src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_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_ElecRFCut_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_ElecRFCut_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     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 139     rcutoff_scalar   = fr->rcoulomb;
 140     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 141     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 142
 143     /* Avoid stupid compiler warnings */
 144     jnrA = jnrB = 0;
 145     j_coord_offsetA = 0;
 146     j_coord_offsetB = 0;
 147
 148     outeriter        = 0;
 149     inneriter        = 0;
 150
 151     /* Start outer loop over neighborlists */
 152     for(iidx=0; iidx<nri; iidx++)
 153     {
 154         /* Load shift vector for this list */
 155         i_shift_offset   = DIM*shiftidx[iidx];
 156
 157         /* Load limits for loop over neighbors */
 158         j_index_start    = jindex[iidx];
 159         j_index_end      = jindex[iidx+1];
 160
 161         /* Get outer coordinate index */
 162         inr              = iinr[iidx];
 163         i_coord_offset   = DIM*inr;
 164
 165         /* Load i particle coords and add shift vector */
 166         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 167                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 168
 169         fix0             = _fjsp_setzero_v2r8();
 170         fiy0             = _fjsp_setzero_v2r8();
 171         fiz0             = _fjsp_setzero_v2r8();
 172         fix1             = _fjsp_setzero_v2r8();
 173         fiy1             = _fjsp_setzero_v2r8();
 174         fiz1             = _fjsp_setzero_v2r8();
 175         fix2             = _fjsp_setzero_v2r8();
 176         fiy2             = _fjsp_setzero_v2r8();
 177         fiz2             = _fjsp_setzero_v2r8();
 178
 179         /* Reset potential sums */
 180         velecsum         = _fjsp_setzero_v2r8();
 181         vvdwsum          = _fjsp_setzero_v2r8();
 182
 183         /* Start inner kernel loop */
 184         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 185         {
 186
 187             /* Get j neighbor index, and coordinate index */
 188             jnrA             = jjnr[jidx];
 189             jnrB             = jjnr[jidx+1];
 190             j_coord_offsetA  = DIM*jnrA;
 191             j_coord_offsetB  = DIM*jnrB;
 192
 193             /* load j atom coordinates */
 194             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 195                                               &jx0,&jy0,&jz0);
 196
 197             /* Calculate displacement vector */
 198             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 199             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 200             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 201             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 202             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 203             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 204             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 205             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 206             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 207
 208             /* Calculate squared distance and things based on it */
 209             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 210             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 211             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216
 217             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 218             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 219             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 220
 221             /* Load parameters for j particles */
 222             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 223             vdwjidx0A        = 2*vdwtype[jnrA+0];
 224             vdwjidx0B        = 2*vdwtype[jnrB+0];
 225
 226             fjx0             = _fjsp_setzero_v2r8();
 227             fjy0             = _fjsp_setzero_v2r8();
 228             fjz0             = _fjsp_setzero_v2r8();
 229
 230             /**************************
 231              * CALCULATE INTERACTIONS *
 232              **************************/
 233
 234             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 235             {
 236
 237             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 238
 239             /* Compute parameters for interactions between i and j atoms */
 240             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 241             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 242                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 243
 244             /* Calculate table index by multiplying r with table scale and truncate to integer */
 245             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 246             itab_tmp         = _fjsp_dtox_v2r8(rt);
 247             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 248             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 249             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 250
 251             vfconv.i[0]     *= 8;
 252             vfconv.i[1]     *= 8;
 253
 254             /* REACTION-FIELD ELECTROSTATICS */
 255             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 256             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 257
 258             /* CUBIC SPLINE TABLE DISPERSION */
 259             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 260             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 261             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 262             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 263             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 264             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 265             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 266             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 267             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 268             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 269             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 270
 271             /* CUBIC SPLINE TABLE REPULSION */
 272             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 273             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 274             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 275             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 276             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 277             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 278             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 279             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 280             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 281             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 282             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 283             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 284             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 285
 286             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 287
 288             /* Update potential sum for this i atom from the interaction with this j atom. */
 289             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 290             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 291             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 292             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 293
 294             fscal            = _fjsp_add_v2r8(felec,fvdw);
 295
 296             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 297
 298             /* Update vectorial force */
 299             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 300             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 301             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 302
 303             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 304             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 305             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 306
 307             }
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 314             {
 315
 316             /* Compute parameters for interactions between i and j atoms */
 317             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 318
 319             /* REACTION-FIELD ELECTROSTATICS */
 320             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 321             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 322
 323             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 324
