544              * CALCULATE INTERACTIONS *
 545              **************************/
 546
 547             /* Compute parameters for interactions between i and j atoms */
 548             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 549
 550             /* COULOMB ELECTROSTATICS */
 551             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 552             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 556             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 561
 562             /* Update vectorial force */
 563             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 564             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 565             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 566
 567             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 568             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 569             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 570
 571             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 572
 573             /* Inner loop uses 155 flops */
 574         }
 575
 576         /* End of innermost loop */
 577
 578         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 579                                               f+i_coord_offset,fshift+i_shift_offset);
 580
 581         ggid                        = gid[iidx];
 582         /* Update potential energies */
 583         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 584         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 585
 586         /* Increment number of inner iterations */
 587         inneriter                  += j_index_end - j_index_start;
 588
 589         /* Outer loop uses 26 flops */
 590     }
 591
 592     /* Increment number of outer iterations */
 593     outeriter        += nri;
 594
 595     /* Update outer/inner flops */
 596
 597     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*155);
 598 }
 599 /*
 600  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 601  * Electrostatics interaction: Coulomb
 602  * VdW interaction:            CubicSplineTable
 603  * Geometry:                   Water4-Particle
 604  * Calculate force/pot:        Force
 605  */
 606 void
 607 nb_kernel_ElecCoul_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 608                     (t_nblist                    * gmx_restrict       nlist,
 609                      rvec                        * gmx_restrict          xx,
 610                      rvec                        * gmx_restrict          ff,
 611                      t_forcerec                  * gmx_restrict          fr,
 612                      t_mdatoms                   * gmx_restrict     mdatoms,
 613                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 614                      t_nrnb                      * gmx_restrict        nrnb)
 615 {
 616     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 617      * just 0 for non-waters.
 618      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 619      * jnr indices corresponding to data put in the four positions in the SIMD register.
 620      */
 621     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 622     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 623     int              jnrA,jnrB;
 624     int              j_coord_offsetA,j_coord_offsetB;
 625     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 626     real             rcutoff_scalar;
 627     real             *shiftvec,*fshift,*x,*f;
 628     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 629     int              vdwioffset0;
 630     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 631     int              vdwioffset1;
 632     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 633     int              vdwioffset2;
 634     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 635     int              vdwioffset3;
 636     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 637     int              vdwjidx0A,vdwjidx0B;
 638     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 639     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 640     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 641     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 642     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 643     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 644     real             *charge;
 645     int              nvdwtype;
 646     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 647     int              *vdwtype;
 648     real             *vdwparam;
 649     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 650     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 651     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 652     real             *vftab;
 653     _fjsp_v2r8       itab_tmp;
 654     _fjsp_v2r8       dummy_mask,cutoff_mask;
 655     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 656     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 657     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 658
 659     x                = xx[0];
 660     f                = ff[0];
 661
 662     nri              = nlist->nri;
 663     iinr             = nlist->iinr;
 664     jindex           = nlist->jindex;
 665     jjnr             = nlist->jjnr;
 666     shiftidx         = nlist->shift;
 667     gid              = nlist->gid;
 668     shiftvec         = fr->shift_vec[0];
 669     fshift           = fr->fshift[0];
 670     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 671     charge           = mdatoms->chargeA;
 672     nvdwtype         = fr->ntype;
 673     vdwparam         = fr->nbfp;
 674     vdwtype          = mdatoms->typeA;
 675
 676     vftab            = kernel_data->table_vdw->data;
 677     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 678
 679     /* Setup water-specific parameters */
 680     inr              = nlist->iinr[0];
 681     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 682     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 683     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 684     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 685
 686     /* Avoid stupid compiler warnings */
 687     jnrA = jnrB = 0;
 688     j_coord_offsetA = 0;
 689     j_coord_offsetB = 0;
 690
 691     outeriter        = 0;
 692     inneriter        = 0;
 693
 694     /* Start outer loop over neighborlists */
 695     for(iidx=0; iidx<nri; iidx++)
 696     {
 697         /* Load shift vector for this list */
 698         i_shift_offset   = DIM*shiftidx[iidx];
 699
 700         /* Load limits for loop over neighbors */
 701         j_index_start    = jindex[iidx];
 702         j_index_end      = jindex[iidx+1];
 703
 704         /* Get outer coordinate index */
 705         inr              = iinr[iidx];
 706         i_coord_offset   = DIM*inr;
 707
 708         /* Load i particle coords and add shift vector */
 709         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 710                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 711
 712         fix0             = _fjsp_setzero_v2r8();
 713         fiy0             = _fjsp_setzero_v2r8();
 714         fiz0             = _fjsp_setzero_v2r8();
 715         fix1             = _fjsp_setzero_v2r8();
 716         fiy1             = _fjsp_setzero_v2r8();
 717         fiz1             = _fjsp_setzero_v2r8();
 718         fix2             = _fjsp_setzero_v2r8();
 719         fiy2             = _fjsp_setzero_v2r8();
 720         fiz2             = _fjsp_setzero_v2r8();
 721         fix3             = _fjsp_setzero_v2r8();
 722         fiy3             = _fjsp_setzero_v2r8();
 723         fiz3             = _fjsp_setzero_v2r8();
 724
 725         /* Start inner kernel loop */
 726         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 727         {
 728
 729             /* Get j neighbor index, and coordinate index */
 730             jnrA             = jjnr[jidx];
 731             jnrB             = jjnr[jidx+1];
 732             j_coord_offsetA  = DIM*jnrA;
 733             j_coord_offsetB  = DIM*jnrB;
 734
 735             /* load j atom coordinates */
 736             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 737                                               &jx0,&jy0,&jz0);
 738
 739             /* Calculate displacement vector */
