src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRF_VdwCSTab_GeomW3P1_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_ElecRF_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_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              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 122     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 123     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_vdw->data;
 129     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 134     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 135     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 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_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 162                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 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
 174         /* Reset potential sums */
 175         velecsum         = _fjsp_setzero_v2r8();
 176         vvdwsum          = _fjsp_setzero_v2r8();
 177
 178         /* Start inner kernel loop */
 179         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 180         {
 181
 182             /* Get j neighbor index, and coordinate index */
 183             jnrA             = jjnr[jidx];
 184             jnrB             = jjnr[jidx+1];
 185             j_coord_offsetA  = DIM*jnrA;
 186             j_coord_offsetB  = DIM*jnrB;
 187
 188             /* load j atom coordinates */
 189             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 190                                               &jx0,&jy0,&jz0);
 191
 192             /* Calculate displacement vector */
 193             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 194             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 195             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 196             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 197             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 198             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 199             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 200             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 201             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 202
 203             /* Calculate squared distance and things based on it */
 204             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 205             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 206             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 207
 208             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 209             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 210             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 211
 212             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 213             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 214             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 215
 216             /* Load parameters for j particles */
 217             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 218             vdwjidx0A        = 2*vdwtype[jnrA+0];
 219             vdwjidx0B        = 2*vdwtype[jnrB+0];
 220
 221             fjx0             = _fjsp_setzero_v2r8();
 222             fjy0             = _fjsp_setzero_v2r8();
 223             fjz0             = _fjsp_setzero_v2r8();
 224
 225             /**************************
 226              * CALCULATE INTERACTIONS *
 227              **************************/
 228
 229             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 230
 231             /* Compute parameters for interactions between i and j atoms */
 232             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 233             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 234                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 235
 236             /* Calculate table index by multiplying r with table scale and truncate to integer */
 237             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 238             itab_tmp         = _fjsp_dtox_v2r8(rt);
 239             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 240             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 241             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 242
 243             vfconv.i[0]     *= 8;
 244             vfconv.i[1]     *= 8;
 245
 246             /* REACTION-FIELD ELECTROSTATICS */
 247             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 248             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 249
 250             /* CUBIC SPLINE TABLE DISPERSION */
 251             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 252             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 253             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 254             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 255             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 256             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 257             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 258             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 259             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 260             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 261             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 262
 263             /* CUBIC SPLINE TABLE REPULSION */
 264             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 265             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 266             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 267             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 268             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 269             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 270             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 271             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 272             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 273             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 274             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 275             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 276             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 277
 278             /* Update potential sum for this i atom from the interaction with this j atom. */
 279             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = _fjsp_add_v2r8(felec,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             /* Compute parameters for interactions between i and j atoms */
 298             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 299
 300             /* REACTION-FIELD ELECTROSTATICS */
 301             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 302             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 303
 304             /* Update potential sum for this i atom from the interaction with this j atom. */
 305             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 306
 307             fscal            = felec;
 308
 309             /* Update vectorial force */
 310             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 311             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 312             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 313
 314             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 315             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 316             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             /* Compute parameters for interactions between i and j atoms */
 323             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 324
 325             /* REACTION-FIELD ELECTROSTATICS */
 326             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 327             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 336             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 337             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 342
 343             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 344
 345             /* Inner loop uses 143 flops */
 346         }
 347
 348         if(jidx<j_index_end)
 349         {
 350
 351             jnrA             = jjnr[jidx];
 352             j_coord_offsetA  = DIM*jnrA;
 353
 354             /* load j atom coordinates */
 355             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 356                                               &jx0,&jy0,&jz0);
 357
 358             /* Calculate displacement vector */
 359             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 360             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 361             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 362             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 363             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 364             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 365             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 366             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 367             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 368
 369             /* Calculate squared distance and things based on it */
 370             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 371             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 372             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 373
