src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_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_elec_vdw->data;
 129     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_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             /* Load parameters for j particles */
 221             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 222             vdwjidx0A        = 2*vdwtype[jnrA+0];
 223             vdwjidx0B        = 2*vdwtype[jnrB+0];
 224
 225             fjx0             = _fjsp_setzero_v2r8();
 226             fjy0             = _fjsp_setzero_v2r8();
 227             fjz0             = _fjsp_setzero_v2r8();
 228
 229             /**************************
 230              * CALCULATE INTERACTIONS *
 231              **************************/
 232
 233             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 234
 235             /* Compute parameters for interactions between i and j atoms */
 236             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 237                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 238
 239             /* Calculate table index by multiplying r with table scale and truncate to integer */
 240             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 241             itab_tmp         = _fjsp_dtox_v2r8(rt);
 242             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 243             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 244             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 245
 246             vfconv.i[0]     *= 12;
 247             vfconv.i[1]     *= 12;
 248
 249             /* CUBIC SPLINE TABLE DISPERSION */
 250             vfconv.i[0]       += 4;
 251             vfconv.i[1]       += 4;
 252             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 253             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 254             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 255             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 256             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 257             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 258             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 259             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 260             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 261             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 262             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 263
 264             /* CUBIC SPLINE TABLE REPULSION */
 265             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 266             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 267             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 268             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 269             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 271             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 272             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 273             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 274             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 275             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 276             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 277             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 278
 279             /* Update potential sum for this i atom from the interaction with this j atom. */
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = fvdw;
 283
 284             /* Update vectorial force */
 285             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 286             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 287             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 288
 289             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 290             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 291             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 298
 299             /* Compute parameters for interactions between i and j atoms */
 300             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 301
 302             /* Calculate table index by multiplying r with table scale and truncate to integer */
 303             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 304             itab_tmp         = _fjsp_dtox_v2r8(rt);
 305             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 306             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 307             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 308
 309             vfconv.i[0]     *= 12;
 310             vfconv.i[1]     *= 12;
 311
 312             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 313             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 314             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 315             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 316             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 317             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 318             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 319             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 320             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 321             velec            = _fjsp_mul_v2r8(qq10,VV);
 322             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 323             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 324
 325             /* Update potential sum for this i atom from the interaction with this j atom. */
 326             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 327
 328             fscal            = felec;
 329
 330             /* Update vectorial force */
 331             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 332             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 333             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 334
 335             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 336             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 337             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 338
 339             /**************************
 340              * CALCULATE INTERACTIONS *
 341              **************************/
 342
 343             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 344
 345             /* Compute parameters for interactions between i and j atoms */
 346             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 347
 348             /* Calculate table index by multiplying r with table scale and truncate to integer */
 349             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 350             itab_tmp         = _fjsp_dtox_v2r8(rt);
 351             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 352             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 353             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 354
 355             vfconv.i[0]     *= 12;
 356             vfconv.i[1]     *= 12;
 357
 358             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 359             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 360             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 361             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 362             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 363             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 364             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 365             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 366             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 367             velec            = _fjsp_mul_v2r8(qq20,VV);
 368             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 369             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 370
 371             /* Update potential sum for this i atom from the interaction with this j atom. */
 372             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 373
 374             fscal            = felec;
 375
 376             /* Update vectorial force */
 377             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 378             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 379             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 380
 381             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 382             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 383             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 384
 385             /**************************
