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, by the GROMACS development team, led by
   5  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   6  * others, as listed in the AUTHORS file in the top-level source
   7  * directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_VdwLJ_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_elec->data;
 129     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->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             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 216             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 217             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 218
 219             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 220
 221             /* Load parameters for j particles */
 222             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 223             vdwjidx0A        = 2*vdwtype[jnrA+0];
 224             vdwjidx0B        = 2*vdwtype[jnrB+0];
 225
 226             fjx0             = _fjsp_setzero_v2r8();
 227             fjy0             = _fjsp_setzero_v2r8();
 228             fjz0             = _fjsp_setzero_v2r8();
 229
 230             /**************************
 231              * CALCULATE INTERACTIONS *
 232              **************************/
 233
 234             /* Compute parameters for interactions between i and j atoms */
 235             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 236                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 237
 238             /* LENNARD-JONES DISPERSION/REPULSION */
 239
 240             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 241             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 242             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 243             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 244             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 245
 246             /* Update potential sum for this i atom from the interaction with this j atom. */
 247             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 248
 249             fscal            = fvdw;
 250
 251             /* Update vectorial force */
 252             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 253             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 254             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 255
 256             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 257             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 258             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 259
 260             /**************************
 261              * CALCULATE INTERACTIONS *
 262              **************************/
 263
 264             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 265
 266             /* Compute parameters for interactions between i and j atoms */
 267             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 268
 269             /* Calculate table index by multiplying r with table scale and truncate to integer */
 270             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 271             itab_tmp         = _fjsp_dtox_v2r8(rt);
 272             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 273             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 274             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 275
 276             vfconv.i[0]     *= 4;
 277             vfconv.i[1]     *= 4;
 278
 279             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 280             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 281             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 282             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 283             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 284             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 285             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 286             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 287             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 288             velec            = _fjsp_mul_v2r8(qq10,VV);
 289             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 290             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 291
 292             /* Update potential sum for this i atom from the interaction with this j atom. */
 293             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 294
 295             fscal            = felec;
 296
 297             /* Update vectorial force */
 298             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 299             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 300             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 301
 302             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 303             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 304             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 305
 306             /**************************
 307              * CALCULATE INTERACTIONS *
 308              **************************/
 309
 310             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 311
 312             /* Compute parameters for interactions between i and j atoms */
 313             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 314
 315             /* Calculate table index by multiplying r with table scale and truncate to integer */
 316             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 317             itab_tmp         = _fjsp_dtox_v2r8(rt);
 318             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 319             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 320             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 321
 322             vfconv.i[0]     *= 4;
 323             vfconv.i[1]     *= 4;
 324
 325             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 326             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 327             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 328             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 329             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 330             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 331             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 332             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 333             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 334             velec            = _fjsp_mul_v2r8(qq20,VV);
 335             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 336             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             /* Update vectorial force */
 344             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 345             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 346             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 347
 348             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 349             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 350             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 354              **************************/
 355
 356             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 360
 361             /* Calculate table index by multiplying r with table scale and truncate to integer */
 362             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 363             itab_tmp         = _fjsp_dtox_v2r8(rt);
 364             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 365             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 366             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 367
 368             vfconv.i[0]     *= 4;
 369             vfconv.i[1]     *= 4;
 370
 371             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 372             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 373             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 374             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 375             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 376             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 377             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 378             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 379             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 380             velec            = _fjsp_mul_v2r8(qq30,VV);
 381             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 382             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 386
 387             fscal            = felec;
 388
 389             /* Update vectorial force */
 390             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 391             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 392             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 393
 394             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 395             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 396             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 397
 398             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 399
 400             /* Inner loop uses 176 flops */
 401         }
 402
 403         if(jidx<j_index_end)
 404         {
 405
 406             jnrA             = jjnr[jidx];
 407             j_coord_offsetA  = DIM*jnrA;
 408
 409             /* load j atom coordinates */
 410             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 411                                               &jx0,&jy0,&jz0);
 412
 413             /* Calculate displacement vector */
 414             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 415             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 416             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 417             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 418             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 419             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 420             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 421             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 422             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 423             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 424             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 425             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 426
 427             /* Calculate squared distance and things based on it */
 428             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 429             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 430             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 431             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 432
 433             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 434             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 435             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 436