 374             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 375             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 376             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 377
 378             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 379             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 380             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 381
 382             /* Load parameters for j particles */
 383             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 384             vdwjidx0A        = 2*vdwtype[jnrA+0];
 385
 386             fjx0             = _fjsp_setzero_v2r8();
 387             fjy0             = _fjsp_setzero_v2r8();
 388             fjz0             = _fjsp_setzero_v2r8();
 389
 390             /**************************
 391              * CALCULATE INTERACTIONS *
 392              **************************/
 393
 394             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 395
 396             /* Compute parameters for interactions between i and j atoms */
 397             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 398             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 399
 400             /* Calculate table index by multiplying r with table scale and truncate to integer */
 401             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 402             itab_tmp         = _fjsp_dtox_v2r8(rt);
 403             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 404             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 405             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 406
 407             vfconv.i[0]     *= 8;
 408             vfconv.i[1]     *= 8;
 409
 410             /* REACTION-FIELD ELECTROSTATICS */
 411             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 412             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 413
 414             /* CUBIC SPLINE TABLE DISPERSION */
 415             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 416             F                = _fjsp_setzero_v2r8();
 417             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 418             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 419             H                = _fjsp_setzero_v2r8();
 420             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 421             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 422             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 423             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 424             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 425             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 426
 427             /* CUBIC SPLINE TABLE REPULSION */
 428             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 429             F                = _fjsp_setzero_v2r8();
 430             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 431             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 432             H                = _fjsp_setzero_v2r8();
 433             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 434             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 435             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 436             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 437             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 438             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 439             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 440             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 441
 442             /* Update potential sum for this i atom from the interaction with this j atom. */
 443             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 444             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 445             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 446             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 447
 448             fscal            = _fjsp_add_v2r8(felec,fvdw);
 449
 450             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 451
 452             /* Update vectorial force */
 453             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 454             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 455             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 456
 457             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 458             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 459             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 460
 461             /**************************
 462              * CALCULATE INTERACTIONS *
 463              **************************/
 464
 465             /* Compute parameters for interactions between i and j atoms */
 466             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 467
 468             /* REACTION-FIELD ELECTROSTATICS */
 469             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 470             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 471
 472             /* Update potential sum for this i atom from the interaction with this j atom. */
 473             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 474             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 475
 476             fscal            = felec;
 477
 478             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 479
 480             /* Update vectorial force */
 481             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 482             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 483             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 484
 485             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 486             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 487             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 488
 489             /**************************
 490              * CALCULATE INTERACTIONS *
 491              **************************/
 492
 493             /* Compute parameters for interactions between i and j atoms */
 494             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 495
 496             /* REACTION-FIELD ELECTROSTATICS */
 497             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 498             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 499
 500             /* Update potential sum for this i atom from the interaction with this j atom. */
 501             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 502             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 503
 504             fscal            = felec;
 505
 506             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 507
 508             /* Update vectorial force */
 509             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 510             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 511             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 512
 513             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 514             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 515             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 516
 517             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 518
 519             /* Inner loop uses 143 flops */
 520         }
 521
 522         /* End of innermost loop */
 523
 524         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 525                                               f+i_coord_offset,fshift+i_shift_offset);
 526
 527         ggid                        = gid[iidx];
 528         /* Update potential energies */
 529         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 530         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 531
 532         /* Increment number of inner iterations */
 533         inneriter                  += j_index_end - j_index_start;
 534
 535         /* Outer loop uses 20 flops */
 536     }
 537
 538     /* Increment number of outer iterations */
 539     outeriter        += nri;
 540
 541     /* Update outer/inner flops */
 542
 543     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*143);
 544 }
 545 /*
 546  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 547  * Electrostatics interaction: ReactionField
 548  * VdW interaction:            CubicSplineTable
 549  * Geometry:                   Water3-Particle
 550  * Calculate force/pot:        Force
 551  */
 552 void
 553 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 554                     (t_nblist                    * gmx_restrict       nlist,
 555                      rvec                        * gmx_restrict          xx,
 556                      rvec                        * gmx_restrict          ff,
 557                      t_forcerec                  * gmx_restrict          fr,
 558                      t_mdatoms                   * gmx_restrict     mdatoms,
 559                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 560                      t_nrnb                      * gmx_restrict        nrnb)
 561 {
 562     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 563      * just 0 for non-waters.
 564      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 565      * jnr indices corresponding to data put in the four positions in the SIMD register.