 386              * CALCULATE INTERACTIONS *
 387              **************************/
 388
 389             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 390
 391             /* Compute parameters for interactions between i and j atoms */
 392             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 393
 394             /* Calculate table index by multiplying r with table scale and truncate to integer */
 395             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 396             itab_tmp         = _fjsp_dtox_v2r8(rt);
 397             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 398             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 399             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 400
 401             vfconv.i[0]     *= 12;
 402             vfconv.i[1]     *= 12;
 403
 404             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 405             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 406             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 407             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 408             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 409             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 410             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 411             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 412             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 413             velec            = _fjsp_mul_v2r8(qq30,VV);
 414             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 415             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 416
 417             /* Update potential sum for this i atom from the interaction with this j atom. */
 418             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             /* Update vectorial force */
 423             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 424             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 425             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 426
 427             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 428             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 429             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 430
 431             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 432
 433             /* Inner loop uses 200 flops */
 434         }
 435
 436         if(jidx<j_index_end)
 437         {
 438
 439             jnrA             = jjnr[jidx];
 440             j_coord_offsetA  = DIM*jnrA;
 441
 442             /* load j atom coordinates */
 443             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 444                                               &jx0,&jy0,&jz0);
 445
 446             /* Calculate displacement vector */
 447             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 448             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 449             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 450             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 451             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 452             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 453             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 454             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 455             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 456             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 457             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 458             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 459
 460             /* Calculate squared distance and things based on it */
 461             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 462             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 463             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 464             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 465
 466             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 467             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 468             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 469             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 470
 471             /* Load parameters for j particles */
 472             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 473             vdwjidx0A        = 2*vdwtype[jnrA+0];
 474
 475             fjx0             = _fjsp_setzero_v2r8();
 476             fjy0             = _fjsp_setzero_v2r8();
 477             fjz0             = _fjsp_setzero_v2r8();
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 484
 485             /* Compute parameters for interactions between i and j atoms */
 486             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 487
 488             /* Calculate table index by multiplying r with table scale and truncate to integer */
 489             rt               = _fjsp_mul_v2r8(r00,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]     *= 12;
 496             vfconv.i[1]     *= 12;
 497
 498             /* CUBIC SPLINE TABLE DISPERSION */
 499             vfconv.i[0]       += 4;
 500             vfconv.i[1]       += 4;
 501             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 502             F                = _fjsp_setzero_v2r8();
 503             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 504             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 505             H                = _fjsp_setzero_v2r8();
 506             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 507             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 508             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 509             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 510             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 511             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 512
 513             /* CUBIC SPLINE TABLE REPULSION */
 514             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 515             F                = _fjsp_setzero_v2r8();
 516             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 517             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 518             H                = _fjsp_setzero_v2r8();
 519             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 520             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 521             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 522             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 523             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 524             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 525             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 526             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 527
 528             /* Update potential sum for this i atom from the interaction with this j atom. */
 529             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 530             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 531
 532             fscal            = fvdw;
 533
 534             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 535
 536             /* Update vectorial force */
 537             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 538             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 539             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 540
 541             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 542             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 543             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 544
 545             /**************************
 546              * CALCULATE INTERACTIONS *
 547              **************************/
 548
 549             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 550
 551             /* Compute parameters for interactions between i and j atoms */
 552             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 553
 554             /* Calculate table index by multiplying r with table scale and truncate to integer */
 555             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 556             itab_tmp         = _fjsp_dtox_v2r8(rt);
 557             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 558             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 559             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 560
 561             vfconv.i[0]     *= 12;