 325             /* Update potential sum for this i atom from the interaction with this j atom. */
 326             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 327             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 328
 329             fscal            = felec;
 330
 331             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 332
 333             /* Update vectorial force */
 334             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 335             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 336             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 337
 338             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 339             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 340             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 341
 342             }
 343
 344             /**************************
 345              * CALCULATE INTERACTIONS *
 346              **************************/
 347
 348             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 349             {
 350
 351             /* Compute parameters for interactions between i and j atoms */
 352             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 353
 354             /* REACTION-FIELD ELECTROSTATICS */
 355             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 356             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 357
 358             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 359
 360             /* Update potential sum for this i atom from the interaction with this j atom. */
 361             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 362             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 363
 364             fscal            = felec;
 365
 366             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 367
 368             /* Update vectorial force */
 369             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 370             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 371             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 372
 373             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 374             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 375             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 376
 377             }
 378
 379             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 380
 381             /* Inner loop uses 156 flops */
 382         }
 383
 384         if(jidx<j_index_end)
 385         {
 386
 387             jnrA             = jjnr[jidx];
 388             j_coord_offsetA  = DIM*jnrA;
 389
 390             /* load j atom coordinates */
 391             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 392                                               &jx0,&jy0,&jz0);
 393
 394             /* Calculate displacement vector */
 395             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 396             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 397             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 398             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 399             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 400             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 401             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 402             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 403             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 404
 405             /* Calculate squared distance and things based on it */
 406             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 407             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 408             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 409
 410             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 411             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 412             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 413
 414             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 415             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 416             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 417
 418             /* Load parameters for j particles */
 419             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 420             vdwjidx0A        = 2*vdwtype[jnrA+0];
 421
 422             fjx0             = _fjsp_setzero_v2r8();
 423             fjy0             = _fjsp_setzero_v2r8();
 424             fjz0             = _fjsp_setzero_v2r8();
 425
 426             /**************************
 427              * CALCULATE INTERACTIONS *
 428              **************************/
 429
 430             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 431             {
 432
 433             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 434
 435             /* Compute parameters for interactions between i and j atoms */
 436             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 437             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 438
 439             /* Calculate table index by multiplying r with table scale and truncate to integer */
 440             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 441             itab_tmp         = _fjsp_dtox_v2r8(rt);
 442             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 443             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 444             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 445
 446             vfconv.i[0]     *= 8;
 447             vfconv.i[1]     *= 8;
 448
 449             /* REACTION-FIELD ELECTROSTATICS */
 450             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 451             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 452
 453             /* CUBIC SPLINE TABLE DISPERSION */
 454             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 455             F                = _fjsp_setzero_v2r8();
 456             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 457             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 458             H                = _fjsp_setzero_v2r8();
 459             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 460             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 461             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 462             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 463             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 464             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 465
 466             /* CUBIC SPLINE TABLE REPULSION */
 467             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 468             F                = _fjsp_setzero_v2r8();
 469             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 470             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 471             H                = _fjsp_setzero_v2r8();
 472             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 473             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 474             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 475             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 476             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 477             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 478             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 479             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 480
 481             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 482
 483             /* Update potential sum for this i atom from the interaction with this j atom. */
 484             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 485             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 486             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 487             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 488             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 489             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 490
 491             fscal            = _fjsp_add_v2r8(felec,fvdw);
 492
 493             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 494
 495             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 496
 497             /* Update vectorial force */
 498             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 499             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 500             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 501
 502             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 503             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 504             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 505
 506             }
 507
 508             /**************************