 740             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 741             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 742             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 743             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 744             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 745             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 746             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 747             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 748             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 749             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 750             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 751             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 752
 753             /* Calculate squared distance and things based on it */
 754             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 755             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 756             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 757             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 758
 759             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 760             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 761             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 762             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 763
 764             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 765             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 766             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 767
 768             /* Load parameters for j particles */
 769             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 770             vdwjidx0A        = 2*vdwtype[jnrA+0];
 771             vdwjidx0B        = 2*vdwtype[jnrB+0];
 772
 773             fjx0             = _fjsp_setzero_v2r8();
 774             fjy0             = _fjsp_setzero_v2r8();
 775             fjz0             = _fjsp_setzero_v2r8();
 776
 777             /**************************
 778              * CALCULATE INTERACTIONS *
 779              **************************/
 780
 781             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 782
 783             /* Compute parameters for interactions between i and j atoms */
 784             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 785                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 786
 787             /* Calculate table index by multiplying r with table scale and truncate to integer */
 788             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 789             itab_tmp         = _fjsp_dtox_v2r8(rt);
 790             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 791             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 792             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 793
 794             vfconv.i[0]     *= 8;
 795             vfconv.i[1]     *= 8;
 796
 797             /* CUBIC SPLINE TABLE DISPERSION */
 798             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 799             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 800             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 801             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 802             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 803             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 804             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 805             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 806             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 807
 808             /* CUBIC SPLINE TABLE REPULSION */
 809             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 810             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 811             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 812             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 813             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 814             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 815             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 816             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 817             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 818             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 819
 820             fscal            = fvdw;
 821
 822             /* Update vectorial force */
 823             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 824             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 825             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 826
 827             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 828             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 829             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 830
 831             /**************************
 832              * CALCULATE INTERACTIONS *
 833              **************************/
 834
 835             /* Compute parameters for interactions between i and j atoms */
 836             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 837
 838             /* COULOMB ELECTROSTATICS */
 839             velec            = _fjsp_mul_v2r8(qq10,rinv10);
 840             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
 841
 842             fscal            = felec;
 843
 844             /* Update vectorial force */
 845             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 846             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 847             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 848
 849             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 850             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 851             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             /* Compute parameters for interactions between i and j atoms */
 858             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 859
 860             /* COULOMB ELECTROSTATICS */
 861             velec            = _fjsp_mul_v2r8(qq20,rinv20);
 862             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
 863
 864             fscal            = felec;
 865
 866             /* Update vectorial force */
 867             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 868             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 869             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 870
 871             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 872             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 873             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 874
 875             /**************************
 876              * CALCULATE INTERACTIONS *
 877              **************************/
 878
 879             /* Compute parameters for interactions between i and j atoms */
 880             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 881
 882             /* COULOMB ELECTROSTATICS */
 883             velec            = _fjsp_mul_v2r8(qq30,rinv30);
 884             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
 885
 886             fscal            = felec;
 887
 888             /* Update vectorial force */
 889             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 890             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 891             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 892
 893             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 894             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 895             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 896
 897             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 898
 899             /* Inner loop uses 144 flops */
 900         }
 901
 902         if(jidx<j_index_end)
 903         {
 904
 905             jnrA             = jjnr[jidx];
 906             j_coord_offsetA  = DIM*jnrA;
 907
 908             /* load j atom coordinates */
 909             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 910                                               &jx0,&jy0,&jz0);
 911
 912             /* Calculate displacement vector */
