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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 438                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 439
 440             /* Calculate table index by multiplying r with table scale and truncate to integer */
 441             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 442             itab_tmp         = _fjsp_dtox_v2r8(rt);
 443             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 444             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 445             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 446
 447             vfconv.i[0]     *= 8;
 448             vfconv.i[1]     *= 8;
 449
 450             /* REACTION-FIELD ELECTROSTATICS */
 451             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 452             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 453
 454             /* CUBIC SPLINE TABLE DISPERSION */
 455             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 456             F                = _fjsp_setzero_v2r8();
 457             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 458             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 459             H                = _fjsp_setzero_v2r8();
 460             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 461             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 462             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 463             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 464             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 465             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 466
 467             /* CUBIC SPLINE TABLE REPULSION */
 468             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 469             F                = _fjsp_setzero_v2r8();
 470             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 471             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 472             H                = _fjsp_setzero_v2r8();
 473             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 474             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 475             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 476             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 477             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 478             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 479             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 480             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 481
 482             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 483
 484             /* Update potential sum for this i atom from the interaction with this j atom. */
 485             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 486             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 487             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 488             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 489             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 490             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 491
 492             fscal            = _fjsp_add_v2r8(felec,fvdw);
 493
 494             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 495
 496             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 497
 498             /* Update vectorial force */
 499             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 500             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 501             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 502
 503             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 504             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 505             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 506
 507             }
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 514             {
 515
 516             /* Compute parameters for interactions between i and j atoms */
 517             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 518
 519             /* REACTION-FIELD ELECTROSTATICS */
 520             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 521             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 522
 523             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 524
 525             /* Update potential sum for this i atom from the interaction with this j atom. */
 526             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 527             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 528             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 533
 534             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 535
 536             /* Update vectorial force */
 537             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 538             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 539             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 540
 541             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 542             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 543             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 544
 545             }
 546
 547             /**************************
 548              * CALCULATE INTERACTIONS *
 549              **************************/
 550
 551             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 552             {
 553
 554             /* Compute parameters for interactions between i and j atoms */
 555             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 556
 557             /* REACTION-FIELD ELECTROSTATICS */
 558             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 559             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 560
 561             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 562
 563             /* Update potential sum for this i atom from the interaction with this j atom. */
 564             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 565             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 566             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 567
 568             fscal            = felec;
 569
 570             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 571
 572             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 573
 574             /* Update vectorial force */
 575             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 576             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 577             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 578
 579             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 580             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 581             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 582
 583             }
 584
 585             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 586
 587             /* Inner loop uses 156 flops */
 588         }
 589
 590         /* End of innermost loop */
 591
 592         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 593                                               f+i_coord_offset,fshift+i_shift_offset);
 594
 595         ggid                        = gid[iidx];
 596         /* Update potential energies */
 597         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 598         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 599
 600         /* Increment number of inner iterations */
 601         inneriter                  += j_index_end - j_index_start;
 602
 603         /* Outer loop uses 20 flops */
 604     }
 605
 606     /* Increment number of outer iterations */
 607     outeriter        += nri;
 608
 609     /* Update outer/inner flops */
 610
 611     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*156);
 612 }
 613 /*
 614  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 615  * Electrostatics interaction: ReactionField
 616  * VdW interaction:            CubicSplineTable
 617  * Geometry:                   Water3-Particle
 618  * Calculate force/pot:        Force
 619  */
 620 void
 621 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 622                     (t_nblist                    * gmx_restrict       nlist,
 623                      rvec                        * gmx_restrict          xx,
 624                      rvec                        * gmx_restrict          ff,
 625                      t_forcerec                  * gmx_restrict          fr,
 626                      t_mdatoms                   * gmx_restrict     mdatoms,
 627                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 628                      t_nrnb                      * gmx_restrict        nrnb)
 629 {
 630     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 631      * just 0 for non-waters.
 632      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 633      * jnr indices corresponding to data put in the four positions in the SIMD register.