 509              * CALCULATE INTERACTIONS *
 510              **************************/
 511
 512             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 513             {
 514
 515             /* Compute parameters for interactions between i and j atoms */
 516             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 517
 518             /* REACTION-FIELD ELECTROSTATICS */
 519             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 520             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 521
 522             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 526             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 527             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 528
 529             fscal            = felec;
 530
 531             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 532
 533             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 534
 535             /* Update vectorial force */
 536             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 537             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 538             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 539
 540             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 541             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 542             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 543
 544             }
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 551             {
 552
 553             /* Compute parameters for interactions between i and j atoms */
 554             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 555
 556             /* REACTION-FIELD ELECTROSTATICS */
 557             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 558             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 559
 560             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 561
 562             /* Update potential sum for this i atom from the interaction with this j atom. */
 563             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 564             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 565             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 566
 567             fscal            = felec;
 568
 569             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 570
 571             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 572
 573             /* Update vectorial force */
 574             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 575             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 576             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 577
 578             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 579             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 580             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 581
 582             }
 583
 584             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 585
 586             /* Inner loop uses 156 flops */
 587         }
 588
 589         /* End of innermost loop */
 590
 591         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 592                                               f+i_coord_offset,fshift+i_shift_offset);
 593
 594         ggid                        = gid[iidx];
 595         /* Update potential energies */
 596         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 597         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 598
 599         /* Increment number of inner iterations */
 600         inneriter                  += j_index_end - j_index_start;
 601
 602         /* Outer loop uses 20 flops */
 603     }
 604
 605     /* Increment number of outer iterations */
 606     outeriter        += nri;
 607
 608     /* Update outer/inner flops */
 609
 610     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*156);
 611 }
 612 /*
 613  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 614  * Electrostatics interaction: ReactionField
 615  * VdW interaction:            CubicSplineTable
 616  * Geometry:                   Water3-Particle
 617  * Calculate force/pot:        Force
 618  */
 619 void
 620 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 621                     (t_nblist                    * gmx_restrict       nlist,
 622                      rvec                        * gmx_restrict          xx,
 623                      rvec                        * gmx_restrict          ff,
 624                      t_forcerec                  * gmx_restrict          fr,
 625                      t_mdatoms                   * gmx_restrict     mdatoms,
 626                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 627                      t_nrnb                      * gmx_restrict        nrnb)
 628 {
 629     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 630      * just 0 for non-waters.
 631      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 632      * jnr indices corresponding to data put in the four positions in the SIMD register.
 633      */
 634     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 635     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 636     int              jnrA,jnrB;
 637     int              j_coord_offsetA,j_coord_offsetB;
 638     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 639     real             rcutoff_scalar;
 640     real             *shiftvec,*fshift,*x,*f;
 641     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 642     int              vdwioffset0;
 643     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 644     int              vdwioffset1;
 645     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 646     int              vdwioffset2;
 647     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 648     int              vdwjidx0A,vdwjidx0B;
 649     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 650     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 651     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 652     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 653     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 654     real             *charge;
 655     int              nvdwtype;
 656     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 657     int              *vdwtype;
 658     real             *vdwparam;
 659     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 660     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 661     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 662     real             *vftab;
 663     _fjsp_v2r8       itab_tmp;
 664     _fjsp_v2r8       dummy_mask,cutoff_mask;
 665     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 666     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 667     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 668
 669     x                = xx[0];
 670     f                = ff[0];
 671
 672     nri              = nlist->nri;
 673     iinr             = nlist->iinr;
 674     jindex           = nlist->jindex;
 675     jjnr             = nlist->jjnr;
 676     shiftidx         = nlist->shift;
 677     gid              = nlist->gid;
 678     shiftvec         = fr->shift_vec[0];
 679     fshift           = fr->fshift[0];
 680     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 681     charge           = mdatoms->chargeA;
 682     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 683     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 684     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 685     nvdwtype         = fr->ntype;
 686     vdwparam         = fr->nbfp;
 687     vdwtype          = mdatoms->typeA;
 688
 689     vftab            = kernel_data->table_vdw->data;
 690     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 691
 692     /* Setup water-specific parameters */
 693     inr              = nlist->iinr[0];
 694     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 695     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 696     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 697     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 698
 699     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 700     rcutoff_scalar   = fr->rcoulomb;
 701     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 702     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 703
 704     /* Avoid stupid compiler warnings */
 705     jnrA = jnrB = 0;
 706     j_coord_offsetA = 0;
 707     j_coord_offsetB = 0;
 708
 709     outeriter        = 0;
 710     inneriter        = 0;
 711
 712     /* Start outer loop over neighborlists */
 713     for(iidx=0; iidx<nri; iidx++)
 714     {
 715         /* Load shift vector for this list */
 716         i_shift_offset   = DIM*shiftidx[iidx];
 717
 718         /* Load limits for loop over neighbors */
 719         j_index_start    = jindex[iidx];
