src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  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       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 102     real             rswitch_scalar,d_scalar;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 123     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 124     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 125     nvdwtype         = fr->ntype;
 126     vdwparam         = fr->nbfp;
 127     vdwtype          = mdatoms->typeA;
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[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     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 137     rcutoff_scalar   = fr->rcoulomb;
 138     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 139     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 140
 141     rswitch_scalar   = fr->rvdw_switch;
 142     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 143     /* Setup switch parameters */
 144     d_scalar         = rcutoff_scalar-rswitch_scalar;
 145     d                = gmx_fjsp_set1_v2r8(d_scalar);
 146     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 147     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 148     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 149     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 150     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 151     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 152
 153     /* Avoid stupid compiler warnings */
 154     jnrA = jnrB = 0;
 155     j_coord_offsetA = 0;
 156     j_coord_offsetB = 0;
 157
 158     outeriter        = 0;
 159     inneriter        = 0;
 160
 161     /* Start outer loop over neighborlists */
 162     for(iidx=0; iidx<nri; iidx++)
 163     {
 164         /* Load shift vector for this list */
 165         i_shift_offset   = DIM*shiftidx[iidx];
 166
 167         /* Load limits for loop over neighbors */
 168         j_index_start    = jindex[iidx];
 169         j_index_end      = jindex[iidx+1];
 170
 171         /* Get outer coordinate index */
 172         inr              = iinr[iidx];
 173         i_coord_offset   = DIM*inr;
 174
 175         /* Load i particle coords and add shift vector */
 176         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 177                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 178
 179         fix0             = _fjsp_setzero_v2r8();
 180         fiy0             = _fjsp_setzero_v2r8();
 181         fiz0             = _fjsp_setzero_v2r8();
 182         fix1             = _fjsp_setzero_v2r8();
 183         fiy1             = _fjsp_setzero_v2r8();
 184         fiz1             = _fjsp_setzero_v2r8();
 185         fix2             = _fjsp_setzero_v2r8();
 186         fiy2             = _fjsp_setzero_v2r8();
 187         fiz2             = _fjsp_setzero_v2r8();
 188         fix3             = _fjsp_setzero_v2r8();
 189         fiy3             = _fjsp_setzero_v2r8();
 190         fiz3             = _fjsp_setzero_v2r8();
 191
 192         /* Reset potential sums */
 193         velecsum         = _fjsp_setzero_v2r8();
 194         vvdwsum          = _fjsp_setzero_v2r8();
 195
 196         /* Start inner kernel loop */
 197         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 198         {
 199
 200             /* Get j neighbor index, and coordinate index */
 201             jnrA             = jjnr[jidx];
 202             jnrB             = jjnr[jidx+1];
 203             j_coord_offsetA  = DIM*jnrA;
 204             j_coord_offsetB  = DIM*jnrB;
 205
 206             /* load j atom coordinates */
 207             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 208                                               &jx0,&jy0,&jz0);
 209
 210             /* Calculate displacement vector */
 211             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 212             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 213             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 214             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 215             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 216             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 217             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 218             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 219             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 220             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 221             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 222             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 223
 224             /* Calculate squared distance and things based on it */
 225             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 226             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 227             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 228             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 229
 230             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 231             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 232             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 233             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 234
 235             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 236             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 237             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 238             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 239
 240             /* Load parameters for j particles */
 241             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 242             vdwjidx0A        = 2*vdwtype[jnrA+0];
 243             vdwjidx0B        = 2*vdwtype[jnrB+0];
 244
 245             fjx0             = _fjsp_setzero_v2r8();
 246             fjy0             = _fjsp_setzero_v2r8();
 247             fjz0             = _fjsp_setzero_v2r8();
 248
 249             /**************************
 250              * CALCULATE INTERACTIONS *
 251              **************************/
 252
 253             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 254             {
 255
 256             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 257
 258             /* Compute parameters for interactions between i and j atoms */
 259             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 260                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 261
 262             /* LENNARD-JONES DISPERSION/REPULSION */
 263
 264             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 265             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 266             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 267             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 268             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 269
 270             d                = _fjsp_sub_v2r8(r00,rswitch);
 271             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 272             d2               = _fjsp_mul_v2r8(d,d);
 273             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 274
 275             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 276
 277             /* Evaluate switch function */
 278             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 279             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 280             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 281             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 282
 283             /* Update potential sum for this i atom from the interaction with this j atom. */
 284             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 285             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 286
 287             fscal            = fvdw;
 288
 289             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 290
 291             /* Update vectorial force */
 292             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 293             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 294             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 295
 296             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 297             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 298             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 299