 562             vfconv.i[1]     *= 12;
 563
 564             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 565             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 566             F                = _fjsp_setzero_v2r8();
 567             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 568             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 569             H                = _fjsp_setzero_v2r8();
 570             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 571             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 572             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 573             velec            = _fjsp_mul_v2r8(qq10,VV);
 574             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 575             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 576
 577             /* Update potential sum for this i atom from the interaction with this j atom. */
 578             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 579             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 580
 581             fscal            = felec;
 582
 583             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 584
 585             /* Update vectorial force */
 586             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 587             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 588             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 589
 590             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 591             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 592             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 599
 600             /* Compute parameters for interactions between i and j atoms */
 601             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 602
 603             /* Calculate table index by multiplying r with table scale and truncate to integer */
 604             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 605             itab_tmp         = _fjsp_dtox_v2r8(rt);
 606             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 607             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 608             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 609
 610             vfconv.i[0]     *= 12;
 611             vfconv.i[1]     *= 12;
 612
 613             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 614             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 615             F                = _fjsp_setzero_v2r8();
 616             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 617             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 618             H                = _fjsp_setzero_v2r8();
 619             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 620             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 621             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 622             velec            = _fjsp_mul_v2r8(qq20,VV);
 623             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 624             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 625
 626             /* Update potential sum for this i atom from the interaction with this j atom. */
 627             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 628             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 629
 630             fscal            = felec;
 631
 632             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 633
 634             /* Update vectorial force */
 635             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 636             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 637             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 638
 639             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 640             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 641             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 642
 643             /**************************
 644              * CALCULATE INTERACTIONS *
 645              **************************/
 646
 647             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 648
 649             /* Compute parameters for interactions between i and j atoms */
 650             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 651
 652             /* Calculate table index by multiplying r with table scale and truncate to integer */
 653             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 654             itab_tmp         = _fjsp_dtox_v2r8(rt);
 655             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 656             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 657             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 658
 659             vfconv.i[0]     *= 12;
 660             vfconv.i[1]     *= 12;
 661
 662             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 663             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 664             F                = _fjsp_setzero_v2r8();
 665             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 666             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 667             H                = _fjsp_setzero_v2r8();
 668             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 669             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 670             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 671             velec            = _fjsp_mul_v2r8(qq30,VV);
 672             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 673             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 674
 675             /* Update potential sum for this i atom from the interaction with this j atom. */
 676             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 677             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 678
 679             fscal            = felec;
 680
 681             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 682
 683             /* Update vectorial force */
 684             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 685             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 686             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 687
 688             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 689             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 690             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 691
 692             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 693
 694             /* Inner loop uses 200 flops */
 695         }
 696
 697         /* End of innermost loop */
 698
 699         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 700                                               f+i_coord_offset,fshift+i_shift_offset);
 701
 702         ggid                        = gid[iidx];
 703         /* Update potential energies */
 704         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 705         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 706
 707         /* Increment number of inner iterations */
 708         inneriter                  += j_index_end - j_index_start;
 709
 710         /* Outer loop uses 26 flops */
 711     }
 712
 713     /* Increment number of outer iterations */
 714     outeriter        += nri;
 715
 716     /* Update outer/inner flops */
 717
 718     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*200);
 719 }
 720 /*
 721  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 722  * Electrostatics interaction: CubicSplineTable
 723  * VdW interaction:            CubicSplineTable
 724  * Geometry:                   Water4-Particle
 725  * Calculate force/pot:        Force
 726  */
 727 void
 728 nb_kernel_ElecCSTab_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 729                     (t_nblist                    * gmx_restrict       nlist,
 730                      rvec                        * gmx_restrict          xx,
 731                      rvec                        * gmx_restrict          ff,
 732                      t_forcerec                  * gmx_restrict          fr,
 733                      t_mdatoms                   * gmx_restrict     mdatoms,
 734                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 735                      t_nrnb                      * gmx_restrict        nrnb)
 736 {
 737     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 738      * just 0 for non-waters.
 739      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 740      * jnr indices corresponding to data put in the four positions in the SIMD register.