 720         j_index_end      = jindex[iidx+1];
 721
 722         /* Get outer coordinate index */
 723         inr              = iinr[iidx];
 724         i_coord_offset   = DIM*inr;
 725
 726         /* Load i particle coords and add shift vector */
 727         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 728                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 729
 730         fix0             = _fjsp_setzero_v2r8();
 731         fiy0             = _fjsp_setzero_v2r8();
 732         fiz0             = _fjsp_setzero_v2r8();
 733         fix1             = _fjsp_setzero_v2r8();
 734         fiy1             = _fjsp_setzero_v2r8();
 735         fiz1             = _fjsp_setzero_v2r8();
 736         fix2             = _fjsp_setzero_v2r8();
 737         fiy2             = _fjsp_setzero_v2r8();
 738         fiz2             = _fjsp_setzero_v2r8();
 739
 740         /* Start inner kernel loop */
 741         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 742         {
 743
 744             /* Get j neighbor index, and coordinate index */
 745             jnrA             = jjnr[jidx];
 746             jnrB             = jjnr[jidx+1];
 747             j_coord_offsetA  = DIM*jnrA;
 748             j_coord_offsetB  = DIM*jnrB;
 749
 750             /* load j atom coordinates */
 751             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 752                                               &jx0,&jy0,&jz0);
 753
 754             /* Calculate displacement vector */
 755             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 756             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 757             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 758             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 759             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 760             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 761             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 762             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 763             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 764
 765             /* Calculate squared distance and things based on it */
 766             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 767             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 768             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 769
 770             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 771             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 772             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 773
 774             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 775             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 776             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 777
 778             /* Load parameters for j particles */
 779             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 780             vdwjidx0A        = 2*vdwtype[jnrA+0];
 781             vdwjidx0B        = 2*vdwtype[jnrB+0];
 782
 783             fjx0             = _fjsp_setzero_v2r8();
 784             fjy0             = _fjsp_setzero_v2r8();
 785             fjz0             = _fjsp_setzero_v2r8();
 786
 787             /**************************
 788              * CALCULATE INTERACTIONS *
 789              **************************/
 790
 791             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 792             {
 793
 794             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 795
 796             /* Compute parameters for interactions between i and j atoms */
 797             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 798             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 799                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 800
 801             /* Calculate table index by multiplying r with table scale and truncate to integer */
 802             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 803             itab_tmp         = _fjsp_dtox_v2r8(rt);
 804             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 805             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 806             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 807
 808             vfconv.i[0]     *= 8;
 809             vfconv.i[1]     *= 8;
 810
 811             /* REACTION-FIELD ELECTROSTATICS */
 812             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 813
 814             /* CUBIC SPLINE TABLE DISPERSION */
 815             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 816             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 817             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 818             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 819             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 820             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 821             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 822             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 823             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 824
 825             /* CUBIC SPLINE TABLE REPULSION */
 826             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 827             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 828             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 829             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 830             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 831             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 832             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 833             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 834             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 835             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 836
 837             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 838
 839             fscal            = _fjsp_add_v2r8(felec,fvdw);
 840
 841             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 842
 843             /* Update vectorial force */
 844             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 845             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 846             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 847
 848             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 849             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 850             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 851
 852             }
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 859             {
 860
 861             /* Compute parameters for interactions between i and j atoms */
 862             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 863
 864             /* REACTION-FIELD ELECTROSTATICS */
 865             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 866
 867             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 868
 869             fscal            = felec;
 870
 871             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 872
 873             /* Update vectorial force */
 874             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 875             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 876             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 877
 878             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 879             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 880             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 881
 882             }
 883
 884             /**************************
 885              * CALCULATE INTERACTIONS *
 886              **************************/
 887
 888             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 889             {
 890
 891             /* Compute parameters for interactions between i and j atoms */
 892             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 893
 894             /* REACTION-FIELD ELECTROSTATICS */
 895             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 896
 897             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 898
 899             fscal            = felec;
 900
 901             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 902
 903             /* Update vectorial force */
 904             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 905             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 906             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 907
 908             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 909             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 910             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 911
 912             }
