src/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012, by the GROMACS development team, led by
   5  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   6  * others, as listed in the AUTHORS file in the top-level source
   7  * 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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwLJSw_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_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       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 101     real             rswitch_scalar,d_scalar;
 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     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 136     rcutoff_scalar   = fr->rcoulomb;
 137     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 138     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 139
 140     rswitch_scalar   = fr->rvdw_switch;
 141     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 142     /* Setup switch parameters */
 143     d_scalar         = rcutoff_scalar-rswitch_scalar;
 144     d                = gmx_fjsp_set1_v2r8(d_scalar);
 145     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 146     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 147     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 148     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 149     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 150     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 151
 152     /* Avoid stupid compiler warnings */
 153     jnrA = jnrB = 0;
 154     j_coord_offsetA = 0;
 155     j_coord_offsetB = 0;
 156
 157     outeriter        = 0;
 158     inneriter        = 0;
 159
 160     /* Start outer loop over neighborlists */
 161     for(iidx=0; iidx<nri; iidx++)
 162     {
 163         /* Load shift vector for this list */
 164         i_shift_offset   = DIM*shiftidx[iidx];
 165
 166         /* Load limits for loop over neighbors */
 167         j_index_start    = jindex[iidx];
 168         j_index_end      = jindex[iidx+1];
 169
 170         /* Get outer coordinate index */
 171         inr              = iinr[iidx];
 172         i_coord_offset   = DIM*inr;
 173
 174         /* Load i particle coords and add shift vector */
 175         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 176                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 177
 178         fix0             = _fjsp_setzero_v2r8();
 179         fiy0             = _fjsp_setzero_v2r8();
 180         fiz0             = _fjsp_setzero_v2r8();
 181         fix1             = _fjsp_setzero_v2r8();
 182         fiy1             = _fjsp_setzero_v2r8();
 183         fiz1             = _fjsp_setzero_v2r8();
 184         fix2             = _fjsp_setzero_v2r8();
 185         fiy2             = _fjsp_setzero_v2r8();
 186         fiz2             = _fjsp_setzero_v2r8();
 187
 188         /* Reset potential sums */
 189         velecsum         = _fjsp_setzero_v2r8();
 190         vvdwsum          = _fjsp_setzero_v2r8();
 191
 192         /* Start inner kernel loop */
 193         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 194         {
 195
 196             /* Get j neighbor index, and coordinate index */
 197             jnrA             = jjnr[jidx];
 198             jnrB             = jjnr[jidx+1];
 199             j_coord_offsetA  = DIM*jnrA;
 200             j_coord_offsetB  = DIM*jnrB;
 201
 202             /* load j atom coordinates */
 203             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 204                                               &jx0,&jy0,&jz0);
 205
 206             /* Calculate displacement vector */
 207             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 208             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 209             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 210             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 211             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 212             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 213             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 214             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 215             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 216
 217             /* Calculate squared distance and things based on it */
 218             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 219             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 220             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 221
 222             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 223             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 224             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 225
 226             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 227             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 228             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 229
 230             /* Load parameters for j particles */
 231             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 232             vdwjidx0A        = 2*vdwtype[jnrA+0];
 233             vdwjidx0B        = 2*vdwtype[jnrB+0];
 234
 235             fjx0             = _fjsp_setzero_v2r8();
 236             fjy0             = _fjsp_setzero_v2r8();
 237             fjz0             = _fjsp_setzero_v2r8();
 238
 239             /**************************
 240              * CALCULATE INTERACTIONS *
 241              **************************/
 242
 243             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 244             {
 245
 246             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 247
 248             /* Compute parameters for interactions between i and j atoms */
 249             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 250             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 251                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 252
 253             /* REACTION-FIELD ELECTROSTATICS */
 254             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 255             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 256
 257             /* LENNARD-JONES DISPERSION/REPULSION */
 258
 259             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 260             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 261             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 262             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 263             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 264
 265             d                = _fjsp_sub_v2r8(r00,rswitch);
 266             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 267             d2               = _fjsp_mul_v2r8(d,d);
 268             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 269
 270             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 271
 272             /* Evaluate switch function */
 273             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 274             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 275             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 276             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 277
 278             /* Update potential sum for this i atom from the interaction with this j atom. */
 279             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 280             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 281             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 282             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 283
 284             fscal            = _fjsp_add_v2r8(felec,fvdw);
 285
 286             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 287
 288             /* Update vectorial force */
 289             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 290             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 291             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 292
 293             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 294             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 295             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 296