 566      */
 567     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 568     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 569     int              jnrA,jnrB;
 570     int              j_coord_offsetA,j_coord_offsetB;
 571     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 572     real             rcutoff_scalar;
 573     real             *shiftvec,*fshift,*x,*f;
 574     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 575     int              vdwioffset0;
 576     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 577     int              vdwioffset1;
 578     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 579     int              vdwioffset2;
 580     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 581     int              vdwjidx0A,vdwjidx0B;
 582     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 583     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 584     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 585     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 586     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 587     real             *charge;
 588     int              nvdwtype;
 589     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 590     int              *vdwtype;
 591     real             *vdwparam;
 592     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 593     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 594     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 595     real             *vftab;
 596     _fjsp_v2r8       itab_tmp;
 597     _fjsp_v2r8       dummy_mask,cutoff_mask;
 598     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 599     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 600     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 601
 602     x                = xx[0];
 603     f                = ff[0];
 604
 605     nri              = nlist->nri;
 606     iinr             = nlist->iinr;
 607     jindex           = nlist->jindex;
 608     jjnr             = nlist->jjnr;
 609     shiftidx         = nlist->shift;
 610     gid              = nlist->gid;
 611     shiftvec         = fr->shift_vec[0];
 612     fshift           = fr->fshift[0];
 613     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 614     charge           = mdatoms->chargeA;
 615     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 616     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 617     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 618     nvdwtype         = fr->ntype;
 619     vdwparam         = fr->nbfp;
 620     vdwtype          = mdatoms->typeA;
 621
 622     vftab            = kernel_data->table_vdw->data;
 623     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 624
 625     /* Setup water-specific parameters */
 626     inr              = nlist->iinr[0];
 627     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 628     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 629     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 630     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 631
 632     /* Avoid stupid compiler warnings */
 633     jnrA = jnrB = 0;
 634     j_coord_offsetA = 0;
 635     j_coord_offsetB = 0;
 636
 637     outeriter        = 0;
 638     inneriter        = 0;
 639
 640     /* Start outer loop over neighborlists */
 641     for(iidx=0; iidx<nri; iidx++)
 642     {
 643         /* Load shift vector for this list */
 644         i_shift_offset   = DIM*shiftidx[iidx];
 645
 646         /* Load limits for loop over neighbors */
 647         j_index_start    = jindex[iidx];
 648         j_index_end      = jindex[iidx+1];
 649
 650         /* Get outer coordinate index */
 651         inr              = iinr[iidx];
 652         i_coord_offset   = DIM*inr;
 653
 654         /* Load i particle coords and add shift vector */
 655         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 656                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 657
 658         fix0             = _fjsp_setzero_v2r8();
 659         fiy0             = _fjsp_setzero_v2r8();
 660         fiz0             = _fjsp_setzero_v2r8();
 661         fix1             = _fjsp_setzero_v2r8();
 662         fiy1             = _fjsp_setzero_v2r8();
 663         fiz1             = _fjsp_setzero_v2r8();
 664         fix2             = _fjsp_setzero_v2r8();
 665         fiy2             = _fjsp_setzero_v2r8();
 666         fiz2             = _fjsp_setzero_v2r8();
 667
 668         /* Start inner kernel loop */
 669         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 670         {
 671
 672             /* Get j neighbor index, and coordinate index */
 673             jnrA             = jjnr[jidx];
 674             jnrB             = jjnr[jidx+1];
 675             j_coord_offsetA  = DIM*jnrA;
 676             j_coord_offsetB  = DIM*jnrB;
 677
 678             /* load j atom coordinates */
 679             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 680                                               &jx0,&jy0,&jz0);
 681
 682             /* Calculate displacement vector */
 683             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 684             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 685             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 686             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 687             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 688             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 689             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 690             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 691             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 692
 693             /* Calculate squared distance and things based on it */
 694             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 695             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 696             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 697
 698             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 699             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 700             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 701
 702             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 703             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 704             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 705
 706             /* Load parameters for j particles */
 707             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 708             vdwjidx0A        = 2*vdwtype[jnrA+0];
 709             vdwjidx0B        = 2*vdwtype[jnrB+0];
 710
 711             fjx0             = _fjsp_setzero_v2r8();
 712             fjy0             = _fjsp_setzero_v2r8();
 713             fjz0             = _fjsp_setzero_v2r8();
 714
 715             /**************************
 716              * CALCULATE INTERACTIONS *
 717              **************************/
 718
 719             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 720
 721             /* Compute parameters for interactions between i and j atoms */
 722             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 723             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 724                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 725
 726             /* Calculate table index by multiplying r with table scale and truncate to integer */
 727             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 728             itab_tmp         = _fjsp_dtox_v2r8(rt);
 729             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 730             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 731             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 732
 733             vfconv.i[0]     *= 8;
 734             vfconv.i[1]     *= 8;
 735
 736             /* REACTION-FIELD ELECTROSTATICS */
 737             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 738
 739             /* CUBIC SPLINE TABLE DISPERSION */
 740             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 741             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 742             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 743             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 744             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 745             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 746             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 747             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 748             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 749
 750             /* CUBIC SPLINE TABLE REPULSION */
 751             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 752             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 753             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 754             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 755             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 756             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 757             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 758             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 759             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 760             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 761
 762             fscal            = _fjsp_add_v2r8(felec,fvdw);
 763
 764             /* Update vectorial force */
 765             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 766             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 767             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 768
 769             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 770             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 771             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 772