 913
 914             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 915
 916             /* Inner loop uses 129 flops */
 917         }
 918
 919         if(jidx<j_index_end)
 920         {
 921
 922             jnrA             = jjnr[jidx];
 923             j_coord_offsetA  = DIM*jnrA;
 924
 925             /* load j atom coordinates */
 926             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 927                                               &jx0,&jy0,&jz0);
 928
 929             /* Calculate displacement vector */
 930             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 931             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 932             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 933             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 934             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 935             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 936             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 937             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 938             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 939
 940             /* Calculate squared distance and things based on it */
 941             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 942             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 943             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 944
 945             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 946             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 947             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 948
 949             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 950             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 951             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 952
 953             /* Load parameters for j particles */
 954             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 955             vdwjidx0A        = 2*vdwtype[jnrA+0];
 956
 957             fjx0             = _fjsp_setzero_v2r8();
 958             fjy0             = _fjsp_setzero_v2r8();
 959             fjz0             = _fjsp_setzero_v2r8();
 960
 961             /**************************
 962              * CALCULATE INTERACTIONS *
 963              **************************/
 964
 965             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 966             {
 967
 968             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 969
 970             /* Compute parameters for interactions between i and j atoms */
 971             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 972             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 973
 974             /* Calculate table index by multiplying r with table scale and truncate to integer */
 975             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 976             itab_tmp         = _fjsp_dtox_v2r8(rt);
 977             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 978             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 979             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 980
 981             vfconv.i[0]     *= 8;
 982             vfconv.i[1]     *= 8;
 983
 984             /* REACTION-FIELD ELECTROSTATICS */
 985             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 986
 987             /* CUBIC SPLINE TABLE DISPERSION */
 988             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 989             F                = _fjsp_setzero_v2r8();
 990             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 991             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 992             H                = _fjsp_setzero_v2r8();
 993             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 994             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 995             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 996             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 997
 998             /* CUBIC SPLINE TABLE REPULSION */
 999             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1000             F                = _fjsp_setzero_v2r8();
1001             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1002             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1003             H                = _fjsp_setzero_v2r8();
1004             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1005             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1006             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1007             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1008             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1009
1010             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1011
1012             fscal            = _fjsp_add_v2r8(felec,fvdw);
1013
1014             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1015
1016             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1017
1018             /* Update vectorial force */
1019             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1020             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1021             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1022
1023             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1024             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1025             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1026
1027             }
1028
1029             /**************************
1030              * CALCULATE INTERACTIONS *
1031              **************************/
1032
1033             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1034             {
1035
1036             /* Compute parameters for interactions between i and j atoms */
1037             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1038
1039             /* REACTION-FIELD ELECTROSTATICS */
1040             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1041
1042             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1043
1044             fscal            = felec;
1045
1046             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1047
1048             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1049
1050             /* Update vectorial force */
1051             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1052             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1053             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1054
1055             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1056             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1057             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1058
1059             }
1060
1061             /**************************
1062              * CALCULATE INTERACTIONS *
1063              **************************/
1064
1065             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1066             {
1067
1068             /* Compute parameters for interactions between i and j atoms */
1069             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1070
1071             /* REACTION-FIELD ELECTROSTATICS */
1072             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1073
1074             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1075
1076             fscal            = felec;
1077
1078             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1079
1080             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1081
1082             /* Update vectorial force */
1083             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1084             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1085             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1086
1087             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1088             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1089             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1090
1091             }
1092
1093             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1094
1095             /* Inner loop uses 129 flops */
1096         }
1097
1098         /* End of innermost loop */
1099
1100         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1101                                               f+i_coord_offset,fshift+i_shift_offset);
1102
1103         /* Increment number of inner iterations */
1104         inneriter                  += j_index_end - j_index_start;
1105
1106         /* Outer loop uses 18 flops */
1107     }
1108
1109     /* Increment number of outer iterations */
1110     outeriter        += nri;
1111
1112     /* Update outer/inner flops */
1113
1114     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*129);
1115 }