 297             }
 298
 299             /**************************
 300              * CALCULATE INTERACTIONS *
 301              **************************/
 302
 303             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 304             {
 305
 306             /* Compute parameters for interactions between i and j atoms */
 307             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 308
 309             /* REACTION-FIELD ELECTROSTATICS */
 310             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 311             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 312
 313             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 314
 315             /* Update potential sum for this i atom from the interaction with this j atom. */
 316             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 317             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 318
 319             fscal            = felec;
 320
 321             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 322
 323             /* Update vectorial force */
 324             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 325             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 326             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 327
 328             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 329             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 330             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 331
 332             }
 333
 334             /**************************
 335              * CALCULATE INTERACTIONS *
 336              **************************/
 337
 338             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 339             {
 340
 341             /* Compute parameters for interactions between i and j atoms */
 342             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 343
 344             /* REACTION-FIELD ELECTROSTATICS */
 345             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 346             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 347
 348             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 349
 350             /* Update potential sum for this i atom from the interaction with this j atom. */
 351             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 352             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 353
 354             fscal            = felec;
 355
 356             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 357
 358             /* Update vectorial force */
 359             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 360             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 361             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 362
 363             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 364             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 365             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 366
 367             }
 368
 369             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 370
 371             /* Inner loop uses 154 flops */
 372         }
 373
 374         if(jidx<j_index_end)
 375         {
 376
 377             jnrA             = jjnr[jidx];
 378             j_coord_offsetA  = DIM*jnrA;
 379
 380             /* load j atom coordinates */
 381             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 382                                               &jx0,&jy0,&jz0);
 383
 384             /* Calculate displacement vector */
 385             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 386             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 387             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 388             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 389             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 390             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 391             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 392             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 393             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 394
 395             /* Calculate squared distance and things based on it */
 396             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 397             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 398             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 399
 400             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 401             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 402             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 403
 404             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 405             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 406             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 407
 408             /* Load parameters for j particles */
 409             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 410             vdwjidx0A        = 2*vdwtype[jnrA+0];
 411
 412             fjx0             = _fjsp_setzero_v2r8();
 413             fjy0             = _fjsp_setzero_v2r8();
 414             fjz0             = _fjsp_setzero_v2r8();
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 421             {
 422
 423             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 424
 425             /* Compute parameters for interactions between i and j atoms */
 426             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 427             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 428
 429             /* REACTION-FIELD ELECTROSTATICS */
 430             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 431             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 432
 433             /* LENNARD-JONES DISPERSION/REPULSION */
 434
 435             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 436             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 437             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 438             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 439             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 440
 441             d                = _fjsp_sub_v2r8(r00,rswitch);
 442             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 443             d2               = _fjsp_mul_v2r8(d,d);
 444             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 445
 446             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 447
 448             /* Evaluate switch function */
 449             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 450             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 451             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 452             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 453
 454             /* Update potential sum for this i atom from the interaction with this j atom. */
 455             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 456             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 457             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 458             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 459             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 460             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 461
 462             fscal            = _fjsp_add_v2r8(felec,fvdw);
 463
 464             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 465
 466             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 467
 468             /* Update vectorial force */
 469             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 470             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 471             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 472
 473             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 474             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 475             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 476
 477             }
