src/gromacs/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) 2012,2013,2014, 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 #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 "gromacs/legacyheaders/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_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       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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 428                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 429
 430             /* REACTION-FIELD ELECTROSTATICS */
 431             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 432             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 433
 434             /* LENNARD-JONES DISPERSION/REPULSION */
 435
 436             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 437             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 438             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 439             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 440             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 441
 442             d                = _fjsp_sub_v2r8(r00,rswitch);
 443             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 444             d2               = _fjsp_mul_v2r8(d,d);
 445             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 446
 447             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 448
 449             /* Evaluate switch function */
 450             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 451             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 452             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 453             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 454
 455             /* Update potential sum for this i atom from the interaction with this j atom. */
 456             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 457             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 458             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 459             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 460             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 461             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 462
 463             fscal            = _fjsp_add_v2r8(felec,fvdw);
 464
 465             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 466
 467             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 468
 469             /* Update vectorial force */
 470             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 471             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 472             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 473
 474             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 475             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 476             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 477
 478             }
 479
 480             /**************************
 481              * CALCULATE INTERACTIONS *
 482              **************************/
 483
 484             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 485             {
 486
 487             /* Compute parameters for interactions between i and j atoms */
 488             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 489
 490             /* REACTION-FIELD ELECTROSTATICS */
 491             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 492             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 493
 494             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 498             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 499             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 500
 501             fscal            = felec;
 502
 503             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 504
 505             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 506
 507             /* Update vectorial force */
 508             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 509             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 510             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 511
 512             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 513             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 514             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 515
 516             }
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 523             {
 524
 525             /* Compute parameters for interactions between i and j atoms */
 526             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 527
 528             /* REACTION-FIELD ELECTROSTATICS */
 529             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 530             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 531
 532             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 533
 534             /* Update potential sum for this i atom from the interaction with this j atom. */
 535             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 536             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 537             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 538
 539             fscal            = felec;
 540
 541             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 542
 543             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 544
 545             /* Update vectorial force */
 546             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 547             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 548             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 549
 550             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 551             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 552             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 553
 554             }
 555
 556             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 557
 558             /* Inner loop uses 154 flops */
 559         }
 560
 561         /* End of innermost loop */
 562
 563         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 564                                               f+i_coord_offset,fshift+i_shift_offset);
 565
 566         ggid                        = gid[iidx];
 567         /* Update potential energies */
 568         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 569         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 570
 571         /* Increment number of inner iterations */
 572         inneriter                  += j_index_end - j_index_start;
 573
 574         /* Outer loop uses 20 flops */
 575     }
 576
 577     /* Increment number of outer iterations */
 578     outeriter        += nri;
 579
 580     /* Update outer/inner flops */
 581
 582     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 583 }
 584 /*
 585  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 586  * Electrostatics interaction: ReactionField
 587  * VdW interaction:            LennardJones
 588  * Geometry:                   Water3-Particle
 589  * Calculate force/pot:        Force
 590  */
 591 void
 592 nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 593                     (t_nblist                    * gmx_restrict       nlist,
 594                      rvec                        * gmx_restrict          xx,
 595                      rvec                        * gmx_restrict          ff,
 596                      t_forcerec                  * gmx_restrict          fr,
 597                      t_mdatoms                   * gmx_restrict     mdatoms,
 598                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 599                      t_nrnb                      * gmx_restrict        nrnb)
 600 {
 601     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 602      * just 0 for non-waters.
 603      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 604      * jnr indices corresponding to data put in the four positions in the SIMD register.
