src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_elec->data;
 124     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     /* Start outer loop over neighborlists */
 142     for(iidx=0; iidx<nri; iidx++)
 143     {
 144         /* Load shift vector for this list */
 145         i_shift_offset   = DIM*shiftidx[iidx];
 146
 147         /* Load limits for loop over neighbors */
 148         j_index_start    = jindex[iidx];
 149         j_index_end      = jindex[iidx+1];
 150
 151         /* Get outer coordinate index */
 152         inr              = iinr[iidx];
 153         i_coord_offset   = DIM*inr;
 154
 155         /* Load i particle coords and add shift vector */
 156         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 157                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 158
 159         fix0             = _fjsp_setzero_v2r8();
 160         fiy0             = _fjsp_setzero_v2r8();
 161         fiz0             = _fjsp_setzero_v2r8();
 162         fix1             = _fjsp_setzero_v2r8();
 163         fiy1             = _fjsp_setzero_v2r8();
 164         fiz1             = _fjsp_setzero_v2r8();
 165         fix2             = _fjsp_setzero_v2r8();
 166         fiy2             = _fjsp_setzero_v2r8();
 167         fiz2             = _fjsp_setzero_v2r8();
 168
 169         /* Reset potential sums */
 170         velecsum         = _fjsp_setzero_v2r8();
 171         vvdwsum          = _fjsp_setzero_v2r8();
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 175         {
 176
 177             /* Get j neighbor index, and coordinate index */
 178             jnrA             = jjnr[jidx];
 179             jnrB             = jjnr[jidx+1];
 180             j_coord_offsetA  = DIM*jnrA;
 181             j_coord_offsetB  = DIM*jnrB;
 182
 183             /* load j atom coordinates */
 184             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                               &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 189             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 190             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 191             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 192             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 193             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 194             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 195             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 196             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 197
 198             /* Calculate squared distance and things based on it */
 199             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 200             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 201             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 202
 203             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 204             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 205             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 206
 207             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 208
 209             /* Load parameters for j particles */
 210             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 211             vdwjidx0A        = 2*vdwtype[jnrA+0];
 212             vdwjidx0B        = 2*vdwtype[jnrB+0];
 213
 214             fjx0             = _fjsp_setzero_v2r8();
 215             fjy0             = _fjsp_setzero_v2r8();
 216             fjz0             = _fjsp_setzero_v2r8();
 217
 218             /**************************
 219              * CALCULATE INTERACTIONS *
 220              **************************/
 221
 222             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 223
 224             /* Compute parameters for interactions between i and j atoms */
 225             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 226             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 227                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 228
 229             /* Calculate table index by multiplying r with table scale and truncate to integer */
 230             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 231             itab_tmp         = _fjsp_dtox_v2r8(rt);
 232             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 233             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 234             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 235
 236             vfconv.i[0]     *= 4;
 237             vfconv.i[1]     *= 4;
 238
 239             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 240             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 241             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 242             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 243             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 244             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 245             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 246             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 247             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 248             velec            = _fjsp_mul_v2r8(qq00,VV);
 249             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 250             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 251
 252             /* LENNARD-JONES DISPERSION/REPULSION */
 253
 254             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 255             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 256             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 257             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 258             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 259
 260             /* Update potential sum for this i atom from the interaction with this j atom. */
 261             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 262             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 263
 264             fscal            = _fjsp_add_v2r8(felec,fvdw);
 265
 266             /* Update vectorial force */
 267             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 268             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 269             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 270
 271             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 272             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 273             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 274
 275             /**************************
 276              * CALCULATE INTERACTIONS *
 277              **************************/
 278
 279             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 280
 281             /* Compute parameters for interactions between i and j atoms */
 282             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 283
 284             /* Calculate table index by multiplying r with table scale and truncate to integer */
 285             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 286             itab_tmp         = _fjsp_dtox_v2r8(rt);
 287             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 288             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 289             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 290
 291             vfconv.i[0]     *= 4;
 292             vfconv.i[1]     *= 4;
 293
 294             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 295             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 296             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 297             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 298             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 299             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 300             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 301             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 302             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 303             velec            = _fjsp_mul_v2r8(qq10,VV);
 304             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 305             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 306
 307             /* Update potential sum for this i atom from the interaction with this j atom. */
 308             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 309
 310             fscal            = felec;
 311
 312             /* Update vectorial force */
 313             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 314             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 315             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 316
 317             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 318             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 319             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 320