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 484             {
 485
 486             /* Compute parameters for interactions between i and j atoms */
 487             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 488
 489             /* REACTION-FIELD ELECTROSTATICS */
 490             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 491             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 492
 493             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 497             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 498             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 499
 500             fscal            = felec;
 501
 502             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 508             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 509             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 514
 515             }
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 522             {
 523
 524             /* Compute parameters for interactions between i and j atoms */
 525             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 526
 527             /* REACTION-FIELD ELECTROSTATICS */
 528             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 529             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 530
 531             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 532
 533             /* Update potential sum for this i atom from the interaction with this j atom. */
 534             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 535             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 536             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 537
 538             fscal            = felec;
 539
 540             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 541
 542             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 543
 544             /* Update vectorial force */
 545             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 546             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 547             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 548
 549             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 550             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 551             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 552
 553             }
 554
 555             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 556
 557             /* Inner loop uses 154 flops */
 558         }
 559
 560         /* End of innermost loop */
 561
 562         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 563                                               f+i_coord_offset,fshift+i_shift_offset);
 564
 565         ggid                        = gid[iidx];
 566         /* Update potential energies */
 567         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 568         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 569
 570         /* Increment number of inner iterations */
 571         inneriter                  += j_index_end - j_index_start;
 572
 573         /* Outer loop uses 20 flops */
 574     }
 575
 576     /* Increment number of outer iterations */
 577     outeriter        += nri;
 578
 579     /* Update outer/inner flops */
 580
 581     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 582 }
 583 /*
 584  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 585  * Electrostatics interaction: ReactionField
 586  * VdW interaction:            LennardJones
 587  * Geometry:                   Water3-Particle
 588  * Calculate force/pot:        Force
 589  */
 590 void
 591 nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 592                     (t_nblist * gmx_restrict                nlist,
 593                      rvec * gmx_restrict                    xx,
 594                      rvec * gmx_restrict                    ff,
 595                      t_forcerec * gmx_restrict              fr,
 596                      t_mdatoms * gmx_restrict               mdatoms,
 597                      nb_kernel_data_t * gmx_restrict        kernel_data,
 598                      t_nrnb * gmx_restrict                  nrnb)
 599 {
 600     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 601      * just 0 for non-waters.
 602      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 603      * jnr indices corresponding to data put in the four positions in the SIMD register.
 604      */
 605     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 606     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 607     int              jnrA,jnrB;
 608     int              j_coord_offsetA,j_coord_offsetB;
 609     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 610     real             rcutoff_scalar;
 611     real             *shiftvec,*fshift,*x,*f;
 612     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 613     int              vdwioffset0;
 614     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 615     int              vdwioffset1;
 616     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 617     int              vdwioffset2;
 618     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 619     int              vdwjidx0A,vdwjidx0B;
 620     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 621     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 622     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 623     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 624     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 625     real             *charge;
 626     int              nvdwtype;
 627     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 628     int              *vdwtype;
 629     real             *vdwparam;
 630     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 631     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 632     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 633     real             rswitch_scalar,d_scalar;
 634     _fjsp_v2r8       itab_tmp;
 635     _fjsp_v2r8       dummy_mask,cutoff_mask;
 636     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 637     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 638     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 639
 640     x                = xx[0];
 641     f                = ff[0];
 642
 643     nri              = nlist->nri;
 644     iinr             = nlist->iinr;
 645     jindex           = nlist->jindex;
 646     jjnr             = nlist->jjnr;
 647     shiftidx         = nlist->shift;
 648     gid              = nlist->gid;
 649     shiftvec         = fr->shift_vec[0];
 650     fshift           = fr->fshift[0];
 651     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 652     charge           = mdatoms->chargeA;
 653     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 654     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 655     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 656     nvdwtype         = fr->ntype;
 657     vdwparam         = fr->nbfp;
 658     vdwtype          = mdatoms->typeA;
 659
 660     /* Setup water-specific parameters */
 661     inr              = nlist->iinr[0];
 662     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 663     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 664     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 665     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 666
 667     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 668     rcutoff_scalar   = fr->rcoulomb;
 669     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 670     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 671
 672     rswitch_scalar   = fr->rvdw_switch;
 673     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 674     /* Setup switch parameters */
 675     d_scalar         = rcutoff_scalar-rswitch_scalar;
 676     d                = gmx_fjsp_set1_v2r8(d_scalar);
 677     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 678     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 679     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 680     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 681     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 682     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 683