 605      */
 606     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 607     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 608     int              jnrA,jnrB;
 609     int              j_coord_offsetA,j_coord_offsetB;
 610     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 611     real             rcutoff_scalar;
 612     real             *shiftvec,*fshift,*x,*f;
 613     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 614     int              vdwioffset0;
 615     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 616     int              vdwioffset1;
 617     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 618     int              vdwioffset2;
 619     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 620     int              vdwjidx0A,vdwjidx0B;
 621     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 622     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 623     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 624     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 625     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 626     real             *charge;
 627     int              nvdwtype;
 628     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 629     int              *vdwtype;
 630     real             *vdwparam;
 631     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 632     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 633     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 634     real             rswitch_scalar,d_scalar;
 635     _fjsp_v2r8       itab_tmp;
 636     _fjsp_v2r8       dummy_mask,cutoff_mask;
 637     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 638     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 639     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 640
 641     x                = xx[0];
 642     f                = ff[0];
 643
 644     nri              = nlist->nri;
 645     iinr             = nlist->iinr;
 646     jindex           = nlist->jindex;
 647     jjnr             = nlist->jjnr;
 648     shiftidx         = nlist->shift;
 649     gid              = nlist->gid;
 650     shiftvec         = fr->shift_vec[0];
 651     fshift           = fr->fshift[0];
 652     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 653     charge           = mdatoms->chargeA;
 654     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 655     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 656     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 657     nvdwtype         = fr->ntype;
 658     vdwparam         = fr->nbfp;
 659     vdwtype          = mdatoms->typeA;
 660
 661     /* Setup water-specific parameters */
 662     inr              = nlist->iinr[0];
 663     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 664     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 665     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 666     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 667
 668     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 669     rcutoff_scalar   = fr->rcoulomb;
 670     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 671     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 672
 673     rswitch_scalar   = fr->rvdw_switch;
 674     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 675     /* Setup switch parameters */
 676     d_scalar         = rcutoff_scalar-rswitch_scalar;
 677     d                = gmx_fjsp_set1_v2r8(d_scalar);
 678     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 679     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 680     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 681     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 682     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 683     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 684
 685     /* Avoid stupid compiler warnings */
 686     jnrA = jnrB = 0;
 687     j_coord_offsetA = 0;
 688     j_coord_offsetB = 0;
 689
 690     outeriter        = 0;
 691     inneriter        = 0;
 692
 693     /* Start outer loop over neighborlists */
 694     for(iidx=0; iidx<nri; iidx++)
 695     {
 696         /* Load shift vector for this list */
 697         i_shift_offset   = DIM*shiftidx[iidx];
 698
 699         /* Load limits for loop over neighbors */
 700         j_index_start    = jindex[iidx];
 701         j_index_end      = jindex[iidx+1];
 702
 703         /* Get outer coordinate index */
 704         inr              = iinr[iidx];
 705         i_coord_offset   = DIM*inr;
 706
 707         /* Load i particle coords and add shift vector */
 708         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 709                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 710
 711         fix0             = _fjsp_setzero_v2r8();
 712         fiy0             = _fjsp_setzero_v2r8();
 713         fiz0             = _fjsp_setzero_v2r8();
 714         fix1             = _fjsp_setzero_v2r8();
 715         fiy1             = _fjsp_setzero_v2r8();
 716         fiz1             = _fjsp_setzero_v2r8();
 717         fix2             = _fjsp_setzero_v2r8();
 718         fiy2             = _fjsp_setzero_v2r8();
 719         fiz2             = _fjsp_setzero_v2r8();
 720
 721         /* Start inner kernel loop */
 722         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 723         {
 724
 725             /* Get j neighbor index, and coordinate index */
 726             jnrA             = jjnr[jidx];
 727             jnrB             = jjnr[jidx+1];
 728             j_coord_offsetA  = DIM*jnrA;
 729             j_coord_offsetB  = DIM*jnrB;
 730
 731             /* load j atom coordinates */
 732             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 733                                               &jx0,&jy0,&jz0);
 734
 735             /* Calculate displacement vector */
 736             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 737             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 738             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 739             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 740             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 741             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 742             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 743             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 744             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 745
 746             /* Calculate squared distance and things based on it */
 747             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 748             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 749             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 750
 751             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 752             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 753             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 754
 755             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 756             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 757             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 758
 759             /* Load parameters for j particles */
 760             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 761             vdwjidx0A        = 2*vdwtype[jnrA+0];
 762             vdwjidx0B        = 2*vdwtype[jnrB+0];
 763
 764             fjx0             = _fjsp_setzero_v2r8();
 765             fjy0             = _fjsp_setzero_v2r8();
 766             fjz0             = _fjsp_setzero_v2r8();
 767
 768             /**************************
 769              * CALCULATE INTERACTIONS *
 770              **************************/
 771
 772             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 773             {
 774
 775             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 776
 777             /* Compute parameters for interactions between i and j atoms */
 778             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 779             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 780                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 781
 782             /* REACTION-FIELD ELECTROSTATICS */
 783             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 784
 785             /* LENNARD-JONES DISPERSION/REPULSION */
 786
 787             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 788             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 789             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 790             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 791             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 792
 793             d                = _fjsp_sub_v2r8(r00,rswitch);
 794             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 795             d2               = _fjsp_mul_v2r8(d,d);
 796             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 797
 798             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 799
 800             /* Evaluate switch function */
 801             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 802             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 803             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 804
 805             fscal            = _fjsp_add_v2r8(felec,fvdw);
 806
 807             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 808
 809             /* Update vectorial force */
 810             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 811             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 812             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 813
 814             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 815             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 816             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 817
 818             }
 819
 820             /**************************
 821              * CALCULATE INTERACTIONS *
 822              **************************/