 437             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 438
 439             /* Load parameters for j particles */
 440             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 441             vdwjidx0A        = 2*vdwtype[jnrA+0];
 442
 443             fjx0             = _fjsp_setzero_v2r8();
 444             fjy0             = _fjsp_setzero_v2r8();
 445             fjz0             = _fjsp_setzero_v2r8();
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 453
 454             /* LENNARD-JONES DISPERSION/REPULSION */
 455
 456             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 457             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 458             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 459             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 460             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 461
 462             /* Update potential sum for this i atom from the interaction with this j atom. */
 463             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 464             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 465
 466             fscal            = fvdw;
 467
 468             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 469
 470             /* Update vectorial force */
 471             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 472             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 473             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 474
 475             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 476             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 477             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 484
 485             /* Compute parameters for interactions between i and j atoms */
 486             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 487
 488             /* Calculate table index by multiplying r with table scale and truncate to integer */
 489             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 490             itab_tmp         = _fjsp_dtox_v2r8(rt);
 491             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 492             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 493             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 494
 495             vfconv.i[0]     *= 4;
 496             vfconv.i[1]     *= 4;
 497
 498             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 499             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 500             F                = _fjsp_setzero_v2r8();
 501             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 502             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 503             H                = _fjsp_setzero_v2r8();
 504             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 505             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 506             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 507             velec            = _fjsp_mul_v2r8(qq10,VV);
 508             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 509             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 510
 511             /* Update potential sum for this i atom from the interaction with this j atom. */
 512             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 513             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 514
 515             fscal            = felec;
 516
 517             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 518
 519             /* Update vectorial force */
 520             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 521             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 522             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 523
 524             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 525             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 526             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 527
 528             /**************************
 529              * CALCULATE INTERACTIONS *
 530              **************************/
 531
 532             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 533
 534             /* Compute parameters for interactions between i and j atoms */
 535             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 536
 537             /* Calculate table index by multiplying r with table scale and truncate to integer */
 538             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 539             itab_tmp         = _fjsp_dtox_v2r8(rt);
 540             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 541             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 542             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 543
 544             vfconv.i[0]     *= 4;
 545             vfconv.i[1]     *= 4;
 546
 547             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 548             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 549             F                = _fjsp_setzero_v2r8();
 550             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 551             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 552             H                = _fjsp_setzero_v2r8();
 553             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 554             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 555             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 556             velec            = _fjsp_mul_v2r8(qq20,VV);
 557             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 558             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 559
 560             /* Update potential sum for this i atom from the interaction with this j atom. */
 561             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 562             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 563
 564             fscal            = felec;
 565
 566             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 567
 568             /* Update vectorial force */
 569             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 570             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 571             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 572
 573             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 574             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 575             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 576
 577             /**************************
 578              * CALCULATE INTERACTIONS *
 579              **************************/
 580
 581             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 582
 583             /* Compute parameters for interactions between i and j atoms */
 584             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 585
 586             /* Calculate table index by multiplying r with table scale and truncate to integer */
 587             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 588             itab_tmp         = _fjsp_dtox_v2r8(rt);
 589             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 590             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 591             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 592
 593             vfconv.i[0]     *= 4;
 594             vfconv.i[1]     *= 4;
 595
 596             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 597             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 598             F                = _fjsp_setzero_v2r8();
 599             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 600             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 601             H                = _fjsp_setzero_v2r8();
 602             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 603             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 604             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 605             velec            = _fjsp_mul_v2r8(qq30,VV);
 606             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 607             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 608
 609             /* Update potential sum for this i atom from the interaction with this j atom. */
 610             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 611             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 612
 613             fscal            = felec;
 614
 615             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 616
 617             /* Update vectorial force */
 618             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 619             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 620             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 621
 622             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 623             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 624             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 625
 626             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 627
 628             /* Inner loop uses 176 flops */
 629         }
 630
 631         /* End of innermost loop */
 632
 633         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 634                                               f+i_coord_offset,fshift+i_shift_offset);
 635
 636         ggid                        = gid[iidx];
 637         /* Update potential energies */
 638         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 639         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 640
 641         /* Increment number of inner iterations */
 642         inneriter                  += j_index_end - j_index_start;
 643
 644         /* Outer loop uses 26 flops */
 645     }
 646
 647     /* Increment number of outer iterations */
 648     outeriter        += nri;
 649
 650     /* Update outer/inner flops */
 651
 652     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*176);
 653 }
 654 /*
 655  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 656  * Electrostatics interaction: CubicSplineTable
 657  * VdW interaction:            LennardJones
 658  * Geometry:                   Water4-Particle
 659  * Calculate force/pot:        Force
 660  */
 661 void
 662 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 663                     (t_nblist                    * gmx_restrict       nlist,
 664                      rvec                        * gmx_restrict          xx,
 665                      rvec                        * gmx_restrict          ff,
 666                      t_forcerec                  * gmx_restrict          fr,
 667                      t_mdatoms                   * gmx_restrict     mdatoms,
 668                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 669                      t_nrnb                      * gmx_restrict        nrnb)
 670 {
 671     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 672      * just 0 for non-waters.