 684     /* Avoid stupid compiler warnings */
 685     jnrA = jnrB = 0;
 686     j_coord_offsetA = 0;
 687     j_coord_offsetB = 0;
 688
 689     outeriter        = 0;
 690     inneriter        = 0;
 691
 692     /* Start outer loop over neighborlists */
 693     for(iidx=0; iidx<nri; iidx++)
 694     {
 695         /* Load shift vector for this list */
 696         i_shift_offset   = DIM*shiftidx[iidx];
 697
 698         /* Load limits for loop over neighbors */
 699         j_index_start    = jindex[iidx];
 700         j_index_end      = jindex[iidx+1];
 701
 702         /* Get outer coordinate index */
 703         inr              = iinr[iidx];
 704         i_coord_offset   = DIM*inr;
 705
 706         /* Load i particle coords and add shift vector */
 707         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 708                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 709
 710         fix0             = _fjsp_setzero_v2r8();
 711         fiy0             = _fjsp_setzero_v2r8();
 712         fiz0             = _fjsp_setzero_v2r8();
 713         fix1             = _fjsp_setzero_v2r8();
 714         fiy1             = _fjsp_setzero_v2r8();
 715         fiz1             = _fjsp_setzero_v2r8();
 716         fix2             = _fjsp_setzero_v2r8();
 717         fiy2             = _fjsp_setzero_v2r8();
 718         fiz2             = _fjsp_setzero_v2r8();
 719
 720         /* Start inner kernel loop */
 721         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 722         {
 723
 724             /* Get j neighbor index, and coordinate index */
 725             jnrA             = jjnr[jidx];
 726             jnrB             = jjnr[jidx+1];
 727             j_coord_offsetA  = DIM*jnrA;
 728             j_coord_offsetB  = DIM*jnrB;
 729
 730             /* load j atom coordinates */
 731             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 732                                               &jx0,&jy0,&jz0);
 733
 734             /* Calculate displacement vector */
 735             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 736             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 737             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 738             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 739             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 740             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 741             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 742             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 743             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 744
 745             /* Calculate squared distance and things based on it */
 746             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 747             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 748             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 749
 750             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 751             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 752             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 753
 754             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 755             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 756             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 757
 758             /* Load parameters for j particles */
 759             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 760             vdwjidx0A        = 2*vdwtype[jnrA+0];
 761             vdwjidx0B        = 2*vdwtype[jnrB+0];
 762
 763             fjx0             = _fjsp_setzero_v2r8();
 764             fjy0             = _fjsp_setzero_v2r8();
 765             fjz0             = _fjsp_setzero_v2r8();
 766
 767             /**************************
 768              * CALCULATE INTERACTIONS *
 769              **************************/
 770
 771             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 772             {
 773
 774             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 775
 776             /* Compute parameters for interactions between i and j atoms */
 777             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 778             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 779                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 780
 781             /* REACTION-FIELD ELECTROSTATICS */
 782             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 783
 784             /* LENNARD-JONES DISPERSION/REPULSION */
 785
 786             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 787             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 788             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 789             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 790             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 791
 792             d                = _fjsp_sub_v2r8(r00,rswitch);
 793             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 794             d2               = _fjsp_mul_v2r8(d,d);
 795             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 796
 797             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 798
 799             /* Evaluate switch function */
 800             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 801             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 802             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 803
 804             fscal            = _fjsp_add_v2r8(felec,fvdw);
 805
 806             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 807
 808             /* Update vectorial force */
 809             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 810             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 811             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 812
 813             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 814             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 815             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 816
 817             }
 818
 819             /**************************
 820              * CALCULATE INTERACTIONS *
 821              **************************/
 822
 823             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 824             {
 825
 826             /* Compute parameters for interactions between i and j atoms */
 827             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 828
 829             /* REACTION-FIELD ELECTROSTATICS */
 830             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 831
 832             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 833
 834             fscal            = felec;
 835
 836             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 837
 838             /* Update vectorial force */
 839             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 840             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 841             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 842
 843             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 844             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 845             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 846
 847             }
 848
 849             /**************************
 850              * CALCULATE INTERACTIONS *
 851              **************************/
 852
 853             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 854             {
 855
 856             /* Compute parameters for interactions between i and j atoms */
 857             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 858
 859             /* REACTION-FIELD ELECTROSTATICS */
 860             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 861
 862             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 863
 864             fscal            = felec;
 865
 866             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 867
 868             /* Update vectorial force */