 913             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 914             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 915             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 916             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 917             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 918             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 919             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 920             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 921             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 922             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 923             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 924             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 925
 926             /* Calculate squared distance and things based on it */
 927             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 928             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 929             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 930             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 931
 932             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 933             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 934             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 935             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 936
 937             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 938             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 939             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 940
 941             /* Load parameters for j particles */
 942             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 943             vdwjidx0A        = 2*vdwtype[jnrA+0];
 944
 945             fjx0             = _fjsp_setzero_v2r8();
 946             fjy0             = _fjsp_setzero_v2r8();
 947             fjz0             = _fjsp_setzero_v2r8();
 948
 949             /**************************
 950              * CALCULATE INTERACTIONS *
 951              **************************/
 952
 953             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 954
 955             /* Compute parameters for interactions between i and j atoms */
 956             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 957                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 958
 959             /* Calculate table index by multiplying r with table scale and truncate to integer */
 960             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 961             itab_tmp         = _fjsp_dtox_v2r8(rt);
 962             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 963             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 964             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 965
 966             vfconv.i[0]     *= 8;
 967             vfconv.i[1]     *= 8;
 968
 969             /* CUBIC SPLINE TABLE DISPERSION */
 970             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 971             F                = _fjsp_setzero_v2r8();
 972             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 973             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 974             H                = _fjsp_setzero_v2r8();
 975             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 976             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 977             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 978             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 979
 980             /* CUBIC SPLINE TABLE REPULSION */
 981             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 982             F                = _fjsp_setzero_v2r8();
 983             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 984             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 985             H                = _fjsp_setzero_v2r8();
 986             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 987             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 988             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 989             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 990             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 991
 992             fscal            = fvdw;
 993
 994             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 995
 996             /* Update vectorial force */
 997             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 998             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 999             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1000
1001             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1002             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1003             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1004
1005             /**************************
1006              * CALCULATE INTERACTIONS *
1007              **************************/
1008
1009             /* Compute parameters for interactions between i and j atoms */
1010             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1011
1012             /* COULOMB ELECTROSTATICS */
1013             velec            = _fjsp_mul_v2r8(qq10,rinv10);
1014             felec            = _fjsp_mul_v2r8(velec,rinvsq10);
1015
1016             fscal            = felec;
1017
1018             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1019
1020             /* Update vectorial force */
1021             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1022             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1023             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1024
1025             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1026             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1027             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1028
1029             /**************************
1030              * CALCULATE INTERACTIONS *
1031              **************************/
1032
1033             /* Compute parameters for interactions between i and j atoms */
1034             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1035
1036             /* COULOMB ELECTROSTATICS */
1037             velec            = _fjsp_mul_v2r8(qq20,rinv20);
1038             felec            = _fjsp_mul_v2r8(velec,rinvsq20);
1039
1040             fscal            = felec;
1041
1042             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1043
1044             /* Update vectorial force */
1045             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1046             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1047             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1048
1049             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1050             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1051             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1052
1053             /**************************
1054              * CALCULATE INTERACTIONS *
1055              **************************/
1056
1057             /* Compute parameters for interactions between i and j atoms */
1058             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1059
1060             /* COULOMB ELECTROSTATICS */
1061             velec            = _fjsp_mul_v2r8(qq30,rinv30);
1062             felec            = _fjsp_mul_v2r8(velec,rinvsq30);
1063
1064             fscal            = felec;
1065
1066             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1067
1068             /* Update vectorial force */
1069             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1070             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1071             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1072
1073             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1074             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1075             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1076
1077             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1078
1079             /* Inner loop uses 144 flops */
1080         }
1081
1082         /* End of innermost loop */
1083
1084         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1085                                               f+i_coord_offset,fshift+i_shift_offset);
1086
1087         /* Increment number of inner iterations */
1088         inneriter                  += j_index_end - j_index_start;
1089
1090         /* Outer loop uses 24 flops */
1091     }
1092
1093     /* Increment number of outer iterations */
1094     outeriter        += nri;
1095
1096     /* Update outer/inner flops */
1097
1098     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1099 }