 673      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 674      * jnr indices corresponding to data put in the four positions in the SIMD register.
 675      */
 676     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 677     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 678     int              jnrA,jnrB;
 679     int              j_coord_offsetA,j_coord_offsetB;
 680     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 681     real             rcutoff_scalar;
 682     real             *shiftvec,*fshift,*x,*f;
 683     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 684     int              vdwioffset0;
 685     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 686     int              vdwioffset1;
 687     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 688     int              vdwioffset2;
 689     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 690     int              vdwioffset3;
 691     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 692     int              vdwjidx0A,vdwjidx0B;
 693     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 694     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 695     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 696     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 697     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 698     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 699     real             *charge;
 700     int              nvdwtype;
 701     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 702     int              *vdwtype;
 703     real             *vdwparam;
 704     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 705     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 706     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 707     real             *vftab;
 708     _fjsp_v2r8       itab_tmp;
 709     _fjsp_v2r8       dummy_mask,cutoff_mask;
 710     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 711     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 712     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 713
 714     x                = xx[0];
 715     f                = ff[0];
 716
 717     nri              = nlist->nri;
 718     iinr             = nlist->iinr;
 719     jindex           = nlist->jindex;
 720     jjnr             = nlist->jjnr;
 721     shiftidx         = nlist->shift;
 722     gid              = nlist->gid;
 723     shiftvec         = fr->shift_vec[0];
 724     fshift           = fr->fshift[0];
 725     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 726     charge           = mdatoms->chargeA;
 727     nvdwtype         = fr->ntype;
 728     vdwparam         = fr->nbfp;
 729     vdwtype          = mdatoms->typeA;
 730
 731     vftab            = kernel_data->table_elec->data;
 732     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 733
 734     /* Setup water-specific parameters */
 735     inr              = nlist->iinr[0];
 736     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 737     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 738     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 739     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 740
 741     /* Avoid stupid compiler warnings */
 742     jnrA = jnrB = 0;
 743     j_coord_offsetA = 0;
 744     j_coord_offsetB = 0;
 745
 746     outeriter        = 0;
 747     inneriter        = 0;
 748
 749     /* Start outer loop over neighborlists */
 750     for(iidx=0; iidx<nri; iidx++)
 751     {
 752         /* Load shift vector for this list */
 753         i_shift_offset   = DIM*shiftidx[iidx];
 754
 755         /* Load limits for loop over neighbors */
 756         j_index_start    = jindex[iidx];
 757         j_index_end      = jindex[iidx+1];
 758
 759         /* Get outer coordinate index */
 760         inr              = iinr[iidx];
 761         i_coord_offset   = DIM*inr;
 762
 763         /* Load i particle coords and add shift vector */
 764         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 765                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 766
 767         fix0             = _fjsp_setzero_v2r8();
 768         fiy0             = _fjsp_setzero_v2r8();
 769         fiz0             = _fjsp_setzero_v2r8();
 770         fix1             = _fjsp_setzero_v2r8();
 771         fiy1             = _fjsp_setzero_v2r8();
 772         fiz1             = _fjsp_setzero_v2r8();
 773         fix2             = _fjsp_setzero_v2r8();
 774         fiy2             = _fjsp_setzero_v2r8();
 775         fiz2             = _fjsp_setzero_v2r8();
 776         fix3             = _fjsp_setzero_v2r8();
 777         fiy3             = _fjsp_setzero_v2r8();
 778         fiz3             = _fjsp_setzero_v2r8();
 779
 780         /* Start inner kernel loop */
 781         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 782         {
 783
 784             /* Get j neighbor index, and coordinate index */
 785             jnrA             = jjnr[jidx];
 786             jnrB             = jjnr[jidx+1];
 787             j_coord_offsetA  = DIM*jnrA;
 788             j_coord_offsetB  = DIM*jnrB;
 789
 790             /* load j atom coordinates */
 791             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 792                                               &jx0,&jy0,&jz0);
 793
 794             /* Calculate displacement vector */
 795             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 796             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 797             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 798             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 799             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 800             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 801             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 802             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 803             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 804             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 805             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 806             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 807
 808             /* Calculate squared distance and things based on it */
 809             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 810             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 811             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 812             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 813
 814             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 815             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 816             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 817
 818             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 819
 820             /* Load parameters for j particles */
 821             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 822             vdwjidx0A        = 2*vdwtype[jnrA+0];
 823             vdwjidx0B        = 2*vdwtype[jnrB+0];
 824
 825             fjx0             = _fjsp_setzero_v2r8();
 826             fjy0             = _fjsp_setzero_v2r8();
 827             fjz0             = _fjsp_setzero_v2r8();
 828
 829             /**************************
 830              * CALCULATE INTERACTIONS *
 831              **************************/
 832
 833             /* Compute parameters for interactions between i and j atoms */
 834             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 835                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 836
 837             /* LENNARD-JONES DISPERSION/REPULSION */
 838
 839             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 840             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 841
 842             fscal            = fvdw;
 843
 844             /* Update vectorial force */
 845             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 846             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 847             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 848
 849             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 850             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 851             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 858
 859             /* Compute parameters for interactions between i and j atoms */
 860             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 861