 321             /**************************
 322              * CALCULATE INTERACTIONS *
 323              **************************/
 324
 325             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 329
 330             /* Calculate table index by multiplying r with table scale and truncate to integer */
 331             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 332             itab_tmp         = _fjsp_dtox_v2r8(rt);
 333             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 334             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 335             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 336
 337             vfconv.i[0]     *= 4;
 338             vfconv.i[1]     *= 4;
 339
 340             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 341             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 342             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 343             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 344             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 345             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 346             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 347             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 348             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 349             velec            = _fjsp_mul_v2r8(qq20,VV);
 350             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 351             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 352
 353             /* Update potential sum for this i atom from the interaction with this j atom. */
 354             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 355
 356             fscal            = felec;
 357
 358             /* Update vectorial force */
 359             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 360             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 361             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 362
 363             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 364             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 365             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 366
 367             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 368
 369             /* Inner loop uses 154 flops */
 370         }
 371
 372         if(jidx<j_index_end)
 373         {
 374
 375             jnrA             = jjnr[jidx];
 376             j_coord_offsetA  = DIM*jnrA;
 377
 378             /* load j atom coordinates */
 379             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 380                                               &jx0,&jy0,&jz0);
 381
 382             /* Calculate displacement vector */
 383             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 384             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 385             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 386             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 387             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 388             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 389             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 390             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 391             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 392
 393             /* Calculate squared distance and things based on it */
 394             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 395             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 396             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 397
 398             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 399             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 400             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 401
 402             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 403
 404             /* Load parameters for j particles */
 405             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 406             vdwjidx0A        = 2*vdwtype[jnrA+0];
 407
 408             fjx0             = _fjsp_setzero_v2r8();
 409             fjy0             = _fjsp_setzero_v2r8();
 410             fjz0             = _fjsp_setzero_v2r8();
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 417
 418             /* Compute parameters for interactions between i and j atoms */
 419             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 420             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 421                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 422
 423             /* Calculate table index by multiplying r with table scale and truncate to integer */
 424             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 425             itab_tmp         = _fjsp_dtox_v2r8(rt);
 426             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 427             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 428             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 429
 430             vfconv.i[0]     *= 4;
 431             vfconv.i[1]     *= 4;
 432
 433             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 434             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 435             F                = _fjsp_setzero_v2r8();
 436             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 437             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 438             H                = _fjsp_setzero_v2r8();
 439             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 440             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 441             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 442             velec            = _fjsp_mul_v2r8(qq00,VV);
 443             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 444             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 445
 446             /* LENNARD-JONES DISPERSION/REPULSION */
 447
 448             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 449             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 450             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 451             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 452             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 453
 454             /* Update potential sum for this i atom from the interaction with this j atom. */
 455             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 456             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 457             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 458             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 459
 460             fscal            = _fjsp_add_v2r8(felec,fvdw);
 461
 462             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 463
 464             /* Update vectorial force */
 465             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 466             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 467             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 468
 469             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 470             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 471             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 472
 473             /**************************
 474              * CALCULATE INTERACTIONS *
 475              **************************/
 476
 477             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 478
 479             /* Compute parameters for interactions between i and j atoms */
 480             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 481
 482             /* Calculate table index by multiplying r with table scale and truncate to integer */
 483             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 484             itab_tmp         = _fjsp_dtox_v2r8(rt);
 485             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 486             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 487             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 488
 489             vfconv.i[0]     *= 4;
 490             vfconv.i[1]     *= 4;
 491
 492             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 493             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 494             F                = _fjsp_setzero_v2r8();
 495             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 496             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 497             H                = _fjsp_setzero_v2r8();
 498             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 499             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 500             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 501             velec            = _fjsp_mul_v2r8(qq10,VV);