 741      */
 742     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 743     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 744     int              jnrA,jnrB;
 745     int              j_coord_offsetA,j_coord_offsetB;
 746     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 747     real             rcutoff_scalar;
 748     real             *shiftvec,*fshift,*x,*f;
 749     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 750     int              vdwioffset0;
 751     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 752     int              vdwioffset1;
 753     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 754     int              vdwioffset2;
 755     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 756     int              vdwioffset3;
 757     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 758     int              vdwjidx0A,vdwjidx0B;
 759     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 760     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 761     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 762     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 763     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 764     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 765     real             *charge;
 766     int              nvdwtype;
 767     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 768     int              *vdwtype;
 769     real             *vdwparam;
 770     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 771     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 772     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 773     real             *vftab;
 774     _fjsp_v2r8       itab_tmp;
 775     _fjsp_v2r8       dummy_mask,cutoff_mask;
 776     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 777     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 778     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 779
 780     x                = xx[0];
 781     f                = ff[0];
 782
 783     nri              = nlist->nri;
 784     iinr             = nlist->iinr;
 785     jindex           = nlist->jindex;
 786     jjnr             = nlist->jjnr;
 787     shiftidx         = nlist->shift;
 788     gid              = nlist->gid;
 789     shiftvec         = fr->shift_vec[0];
 790     fshift           = fr->fshift[0];
 791     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 792     charge           = mdatoms->chargeA;
 793     nvdwtype         = fr->ntype;
 794     vdwparam         = fr->nbfp;
 795     vdwtype          = mdatoms->typeA;
 796
 797     vftab            = kernel_data->table_elec_vdw->data;
 798     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 799
 800     /* Setup water-specific parameters */
 801     inr              = nlist->iinr[0];
 802     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 803     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 804     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 805     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 806
 807     /* Avoid stupid compiler warnings */
 808     jnrA = jnrB = 0;
 809     j_coord_offsetA = 0;
 810     j_coord_offsetB = 0;
 811
 812     outeriter        = 0;
 813     inneriter        = 0;
 814
 815     /* Start outer loop over neighborlists */
 816     for(iidx=0; iidx<nri; iidx++)
 817     {
 818         /* Load shift vector for this list */
 819         i_shift_offset   = DIM*shiftidx[iidx];
 820
 821         /* Load limits for loop over neighbors */
 822         j_index_start    = jindex[iidx];
 823         j_index_end      = jindex[iidx+1];
 824
 825         /* Get outer coordinate index */
 826         inr              = iinr[iidx];
 827         i_coord_offset   = DIM*inr;
 828
 829         /* Load i particle coords and add shift vector */
 830         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 831                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 832
 833         fix0             = _fjsp_setzero_v2r8();
 834         fiy0             = _fjsp_setzero_v2r8();
 835         fiz0             = _fjsp_setzero_v2r8();
 836         fix1             = _fjsp_setzero_v2r8();
 837         fiy1             = _fjsp_setzero_v2r8();
 838         fiz1             = _fjsp_setzero_v2r8();
 839         fix2             = _fjsp_setzero_v2r8();
 840         fiy2             = _fjsp_setzero_v2r8();
 841         fiz2             = _fjsp_setzero_v2r8();
 842         fix3             = _fjsp_setzero_v2r8();
 843         fiy3             = _fjsp_setzero_v2r8();
 844         fiz3             = _fjsp_setzero_v2r8();
 845
 846         /* Start inner kernel loop */
 847         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 848         {
 849
 850             /* Get j neighbor index, and coordinate index */
 851             jnrA             = jjnr[jidx];
 852             jnrB             = jjnr[jidx+1];
 853             j_coord_offsetA  = DIM*jnrA;
 854             j_coord_offsetB  = DIM*jnrB;
 855
 856             /* load j atom coordinates */
 857             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 858                                               &jx0,&jy0,&jz0);
 859
 860             /* Calculate displacement vector */
 861             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 862             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 863             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 864             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 865             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 866             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 867             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 868             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 869             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 870             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 871             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 872             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 873
 874             /* Calculate squared distance and things based on it */
 875             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 876             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 877             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 878             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 879
 880             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 881             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 882             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 883             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 884
 885             /* Load parameters for j particles */
 886             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 887             vdwjidx0A        = 2*vdwtype[jnrA+0];
 888             vdwjidx0B        = 2*vdwtype[jnrB+0];
 889
 890             fjx0             = _fjsp_setzero_v2r8();
 891             fjy0             = _fjsp_setzero_v2r8();
 892             fjz0             = _fjsp_setzero_v2r8();
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 899
 900             /* Compute parameters for interactions between i and j atoms */
 901             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 902                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 903
 904             /* Calculate table index by multiplying r with table scale and truncate to integer */
 905             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 906             itab_tmp         = _fjsp_dtox_v2r8(rt);
 907             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 908             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 909             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 910
 911             vfconv.i[0]     *= 12;
 912             vfconv.i[1]     *= 12;
 913
 914             /* CUBIC SPLINE TABLE DISPERSION */
 915             vfconv.i[0]       += 4;
 916             vfconv.i[1]       += 4;
 917             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 918             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 919             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 920             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 921             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 922             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 923             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 924             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 925             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 926
 927             /* CUBIC SPLINE TABLE REPULSION */
 928             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 929             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 930             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 931             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 932             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 933             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 934             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 935             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 936             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 937             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 938