 862             /* Calculate table index by multiplying r with table scale and truncate to integer */
 863             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 864             itab_tmp         = _fjsp_dtox_v2r8(rt);
 865             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 866             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 867             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 868
 869             vfconv.i[0]     *= 4;
 870             vfconv.i[1]     *= 4;
 871
 872             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 873             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 874             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 875             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 876             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 877             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 878             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 879             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 880             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 881             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 882
 883             fscal            = felec;
 884
 885             /* Update vectorial force */
 886             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 887             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 888             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 889
 890             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 891             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 892             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 899
 900             /* Compute parameters for interactions between i and j atoms */
 901             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 902
 903             /* Calculate table index by multiplying r with table scale and truncate to integer */
 904             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 905             itab_tmp         = _fjsp_dtox_v2r8(rt);
 906             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 907             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 908             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 909
 910             vfconv.i[0]     *= 4;
 911             vfconv.i[1]     *= 4;
 912
 913             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 914             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 915             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 916             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 917             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 918             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 919             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 920             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 921             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 922             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 923
 924             fscal            = felec;
 925
 926             /* Update vectorial force */
 927             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 928             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 929             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 930
 931             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 932             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 933             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 934
 935             /**************************
 936              * CALCULATE INTERACTIONS *
 937              **************************/
 938
 939             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 940
 941             /* Compute parameters for interactions between i and j atoms */
 942             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 943
 944             /* Calculate table index by multiplying r with table scale and truncate to integer */
 945             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 946             itab_tmp         = _fjsp_dtox_v2r8(rt);
 947             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 948             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 949             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 950
 951             vfconv.i[0]     *= 4;
 952             vfconv.i[1]     *= 4;
 953
 954             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 955             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 956             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 957             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 958             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 959             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 960             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 961             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 962             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 963             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 964
 965             fscal            = felec;
 966
 967             /* Update vectorial force */
 968             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 969             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 970             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 971
 972             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 973             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 974             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 975
 976             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 977
 978             /* Inner loop uses 159 flops */
 979         }
 980
 981         if(jidx<j_index_end)
 982         {
 983
 984             jnrA             = jjnr[jidx];
 985             j_coord_offsetA  = DIM*jnrA;
 986
 987             /* load j atom coordinates */
 988             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 989                                               &jx0,&jy0,&jz0);
 990
 991             /* Calculate displacement vector */
 992             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 993             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 994             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 995             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 996             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 997             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 998             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 999             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1000             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1001             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1002             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1003             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1004
1005             /* Calculate squared distance and things based on it */
1006             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1007             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1008             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1009             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1010
1011             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1012             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1013             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1014
1015             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
1016
1017             /* Load parameters for j particles */
1018             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1019             vdwjidx0A        = 2*vdwtype[jnrA+0];
1020
1021             fjx0             = _fjsp_setzero_v2r8();
1022             fjy0             = _fjsp_setzero_v2r8();
1023             fjz0             = _fjsp_setzero_v2r8();
1024
1025             /**************************
1026              * CALCULATE INTERACTIONS *
1027              **************************/
1028
1029             /* Compute parameters for interactions between i and j atoms */
1030             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&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 }