 502             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 503             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 504
 505             /* Update potential sum for this i atom from the interaction with this j atom. */
 506             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 507             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 508
 509             fscal            = felec;
 510
 511             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 512
 513             /* Update vectorial force */
 514             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 515             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 516             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 517
 518             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 519             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 520             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 521
 522             /**************************
 523              * CALCULATE INTERACTIONS *
 524              **************************/
 525
 526             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 527
 528             /* Compute parameters for interactions between i and j atoms */
 529             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 530
 531             /* Calculate table index by multiplying r with table scale and truncate to integer */
 532             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 533             itab_tmp         = _fjsp_dtox_v2r8(rt);
 534             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 535             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 536             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 537
 538             vfconv.i[0]     *= 4;
 539             vfconv.i[1]     *= 4;
 540
 541             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 542             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 543             F                = _fjsp_setzero_v2r8();
 544             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 545             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 546             H                = _fjsp_setzero_v2r8();
 547             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 548             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 549             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 550             velec            = _fjsp_mul_v2r8(qq20,VV);
 551             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 552             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 556             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 561
 562             /* Update vectorial force */
 563             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 564             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 565             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 566
 567             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 568             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 569             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 570
 571             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 572
 573             /* Inner loop uses 154 flops */
 574         }
 575
 576         /* End of innermost loop */
 577
 578         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 579                                               f+i_coord_offset,fshift+i_shift_offset);
 580
 581         ggid                        = gid[iidx];
 582         /* Update potential energies */
 583         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 584         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 585
 586         /* Increment number of inner iterations */
 587         inneriter                  += j_index_end - j_index_start;
 588
 589         /* Outer loop uses 20 flops */
 590     }
 591
 592     /* Increment number of outer iterations */
 593     outeriter        += nri;
 594
 595     /* Update outer/inner flops */
 596
 597     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 598 }
 599 /*
 600  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 601  * Electrostatics interaction: CubicSplineTable
 602  * VdW interaction:            LennardJones
 603  * Geometry:                   Water3-Particle
 604  * Calculate force/pot:        Force
 605  */
 606 void
 607 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 608                     (t_nblist                    * gmx_restrict       nlist,
 609                      rvec                        * gmx_restrict          xx,
 610                      rvec                        * gmx_restrict          ff,
 611                      struct t_forcerec           * gmx_restrict          fr,
 612                      t_mdatoms                   * gmx_restrict     mdatoms,
 613                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 614                      t_nrnb                      * gmx_restrict        nrnb)
 615 {
 616     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 617      * just 0 for non-waters.
 618      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 619      * jnr indices corresponding to data put in the four positions in the SIMD register.
 620      */
 621     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 622     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 623     int              jnrA,jnrB;
 624     int              j_coord_offsetA,j_coord_offsetB;
 625     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 626     real             rcutoff_scalar;
 627     real             *shiftvec,*fshift,*x,*f;
 628     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 629     int              vdwioffset0;
 630     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 631     int              vdwioffset1;
 632     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 633     int              vdwioffset2;
 634     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 635     int              vdwjidx0A,vdwjidx0B;
 636     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 637     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 638     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 639     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 640     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 641     real             *charge;
 642     int              nvdwtype;
 643     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 644     int              *vdwtype;
 645     real             *vdwparam;
 646     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 647     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 648     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 649     real             *vftab;
 650     _fjsp_v2r8       itab_tmp;
 651     _fjsp_v2r8       dummy_mask,cutoff_mask;
 652     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 653     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 654     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 655
 656     x                = xx[0];
 657     f                = ff[0];
 658
 659     nri              = nlist->nri;
 660     iinr             = nlist->iinr;
 661     jindex           = nlist->jindex;
 662     jjnr             = nlist->jjnr;
 663     shiftidx         = nlist->shift;
 664     gid              = nlist->gid;
 665     shiftvec         = fr->shift_vec[0];
 666     fshift           = fr->fshift[0];
 667     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 668     charge           = mdatoms->chargeA;
 669     nvdwtype         = fr->ntype;
 670     vdwparam         = fr->nbfp;
 671     vdwtype          = mdatoms->typeA;
 672
 673     vftab            = kernel_data->table_elec->data;
 674     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 675
 676     /* Setup water-specific parameters */
 677     inr              = nlist->iinr[0];
 678     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 679     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 680     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 681     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 682
 683     /* Avoid stupid compiler warnings */
 684     jnrA = jnrB = 0;
 685     j_coord_offsetA = 0;
 686     j_coord_offsetB = 0;
 687
 688     outeriter        = 0;
 689     inneriter        = 0;
 690
 691     /* Start outer loop over neighborlists */
 692     for(iidx=0; iidx<nri; iidx++)
 693     {
 694         /* Load shift vector for this list */
 695         i_shift_offset   = DIM*shiftidx[iidx];
 696
 697         /* Load limits for loop over neighbors */
 698         j_index_start    = jindex[iidx];
 699         j_index_end      = jindex[iidx+1];