 939             fscal            = fvdw;
 940
 941             /* Update vectorial force */
 942             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 943             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 944             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 945
 946             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 947             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 948             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 949
 950             /**************************
 951              * CALCULATE INTERACTIONS *
 952              **************************/
 953
 954             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 955
 956             /* Compute parameters for interactions between i and j atoms */
 957             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 958
 959             /* Calculate table index by multiplying r with table scale and truncate to integer */
 960             rt               = _fjsp_mul_v2r8(r10,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]     *= 12;
 967             vfconv.i[1]     *= 12;
 968
 969             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 970             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 971             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 972             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 973             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 974             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 975             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 976             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 977             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 978             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 979
 980             fscal            = felec;
 981
 982             /* Update vectorial force */
 983             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 984             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 985             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 986
 987             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 988             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 989             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 996
 997             /* Compute parameters for interactions between i and j atoms */
 998             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 999
1000             /* Calculate table index by multiplying r with table scale and truncate to integer */
1001             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1002             itab_tmp         = _fjsp_dtox_v2r8(rt);
1003             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1004             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1005             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1006
1007             vfconv.i[0]     *= 12;
1008             vfconv.i[1]     *= 12;
1009
1010             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1011             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1012             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
1013             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1014             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1015             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
1016             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1017             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1018             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1019             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1020
1021             fscal            = felec;
1022
1023             /* Update vectorial force */
1024             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1025             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1026             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1027
1028             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1029             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1030             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1031
1032             /**************************
1033              * CALCULATE INTERACTIONS *
1034              **************************/
1035
1036             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1037
1038             /* Compute parameters for interactions between i and j atoms */
1039             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1040
1041             /* Calculate table index by multiplying r with table scale and truncate to integer */
1042             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1043             itab_tmp         = _fjsp_dtox_v2r8(rt);
1044             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1045             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1046             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1047
1048             vfconv.i[0]     *= 12;
1049             vfconv.i[1]     *= 12;
1050
1051             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1052             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1053             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
1054             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1055             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1056             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
1057             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1058             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1059             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1060             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1061
1062             fscal            = felec;
1063
1064             /* Update vectorial force */
1065             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1066             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1067             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1068
1069             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1070             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1071             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1072
1073             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1074
1075             /* Inner loop uses 180 flops */
1076         }
1077
1078         if(jidx<j_index_end)
1079         {
1080
1081             jnrA             = jjnr[jidx];
1082             j_coord_offsetA  = DIM*jnrA;
1083
1084             /* load j atom coordinates */
1085             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1086                                               &jx0,&jy0,&jz0);
1087
1088             /* Calculate displacement vector */
1089             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1090             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1091             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1092             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1093             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1094             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1095             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1096             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1097             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1098             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1099             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1100             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1101
1102             /* Calculate squared distance and things based on it */
1103             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1104             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1105             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1106             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1107
1108             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1109             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1110             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1111             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1112
1113             /* Load parameters for j particles */
1114             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1115             vdwjidx0A        = 2*vdwtype[jnrA+0];
1116
1117             fjx0             = _fjsp_setzero_v2r8();
1118             fjy0             = _fjsp_setzero_v2r8();
1119             fjz0             = _fjsp_setzero_v2r8();
1120
1121             /**************************
1122              * CALCULATE INTERACTIONS *
1123              **************************/
1124
1125             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1126
1127             /* Compute parameters for interactions between i and j atoms */
1128             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1129
1130             /* Calculate table index by multiplying r with table scale and truncate to integer */
1131             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1132             itab_tmp         = _fjsp_dtox_v2r8(rt);
1133             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1134             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1135             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1136