 869             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 870             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 871             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 872
 873             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 874             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 875             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 876
 877             }
 878
 879             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 880
 881             /* Inner loop uses 133 flops */
 882         }
 883
 884         if(jidx<j_index_end)
 885         {
 886
 887             jnrA             = jjnr[jidx];
 888             j_coord_offsetA  = DIM*jnrA;
 889
 890             /* load j atom coordinates */
 891             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 892                                               &jx0,&jy0,&jz0);
 893
 894             /* Calculate displacement vector */
 895             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 896             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 897             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 898             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 899             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 900             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 901             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 902             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 903             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 904
 905             /* Calculate squared distance and things based on it */
 906             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 907             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 908             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 909
 910             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 911             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 912             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 913
 914             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 915             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 916             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 917
 918             /* Load parameters for j particles */
 919             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 920             vdwjidx0A        = 2*vdwtype[jnrA+0];
 921
 922             fjx0             = _fjsp_setzero_v2r8();
 923             fjy0             = _fjsp_setzero_v2r8();
 924             fjz0             = _fjsp_setzero_v2r8();
 925
 926             /**************************
 927              * CALCULATE INTERACTIONS *
 928              **************************/
 929
 930             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 931             {
 932
 933             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 934
 935             /* Compute parameters for interactions between i and j atoms */
 936             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 937             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 938
 939             /* REACTION-FIELD ELECTROSTATICS */
 940             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 941
 942             /* LENNARD-JONES DISPERSION/REPULSION */
 943
 944             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 945             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 946             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 947             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 948             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 949
 950             d                = _fjsp_sub_v2r8(r00,rswitch);
 951             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 952             d2               = _fjsp_mul_v2r8(d,d);
 953             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 954
 955             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 956
 957             /* Evaluate switch function */
 958             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 959             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 960             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 961
 962             fscal            = _fjsp_add_v2r8(felec,fvdw);
 963
 964             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 965
 966             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 967
 968             /* Update vectorial force */
 969             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 970             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 971             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 972
 973             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 974             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 975             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 976
 977             }
 978
 979             /**************************
 980              * CALCULATE INTERACTIONS *
 981              **************************/
 982
 983             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 984             {
 985
 986             /* Compute parameters for interactions between i and j atoms */
 987             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 988
 989             /* REACTION-FIELD ELECTROSTATICS */
 990             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 991
 992             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 993
 994             fscal            = felec;
 995
 996             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 997
 998             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 999
1000             /* Update vectorial force */
1001             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1002             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1003             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1004
1005             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1006             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1007             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1008
1009             }
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1016             {
1017
1018             /* Compute parameters for interactions between i and j atoms */
1019             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1020
1021             /* REACTION-FIELD ELECTROSTATICS */
1022             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1023
1024             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1025
1026             fscal            = felec;
1027
1028             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1029
1030             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1031
1032             /* Update vectorial force */
1033             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1034             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1035             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1036
1037             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1038             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1039             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1040
1041             }
1042
1043             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1044
1045             /* Inner loop uses 133 flops */
1046         }
1047
1048         /* End of innermost loop */
1049
1050         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1051                                               f+i_coord_offset,fshift+i_shift_offset);
1052
1053         /* Increment number of inner iterations */
1054         inneriter                  += j_index_end - j_index_start;
1055
1056         /* Outer loop uses 18 flops */
1057     }
1058
1059     /* Increment number of outer iterations */
1060     outeriter        += nri;
1061
1062     /* Update outer/inner flops */
1063
1064     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*133);
1065 }