 823
 824             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 825             {
 826
 827             /* Compute parameters for interactions between i and j atoms */
 828             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 829
 830             /* REACTION-FIELD ELECTROSTATICS */
 831             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 832
 833             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 834
 835             fscal            = felec;
 836
 837             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 838
 839             /* Update vectorial force */
 840             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 841             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 842             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 843
 844             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 845             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 846             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 847
 848             }
 849
 850             /**************************
 851              * CALCULATE INTERACTIONS *
 852              **************************/
 853
 854             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 855             {
 856
 857             /* Compute parameters for interactions between i and j atoms */
 858             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 859
 860             /* REACTION-FIELD ELECTROSTATICS */
 861             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 862
 863             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 864
 865             fscal            = felec;
 866
 867             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 868
 869             /* Update vectorial force */
 870             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 871             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 872             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 873
 874             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 875             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 876             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 877
 878             }
 879
 880             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 881
 882             /* Inner loop uses 133 flops */
 883         }
 884
 885         if(jidx<j_index_end)
 886         {
 887
 888             jnrA             = jjnr[jidx];
 889             j_coord_offsetA  = DIM*jnrA;
 890
 891             /* load j atom coordinates */
 892             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 893                                               &jx0,&jy0,&jz0);
 894
 895             /* Calculate displacement vector */
 896             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 897             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 898             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 899             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 900             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 901             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 902             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 903             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 904             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 905
 906             /* Calculate squared distance and things based on it */
 907             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 908             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 909             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 910
 911             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 912             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 913             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 914
 915             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 916             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 917             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 918
 919             /* Load parameters for j particles */
 920             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 921             vdwjidx0A        = 2*vdwtype[jnrA+0];
 922
 923             fjx0             = _fjsp_setzero_v2r8();
 924             fjy0             = _fjsp_setzero_v2r8();
 925             fjz0             = _fjsp_setzero_v2r8();
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 932             {
 933
 934             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 935
 936             /* Compute parameters for interactions between i and j atoms */
 937             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 938             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 939                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 940
 941             /* REACTION-FIELD ELECTROSTATICS */
 942             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 943
 944             /* LENNARD-JONES DISPERSION/REPULSION */
 945
 946             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 947             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 948             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 949             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 950             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 951
 952             d                = _fjsp_sub_v2r8(r00,rswitch);
 953             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 954             d2               = _fjsp_mul_v2r8(d,d);
 955             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 956
 957             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 958
 959             /* Evaluate switch function */
 960             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 961             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 962             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 963
 964             fscal            = _fjsp_add_v2r8(felec,fvdw);
 965
 966             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 967
 968             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 969
 970             /* Update vectorial force */
 971             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 972             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 973             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 974
 975             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 976             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 977             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 978
 979             }
 980
 981             /**************************
 982              * CALCULATE INTERACTIONS *
 983              **************************/
 984
 985             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 986             {
 987
 988             /* Compute parameters for interactions between i and j atoms */
 989             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 990
 991             /* REACTION-FIELD ELECTROSTATICS */
 992             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 993
 994             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 995
 996             fscal            = felec;
 997
 998             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 999
1000             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1001
1002             /* Update vectorial force */
1003             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1004             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1005             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1006
1007             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1008             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1009             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1010
1011             }
1012
1013             /**************************
1014              * CALCULATE INTERACTIONS *
1015              **************************/
1016
1017             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1018             {
1019
1020             /* Compute parameters for interactions between i and j atoms */
1021             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1022
1023             /* REACTION-FIELD ELECTROSTATICS */
1024             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1025
1026             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1027
1028             fscal            = felec;
1029
1030             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1031
1032             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1033
1034             /* Update vectorial force */
1035             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1036             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1037             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1038
1039             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1040             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1041             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1042
1043             }
1044
1045             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1046
1047             /* Inner loop uses 133 flops */
1048         }
1049
1050         /* End of innermost loop */
1051
1052         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1053                                               f+i_coord_offset,fshift+i_shift_offset);
1054
1055         /* Increment number of inner iterations */
1056         inneriter                  += j_index_end - j_index_start;
1057
1058         /* Outer loop uses 18 flops */
1059     }
1060
1061     /* Increment number of outer iterations */
1062     outeriter        += nri;
1063
1064     /* Update outer/inner flops */
1065
1066     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*133);
1067 }