 634      */
 635     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 636     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 637     int              jnrA,jnrB;
 638     int              j_coord_offsetA,j_coord_offsetB;
 639     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 640     real             rcutoff_scalar;
 641     real             *shiftvec,*fshift,*x,*f;
 642     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 643     int              vdwioffset0;
 644     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 645     int              vdwioffset1;
 646     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 647     int              vdwioffset2;
 648     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 649     int              vdwjidx0A,vdwjidx0B;
 650     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 651     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 652     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 653     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 654     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 655     real             *charge;
 656     int              nvdwtype;
 657     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 658     int              *vdwtype;
 659     real             *vdwparam;
 660     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 661     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 662     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 663     real             *vftab;
 664     _fjsp_v2r8       itab_tmp;
 665     _fjsp_v2r8       dummy_mask,cutoff_mask;
 666     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 667     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 668     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 669
 670     x                = xx[0];
 671     f                = ff[0];
 672
 673     nri              = nlist->nri;
 674     iinr             = nlist->iinr;
 675     jindex           = nlist->jindex;
 676     jjnr             = nlist->jjnr;
 677     shiftidx         = nlist->shift;
 678     gid              = nlist->gid;
 679     shiftvec         = fr->shift_vec[0];
 680     fshift           = fr->fshift[0];
 681     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 682     charge           = mdatoms->chargeA;
 683     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 684     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 685     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 686     nvdwtype         = fr->ntype;
 687     vdwparam         = fr->nbfp;
 688     vdwtype          = mdatoms->typeA;
 689
 690     vftab            = kernel_data->table_vdw->data;
 691     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 692
 693     /* Setup water-specific parameters */
 694     inr              = nlist->iinr[0];
 695     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 696     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 697     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 698     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 699
 700     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 701     rcutoff_scalar   = fr->rcoulomb;
 702     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 703     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 704
 705     /* Avoid stupid compiler warnings */
 706     jnrA = jnrB = 0;
 707     j_coord_offsetA = 0;
 708     j_coord_offsetB = 0;
 709
 710     outeriter        = 0;
 711     inneriter        = 0;
 712
 713     /* Start outer loop over neighborlists */
 714     for(iidx=0; iidx<nri; iidx++)
 715     {
 716         /* Load shift vector for this list */
 717         i_shift_offset   = DIM*shiftidx[iidx];
 718
 719         /* Load limits for loop over neighbors */
 720         j_index_start    = jindex[iidx];
 721         j_index_end      = jindex[iidx+1];
 722
 723         /* Get outer coordinate index */
 724         inr              = iinr[iidx];
 725         i_coord_offset   = DIM*inr;
 726
 727         /* Load i particle coords and add shift vector */
 728         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 729                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 730
 731         fix0             = _fjsp_setzero_v2r8();
 732         fiy0             = _fjsp_setzero_v2r8();
 733         fiz0             = _fjsp_setzero_v2r8();
 734         fix1             = _fjsp_setzero_v2r8();
 735         fiy1             = _fjsp_setzero_v2r8();
 736         fiz1             = _fjsp_setzero_v2r8();
 737         fix2             = _fjsp_setzero_v2r8();
 738         fiy2             = _fjsp_setzero_v2r8();
 739         fiz2             = _fjsp_setzero_v2r8();
 740
 741         /* Start inner kernel loop */
 742         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 743         {
 744
 745             /* Get j neighbor index, and coordinate index */
 746             jnrA             = jjnr[jidx];
 747             jnrB             = jjnr[jidx+1];
 748             j_coord_offsetA  = DIM*jnrA;
 749             j_coord_offsetB  = DIM*jnrB;
 750
 751             /* load j atom coordinates */
 752             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 753                                               &jx0,&jy0,&jz0);
 754
 755             /* Calculate displacement vector */
 756             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 757             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 758             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 759             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 760             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 761             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 762             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 763             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 764             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 765
 766             /* Calculate squared distance and things based on it */
 767             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 768             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 769             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 770
 771             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 772             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 773             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 774
 775             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 776             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 777             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 778
 779             /* Load parameters for j particles */
 780             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 781             vdwjidx0A        = 2*vdwtype[jnrA+0];
 782             vdwjidx0B        = 2*vdwtype[jnrB+0];
 783
 784             fjx0             = _fjsp_setzero_v2r8();
 785             fjy0             = _fjsp_setzero_v2r8();
 786             fjz0             = _fjsp_setzero_v2r8();
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 793             {
 794
 795             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 796
 797             /* Compute parameters for interactions between i and j atoms */
 798             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 799             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 800                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 801
 802             /* Calculate table index by multiplying r with table scale and truncate to integer */
 803             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 804             itab_tmp         = _fjsp_dtox_v2r8(rt);
 805             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 806             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 807             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 808
 809             vfconv.i[0]     *= 8;
 810             vfconv.i[1]     *= 8;
 811
 812             /* REACTION-FIELD ELECTROSTATICS */
 813             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 814
 815             /* CUBIC SPLINE TABLE DISPERSION */
 816             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 817             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 818             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 819             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 820             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 821             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 822             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 823             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 824             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 825
 826             /* CUBIC SPLINE TABLE REPULSION */
 827             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 828             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 829             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 830             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 831             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 832             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 833             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 834             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 835             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 836             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 837
 838             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 839
 840             fscal            = _fjsp_add_v2r8(felec,fvdw);
 841
 842             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 843
 844             /* Update vectorial force */
 845             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 846             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 847             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 848
 849             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 850             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 851             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 852
 853             }
 854
 855             /**************************
 856              * CALCULATE INTERACTIONS *
 857              **************************/
 858
 859             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 860             {
 861
 862             /* Compute parameters for interactions between i and j atoms */
 863             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 864
 865             /* REACTION-FIELD ELECTROSTATICS */
 866             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 867