1137             vfconv.i[0]     *= 12;
1138             vfconv.i[1]     *= 12;
1139
1140             /* CUBIC SPLINE TABLE DISPERSION */
1141             vfconv.i[0]       += 4;
1142             vfconv.i[1]       += 4;
1143             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1144             F                = _fjsp_setzero_v2r8();
1145             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1146             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1147             H                = _fjsp_setzero_v2r8();
1148             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1149             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1150             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1151             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1152
1153             /* CUBIC SPLINE TABLE REPULSION */
1154             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1155             F                = _fjsp_setzero_v2r8();
1156             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1157             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1158             H                = _fjsp_setzero_v2r8();
1159             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1160             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1161             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1162             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1163             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1164
1165             fscal            = fvdw;
1166
1167             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1168
1169             /* Update vectorial force */
1170             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1171             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1172             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1173
1174             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1175             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1176             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1177
1178             /**************************
1179              * CALCULATE INTERACTIONS *
1180              **************************/
1181
1182             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1183
1184             /* Compute parameters for interactions between i and j atoms */
1185             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1186
1187             /* Calculate table index by multiplying r with table scale and truncate to integer */
1188             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1189             itab_tmp         = _fjsp_dtox_v2r8(rt);
1190             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1191             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1192             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1193
1194             vfconv.i[0]     *= 12;
1195             vfconv.i[1]     *= 12;
1196
1197             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1198             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1199             F                = _fjsp_setzero_v2r8();
1200             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1201             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1202             H                = _fjsp_setzero_v2r8();
1203             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1204             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1205             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1206             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1207
1208             fscal            = felec;
1209
1210             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1211
1212             /* Update vectorial force */
1213             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1214             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1215             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1216
1217             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1218             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1219             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1220
1221             /**************************
1222              * CALCULATE INTERACTIONS *
1223              **************************/
1224
1225             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1226
1227             /* Compute parameters for interactions between i and j atoms */
1228             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1229
1230             /* Calculate table index by multiplying r with table scale and truncate to integer */
1231             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1232             itab_tmp         = _fjsp_dtox_v2r8(rt);
1233             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1234             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1235             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1236
1237             vfconv.i[0]     *= 12;
1238             vfconv.i[1]     *= 12;
1239
1240             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1241             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1242             F                = _fjsp_setzero_v2r8();
1243             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1244             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1245             H                = _fjsp_setzero_v2r8();
1246             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1247             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1248             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1249             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1250
1251             fscal            = felec;
1252
1253             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1254
1255             /* Update vectorial force */
1256             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1257             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1258             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1259
1260             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1261             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1262             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1263
1264             /**************************
1265              * CALCULATE INTERACTIONS *
1266              **************************/
1267
1268             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1269
1270             /* Compute parameters for interactions between i and j atoms */
1271             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1272
1273             /* Calculate table index by multiplying r with table scale and truncate to integer */
1274             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1275             itab_tmp         = _fjsp_dtox_v2r8(rt);
1276             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1277             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1278             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1279
1280             vfconv.i[0]     *= 12;
1281             vfconv.i[1]     *= 12;
1282
1283             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1284             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1285             F                = _fjsp_setzero_v2r8();
1286             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1287             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1288             H                = _fjsp_setzero_v2r8();
1289             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1290             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1291             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1292             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1293
1294             fscal            = felec;
1295
1296             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1297
1298             /* Update vectorial force */
1299             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1300             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1301             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1302
1303             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1304             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1305             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1306
1307             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1308
1309             /* Inner loop uses 180 flops */
1310         }
1311
1312         /* End of innermost loop */
1313
1314         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1315                                               f+i_coord_offset,fshift+i_shift_offset);
1316
1317         /* Increment number of inner iterations */
1318         inneriter                  += j_index_end - j_index_start;
1319
1320         /* Outer loop uses 24 flops */
1321     }
1322
1323     /* Increment number of outer iterations */
1324     outeriter        += nri;
1325
1326     /* Update outer/inner flops */
1327
1328     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*180);
1329 }