 700
 701         /* Get outer coordinate index */
 702         inr              = iinr[iidx];
 703         i_coord_offset   = DIM*inr;
 704
 705         /* Load i particle coords and add shift vector */
 706         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 707                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 708
 709         fix0             = _fjsp_setzero_v2r8();
 710         fiy0             = _fjsp_setzero_v2r8();
 711         fiz0             = _fjsp_setzero_v2r8();
 712         fix1             = _fjsp_setzero_v2r8();
 713         fiy1             = _fjsp_setzero_v2r8();
 714         fiz1             = _fjsp_setzero_v2r8();
 715         fix2             = _fjsp_setzero_v2r8();
 716         fiy2             = _fjsp_setzero_v2r8();
 717         fiz2             = _fjsp_setzero_v2r8();
 718
 719         /* Start inner kernel loop */
 720         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 721         {
 722
 723             /* Get j neighbor index, and coordinate index */
 724             jnrA             = jjnr[jidx];
 725             jnrB             = jjnr[jidx+1];
 726             j_coord_offsetA  = DIM*jnrA;
 727             j_coord_offsetB  = DIM*jnrB;
 728
 729             /* load j atom coordinates */
 730             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 731                                               &jx0,&jy0,&jz0);
 732
 733             /* Calculate displacement vector */
 734             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 735             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 736             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 737             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 738             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 739             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 740             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 741             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 742             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 743
 744             /* Calculate squared distance and things based on it */
 745             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 746             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 747             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 748
 749             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 750             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 751             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 752
 753             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 754
 755             /* Load parameters for j particles */
 756             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 757             vdwjidx0A        = 2*vdwtype[jnrA+0];
 758             vdwjidx0B        = 2*vdwtype[jnrB+0];
 759
 760             fjx0             = _fjsp_setzero_v2r8();
 761             fjy0             = _fjsp_setzero_v2r8();
 762             fjz0             = _fjsp_setzero_v2r8();
 763
 764             /**************************
 765              * CALCULATE INTERACTIONS *
 766              **************************/
 767
 768             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 769
 770             /* Compute parameters for interactions between i and j atoms */
 771             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 772             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 773                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 774
 775             /* Calculate table index by multiplying r with table scale and truncate to integer */
 776             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 777             itab_tmp         = _fjsp_dtox_v2r8(rt);
 778             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 779             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 780             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 781
 782             vfconv.i[0]     *= 4;
 783             vfconv.i[1]     *= 4;
 784
 785             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 786             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 787             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 788             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 789             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 790             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 791             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 792             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 793             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 794             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 795
 796             /* LENNARD-JONES DISPERSION/REPULSION */
 797
 798             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 799             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 800
 801             fscal            = _fjsp_add_v2r8(felec,fvdw);
 802
 803             /* Update vectorial force */
 804             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 805             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 806             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 807
 808             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 809             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 810             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 811
 812             /**************************
 813              * CALCULATE INTERACTIONS *
 814              **************************/
 815
 816             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 817
 818             /* Compute parameters for interactions between i and j atoms */
 819             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 820
 821             /* Calculate table index by multiplying r with table scale and truncate to integer */
 822             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 823             itab_tmp         = _fjsp_dtox_v2r8(rt);
 824             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 825             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 826             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 827
 828             vfconv.i[0]     *= 4;
 829             vfconv.i[1]     *= 4;
 830
 831             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 832             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 833             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 834             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 835             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 836             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 837             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 838             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 839             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 840             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 841
 842             fscal            = felec;
 843
 844             /* Update vectorial force */
 845             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 846             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 847             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 848
 849             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 850             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 851             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 858
 859             /* Compute parameters for interactions between i and j atoms */
 860             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 861
 862             /* Calculate table index by multiplying r with table scale and truncate to integer */
 863             rt               = _fjsp_mul_v2r8(r20,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(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 882
 883             fscal            = felec;
 884
 885             /* Update vectorial force */
 886             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 887             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 888             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 889
 890             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 891             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 892             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 893
 894             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 895
 896             /* Inner loop uses 137 flops */
 897         }
 898
 899         if(jidx<j_index_end)
 900         {
 901
 902             jnrA             = jjnr[jidx];
 903             j_coord_offsetA  = DIM*jnrA;
 904
 905             /* load j atom coordinates */