 300             }
 301
 302             /**************************
 303              * CALCULATE INTERACTIONS *
 304              **************************/
 305
 306             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 307             {
 308
 309             /* Compute parameters for interactions between i and j atoms */
 310             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 311
 312             /* REACTION-FIELD ELECTROSTATICS */
 313             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 314             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 315
 316             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 317
 318             /* Update potential sum for this i atom from the interaction with this j atom. */
 319             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 320             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 321
 322             fscal            = felec;
 323
 324             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 325
 326             /* Update vectorial force */
 327             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 328             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 329             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 330
 331             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 332             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 333             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 334
 335             }
 336
 337             /**************************
 338              * CALCULATE INTERACTIONS *
 339              **************************/
 340
 341             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 342             {
 343
 344             /* Compute parameters for interactions between i and j atoms */
 345             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 346
 347             /* REACTION-FIELD ELECTROSTATICS */
 348             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 349             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 350
 351             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 352
 353             /* Update potential sum for this i atom from the interaction with this j atom. */
 354             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 355             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 356
 357             fscal            = felec;
 358
 359             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 360
 361             /* Update vectorial force */
 362             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 363             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 364             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 365
 366             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 367             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 368             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 369
 370             }
 371
 372             /**************************
 373              * CALCULATE INTERACTIONS *
 374              **************************/
 375
 376             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 377             {
 378
 379             /* Compute parameters for interactions between i and j atoms */
 380             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 381
 382             /* REACTION-FIELD ELECTROSTATICS */
 383             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 384             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 385
 386             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 387
 388             /* Update potential sum for this i atom from the interaction with this j atom. */
 389             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 390             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 391
 392             fscal            = felec;
 393
 394             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 395
 396             /* Update vectorial force */
 397             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 398             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 399             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 400
 401             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 402             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 403             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 404
 405             }
 406
 407             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 408
 409             /* Inner loop uses 182 flops */
 410         }
 411
 412         if(jidx<j_index_end)
 413         {
 414
 415             jnrA             = jjnr[jidx];
 416             j_coord_offsetA  = DIM*jnrA;
 417
 418             /* load j atom coordinates */
 419             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 420                                               &jx0,&jy0,&jz0);
 421
 422             /* Calculate displacement vector */
 423             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 424             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 425             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 426             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 427             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 428             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 429             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 430             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 431             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 432             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 433             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 434             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 435
 436             /* Calculate squared distance and things based on it */
 437             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 438             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 439             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 440             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 441
 442             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 443             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 444             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 445             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 446
 447             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 448             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 449             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 450             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 451
 452             /* Load parameters for j particles */
 453             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 454             vdwjidx0A        = 2*vdwtype[jnrA+0];
 455
 456             fjx0             = _fjsp_setzero_v2r8();
 457             fjy0             = _fjsp_setzero_v2r8();
 458             fjz0             = _fjsp_setzero_v2r8();
 459
 460             /**************************
 461              * CALCULATE INTERACTIONS *
 462              **************************/
 463
 464             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 465             {
 466
 467             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 468
 469             /* Compute parameters for interactions between i and j atoms */
 470             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 471
 472             /* LENNARD-JONES DISPERSION/REPULSION */
 473
 474             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 475             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 476             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 477             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 478             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 479
 480             d                = _fjsp_sub_v2r8(r00,rswitch);
 481             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 482             d2               = _fjsp_mul_v2r8(d,d);
 483             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 484