 773             /**************************
 774              * CALCULATE INTERACTIONS *
 775              **************************/
 776
 777             /* Compute parameters for interactions between i and j atoms */
 778             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 779
 780             /* REACTION-FIELD ELECTROSTATICS */
 781             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 782
 783             fscal            = felec;
 784
 785             /* Update vectorial force */
 786             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 787             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 788             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 789
 790             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 791             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 792             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 793
 794             /**************************
 795              * CALCULATE INTERACTIONS *
 796              **************************/
 797
 798             /* Compute parameters for interactions between i and j atoms */
 799             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 800
 801             /* REACTION-FIELD ELECTROSTATICS */
 802             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 803
 804             fscal            = felec;
 805
 806             /* Update vectorial force */
 807             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 808             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 809             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 810
 811             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 812             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 813             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 814
 815             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 816
 817             /* Inner loop uses 120 flops */
 818         }
 819
 820         if(jidx<j_index_end)
 821         {
 822
 823             jnrA             = jjnr[jidx];
 824             j_coord_offsetA  = DIM*jnrA;
 825
 826             /* load j atom coordinates */
 827             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 828                                               &jx0,&jy0,&jz0);
 829
 830             /* Calculate displacement vector */
 831             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 832             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 833             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 834             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 835             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 836             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 837             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 838             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 839             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 840
 841             /* Calculate squared distance and things based on it */
 842             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 843             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 844             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 845
 846             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 847             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 848             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 849
 850             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 851             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 852             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 853
 854             /* Load parameters for j particles */
 855             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 856             vdwjidx0A        = 2*vdwtype[jnrA+0];
 857
 858             fjx0             = _fjsp_setzero_v2r8();
 859             fjy0             = _fjsp_setzero_v2r8();
 860             fjz0             = _fjsp_setzero_v2r8();
 861
 862             /**************************
 863              * CALCULATE INTERACTIONS *
 864              **************************/
 865
 866             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 867
 868             /* Compute parameters for interactions between i and j atoms */
 869             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 870             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 871
 872             /* Calculate table index by multiplying r with table scale and truncate to integer */
 873             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 874             itab_tmp         = _fjsp_dtox_v2r8(rt);
 875             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 876             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 877             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 878
 879             vfconv.i[0]     *= 8;
 880             vfconv.i[1]     *= 8;
 881
 882             /* REACTION-FIELD ELECTROSTATICS */
 883             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 884
 885             /* CUBIC SPLINE TABLE DISPERSION */
 886             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 887             F                = _fjsp_setzero_v2r8();
 888             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 889             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 890             H                = _fjsp_setzero_v2r8();
 891             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 892             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 893             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 894             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 895
 896             /* CUBIC SPLINE TABLE REPULSION */
 897             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 898             F                = _fjsp_setzero_v2r8();
 899             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 900             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 901             H                = _fjsp_setzero_v2r8();
 902             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 903             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 904             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 905             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 906             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 907
 908             fscal            = _fjsp_add_v2r8(felec,fvdw);
 909
 910             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 911
 912             /* Update vectorial force */
 913             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 914             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 915             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 916
 917             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 918             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 919             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             /* Compute parameters for interactions between i and j atoms */
 926             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 927
 928             /* REACTION-FIELD ELECTROSTATICS */
 929             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 930
 931             fscal            = felec;
 932
 933             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 934
 935             /* Update vectorial force */
 936             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 937             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 938             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 939
 940             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 941             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 942             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 943
 944             /**************************
 945              * CALCULATE INTERACTIONS *
 946              **************************/
 947
 948             /* Compute parameters for interactions between i and j atoms */
 949             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 950
 951             /* REACTION-FIELD ELECTROSTATICS */
 952             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 953
 954             fscal            = felec;
 955
 956             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 957
 958             /* Update vectorial force */
 959             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 960             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 961             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 962
 963             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 964             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 965             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 966
 967             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 968
 969             /* Inner loop uses 120 flops */
 970         }
 971
 972         /* End of innermost loop */
 973
 974         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 975                                               f+i_coord_offset,fshift+i_shift_offset);
 976
 977         /* Increment number of inner iterations */
 978         inneriter                  += j_index_end - j_index_start;
 979
 980         /* Outer loop uses 18 flops */
 981     }
 982
 983     /* Increment number of outer iterations */
 984     outeriter        += nri;
 985
 986     /* Update outer/inner flops */
 987
 988     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*120);
 989 }