 906             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 907                                               &jx0,&jy0,&jz0);
 908
 909             /* Calculate displacement vector */
 910             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 911             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 912             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 913             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 914             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 915             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 916             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 917             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 918             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 919
 920             /* Calculate squared distance and things based on it */
 921             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 922             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 923             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 924
 925             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 926             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 927             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 928
 929             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 930
 931             /* Load parameters for j particles */
 932             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 933             vdwjidx0A        = 2*vdwtype[jnrA+0];
 934
 935             fjx0             = _fjsp_setzero_v2r8();
 936             fjy0             = _fjsp_setzero_v2r8();
 937             fjz0             = _fjsp_setzero_v2r8();
 938
 939             /**************************
 940              * CALCULATE INTERACTIONS *
 941              **************************/
 942
 943             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 944
 945             /* Compute parameters for interactions between i and j atoms */
 946             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 947             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 948                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 949
 950             /* Calculate table index by multiplying r with table scale and truncate to integer */
 951             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 952             itab_tmp         = _fjsp_dtox_v2r8(rt);
 953             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 954             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 955             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 956
 957             vfconv.i[0]     *= 4;
 958             vfconv.i[1]     *= 4;
 959
 960             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 961             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 962             F                = _fjsp_setzero_v2r8();
 963             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 964             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 965             H                = _fjsp_setzero_v2r8();
 966             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 967             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 968             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 969             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 970
 971             /* LENNARD-JONES DISPERSION/REPULSION */
 972
 973             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 974             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 975
 976             fscal            = _fjsp_add_v2r8(felec,fvdw);
 977
 978             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 979
 980             /* Update vectorial force */
 981             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 982             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 983             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 984
 985             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 986             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 987             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 988
 989             /**************************
 990              * CALCULATE INTERACTIONS *
 991              **************************/
 992
 993             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 994
 995             /* Compute parameters for interactions between i and j atoms */
 996             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 997
 998             /* Calculate table index by multiplying r with table scale and truncate to integer */
 999             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1000             itab_tmp         = _fjsp_dtox_v2r8(rt);
1001             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1002             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1003             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1004
1005             vfconv.i[0]     *= 4;
1006             vfconv.i[1]     *= 4;
1007
1008             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1009             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1010             F                = _fjsp_setzero_v2r8();
1011             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1012             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1013             H                = _fjsp_setzero_v2r8();
1014             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1015             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1016             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1017             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1018
1019             fscal            = felec;
1020
1021             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1022
1023             /* Update vectorial force */
1024             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1025             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1026             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1027
1028             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1029             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1030             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1031
1032             /**************************
1033              * CALCULATE INTERACTIONS *
1034              **************************/
1035
1036             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1037
1038             /* Compute parameters for interactions between i and j atoms */
1039             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1040
1041             /* Calculate table index by multiplying r with table scale and truncate to integer */
1042             rt               = _fjsp_mul_v2r8(r20,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]     *= 4;
1049             vfconv.i[1]     *= 4;
1050
1051             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1052             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1053             F                = _fjsp_setzero_v2r8();
1054             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1055             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1056             H                = _fjsp_setzero_v2r8();
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(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1061
1062             fscal            = felec;
1063
1064             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1065
1066             /* Update vectorial force */
1067             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1068             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1069             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1070
1071             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1072             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1073             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1074
1075             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1076
1077             /* Inner loop uses 137 flops */
1078         }
1079
1080         /* End of innermost loop */
1081
1082         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1083                                               f+i_coord_offset,fshift+i_shift_offset);
1084
1085         /* Increment number of inner iterations */
1086         inneriter                  += j_index_end - j_index_start;
1087
1088         /* Outer loop uses 18 flops */
1089     }
1090
1091     /* Increment number of outer iterations */
1092     outeriter        += nri;
1093
1094     /* Update outer/inner flops */
1095
1096     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*137);
1097 }