 485             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 486
 487             /* Evaluate switch function */
 488             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 489             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 490             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 491             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 492
 493             /* Update potential sum for this i atom from the interaction with this j atom. */
 494             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 495             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 496             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 497
 498             fscal            = fvdw;
 499
 500             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 501
 502             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 503
 504             /* Update vectorial force */
 505             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 506             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 507             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 508
 509             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 510             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 511             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 512
 513             }
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 520             {
 521
 522             /* Compute parameters for interactions between i and j atoms */
 523             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 524
 525             /* REACTION-FIELD ELECTROSTATICS */
 526             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 527             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 528
 529             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 530
 531             /* Update potential sum for this i atom from the interaction with this j atom. */
 532             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 533             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 534             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 535
 536             fscal            = felec;
 537
 538             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 539
 540             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 541
 542             /* Update vectorial force */
 543             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 544             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 545             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 546
 547             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 548             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 549             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 550
 551             }
 552
 553             /**************************
 554              * CALCULATE INTERACTIONS *
 555              **************************/
 556
 557             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 558             {
 559
 560             /* Compute parameters for interactions between i and j atoms */
 561             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 562
 563             /* REACTION-FIELD ELECTROSTATICS */
 564             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 565             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 566
 567             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 568
 569             /* Update potential sum for this i atom from the interaction with this j atom. */
 570             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 571             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 572             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 573
 574             fscal            = felec;
 575
 576             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 577
 578             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 579
 580             /* Update vectorial force */
 581             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 582             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 583             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 584
 585             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 586             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 587             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 588
 589             }
 590
 591             /**************************
 592              * CALCULATE INTERACTIONS *
 593              **************************/
 594
 595             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 596             {
 597
 598             /* Compute parameters for interactions between i and j atoms */
 599             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 600
 601             /* REACTION-FIELD ELECTROSTATICS */
 602             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 603             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 604
 605             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 606
 607             /* Update potential sum for this i atom from the interaction with this j atom. */
 608             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 609             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 610             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 611
 612             fscal            = felec;
 613
 614             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 615
 616             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 617
 618             /* Update vectorial force */
 619             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 620             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 621             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 622
 623             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 624             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 625             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 626
 627             }
 628
 629             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 630
 631             /* Inner loop uses 182 flops */
 632         }
 633
 634         /* End of innermost loop */
 635
 636         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 637                                               f+i_coord_offset,fshift+i_shift_offset);
 638
 639         ggid                        = gid[iidx];
 640         /* Update potential energies */
 641         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 642         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 643
 644         /* Increment number of inner iterations */
 645         inneriter                  += j_index_end - j_index_start;
 646
 647         /* Outer loop uses 26 flops */
 648     }
 649
 650     /* Increment number of outer iterations */
 651     outeriter        += nri;
 652
 653     /* Update outer/inner flops */
 654
 655     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*182);
 656 }
 657 /*
 658  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 659  * Electrostatics interaction: ReactionField
 660  * VdW interaction:            LennardJones
 661  * Geometry:                   Water4-Particle
 662  * Calculate force/pot:        Force
 663  */
 664 void
 665 nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 666                     (t_nblist                    * gmx_restrict       nlist,
 667                      rvec                        * gmx_restrict          xx,
 668                      rvec                        * gmx_restrict          ff,
 669                      t_forcerec                  * gmx_restrict          fr,
 670                      t_mdatoms                   * gmx_restrict     mdatoms,
 671                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 672                      t_nrnb                      * gmx_restrict        nrnb)
 673 {
 674     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 675      * just 0 for non-waters.
 676      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 677      * jnr indices corresponding to data put in the four positions in the SIMD register.