 868             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 869
 870             fscal            = felec;
 871
 872             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 873
 874             /* Update vectorial force */
 875             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 876             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 877             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 878
 879             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 880             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 881             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 882
 883             }
 884
 885             /**************************
 886              * CALCULATE INTERACTIONS *
 887              **************************/
 888
 889             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 890             {
 891
 892             /* Compute parameters for interactions between i and j atoms */
 893             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 894
 895             /* REACTION-FIELD ELECTROSTATICS */
 896             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 897
 898             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 899
 900             fscal            = felec;
 901
 902             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 903
 904             /* Update vectorial force */
 905             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 906             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 907             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 908
 909             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 910             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 911             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 912
 913             }
 914
 915             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 916
 917             /* Inner loop uses 129 flops */
 918         }
 919
 920         if(jidx<j_index_end)
 921         {
 922
 923             jnrA             = jjnr[jidx];
 924             j_coord_offsetA  = DIM*jnrA;
 925
 926             /* load j atom coordinates */
 927             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 928                                               &jx0,&jy0,&jz0);
 929
 930             /* Calculate displacement vector */
 931             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 932             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 933             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 934             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 935             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 936             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 937             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 938             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 939             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 940
 941             /* Calculate squared distance and things based on it */
 942             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 943             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 944             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 945
 946             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 947             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 948             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 949
 950             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 951             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 952             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 953
 954             /* Load parameters for j particles */
 955             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 956             vdwjidx0A        = 2*vdwtype[jnrA+0];
 957
 958             fjx0             = _fjsp_setzero_v2r8();
 959             fjy0             = _fjsp_setzero_v2r8();
 960             fjz0             = _fjsp_setzero_v2r8();
 961
 962             /**************************
 963              * CALCULATE INTERACTIONS *
 964              **************************/
 965
 966             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 967             {
 968
 969             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 970
 971             /* Compute parameters for interactions between i and j atoms */
 972             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 973             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 974                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 975
 976             /* Calculate table index by multiplying r with table scale and truncate to integer */
 977             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 978             itab_tmp         = _fjsp_dtox_v2r8(rt);
 979             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 980             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 981             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 982
 983             vfconv.i[0]     *= 8;
 984             vfconv.i[1]     *= 8;
 985
 986             /* REACTION-FIELD ELECTROSTATICS */
 987             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 988
 989             /* CUBIC SPLINE TABLE DISPERSION */
 990             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 991             F                = _fjsp_setzero_v2r8();
 992             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 993             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 994             H                = _fjsp_setzero_v2r8();
 995             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 996             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 997             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 998             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 999
1000             /* CUBIC SPLINE TABLE REPULSION */
1001             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1002             F                = _fjsp_setzero_v2r8();
1003             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1004             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1005             H                = _fjsp_setzero_v2r8();
1006             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1007             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1008             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1009             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1010             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1011
1012             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1013
1014             fscal            = _fjsp_add_v2r8(felec,fvdw);
1015
1016             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1017
1018             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1019
1020             /* Update vectorial force */
1021             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1022             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1023             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1024
1025             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1026             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1027             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1028
1029             }
1030
1031             /**************************
1032              * CALCULATE INTERACTIONS *
1033              **************************/
1034
1035             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1036             {
1037
1038             /* Compute parameters for interactions between i and j atoms */
1039             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1040
1041             /* REACTION-FIELD ELECTROSTATICS */
1042             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1043
1044             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1045
1046             fscal            = felec;
1047
1048             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1049
1050             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1051
1052             /* Update vectorial force */
1053             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1054             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1055             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1056
1057             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1058             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1059             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1060
1061             }
1062
1063             /**************************
1064              * CALCULATE INTERACTIONS *
1065              **************************/
1066
1067             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1068             {
1069
1070             /* Compute parameters for interactions between i and j atoms */
1071             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1072
1073             /* REACTION-FIELD ELECTROSTATICS */
1074             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1075
1076             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1077
1078             fscal            = felec;
1079
1080             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1081
1082             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1083
1084             /* Update vectorial force */
1085             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1086             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1087             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1088
1089             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1090             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1091             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1092
1093             }
1094
1095             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1096
1097             /* Inner loop uses 129 flops */
1098         }
1099
1100         /* End of innermost loop */
1101
1102         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1103                                               f+i_coord_offset,fshift+i_shift_offset);
1104
1105         /* Increment number of inner iterations */
1106         inneriter                  += j_index_end - j_index_start;
1107
1108         /* Outer loop uses 18 flops */
1109     }
1110
1111     /* Increment number of outer iterations */
1112     outeriter        += nri;
1113
1114     /* Update outer/inner flops */
1115
1116     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*129);
1117 }