 678      */
 679     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 680     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 681     int              jnrA,jnrB;
 682     int              j_coord_offsetA,j_coord_offsetB;
 683     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 684     real             rcutoff_scalar;
 685     real             *shiftvec,*fshift,*x,*f;
 686     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 687     int              vdwioffset0;
 688     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 689     int              vdwioffset1;
 690     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 691     int              vdwioffset2;
 692     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 693     int              vdwioffset3;
 694     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 695     int              vdwjidx0A,vdwjidx0B;
 696     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 697     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 698     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 699     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 700     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 701     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 702     real             *charge;
 703     int              nvdwtype;
 704     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 705     int              *vdwtype;
 706     real             *vdwparam;
 707     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 708     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 709     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 710     real             rswitch_scalar,d_scalar;
 711     _fjsp_v2r8       itab_tmp;
 712     _fjsp_v2r8       dummy_mask,cutoff_mask;
 713     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 714     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 715     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 716
 717     x                = xx[0];
 718     f                = ff[0];
 719
 720     nri              = nlist->nri;
 721     iinr             = nlist->iinr;
 722     jindex           = nlist->jindex;
 723     jjnr             = nlist->jjnr;
 724     shiftidx         = nlist->shift;
 725     gid              = nlist->gid;
 726     shiftvec         = fr->shift_vec[0];
 727     fshift           = fr->fshift[0];
 728     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 729     charge           = mdatoms->chargeA;
 730     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 731     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 732     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 733     nvdwtype         = fr->ntype;
 734     vdwparam         = fr->nbfp;
 735     vdwtype          = mdatoms->typeA;
 736
 737     /* Setup water-specific parameters */
 738     inr              = nlist->iinr[0];
 739     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 740     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 741     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 742     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 743
 744     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 745     rcutoff_scalar   = fr->rcoulomb;
 746     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 747     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 748
 749     rswitch_scalar   = fr->rvdw_switch;
 750     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 751     /* Setup switch parameters */
 752     d_scalar         = rcutoff_scalar-rswitch_scalar;
 753     d                = gmx_fjsp_set1_v2r8(d_scalar);
 754     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 755     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 756     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 757     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 758     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 759     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 760
 761     /* Avoid stupid compiler warnings */
 762     jnrA = jnrB = 0;
 763     j_coord_offsetA = 0;
 764     j_coord_offsetB = 0;
 765
 766     outeriter        = 0;
 767     inneriter        = 0;
 768
 769     /* Start outer loop over neighborlists */
 770     for(iidx=0; iidx<nri; iidx++)
 771     {
 772         /* Load shift vector for this list */
 773         i_shift_offset   = DIM*shiftidx[iidx];
 774
 775         /* Load limits for loop over neighbors */
 776         j_index_start    = jindex[iidx];
 777         j_index_end      = jindex[iidx+1];
 778
 779         /* Get outer coordinate index */
 780         inr              = iinr[iidx];
 781         i_coord_offset   = DIM*inr;
 782
 783         /* Load i particle coords and add shift vector */
 784         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 785                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 786
 787         fix0             = _fjsp_setzero_v2r8();
 788         fiy0             = _fjsp_setzero_v2r8();
 789         fiz0             = _fjsp_setzero_v2r8();
 790         fix1             = _fjsp_setzero_v2r8();
 791         fiy1             = _fjsp_setzero_v2r8();
 792         fiz1             = _fjsp_setzero_v2r8();
 793         fix2             = _fjsp_setzero_v2r8();
 794         fiy2             = _fjsp_setzero_v2r8();
 795         fiz2             = _fjsp_setzero_v2r8();
 796         fix3             = _fjsp_setzero_v2r8();
 797         fiy3             = _fjsp_setzero_v2r8();
 798         fiz3             = _fjsp_setzero_v2r8();
 799
 800         /* Start inner kernel loop */
 801         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 802         {
 803
 804             /* Get j neighbor index, and coordinate index */
 805             jnrA             = jjnr[jidx];
 806             jnrB             = jjnr[jidx+1];
 807             j_coord_offsetA  = DIM*jnrA;
 808             j_coord_offsetB  = DIM*jnrB;
 809
 810             /* load j atom coordinates */
 811             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 812                                               &jx0,&jy0,&jz0);
 813
 814             /* Calculate displacement vector */
 815             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 816             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 817             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 818             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 819             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 820             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 821             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 822             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 823             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 824             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 825             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 826             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 827
 828             /* Calculate squared distance and things based on it */
 829             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 830             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 831             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 832             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 833
 834             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 835             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 836             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 837             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 838
 839             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 840             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 841             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 842             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 843
 844             /* Load parameters for j particles */
 845             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 846             vdwjidx0A        = 2*vdwtype[jnrA+0];
 847             vdwjidx0B        = 2*vdwtype[jnrB+0];
 848
 849             fjx0             = _fjsp_setzero_v2r8();
 850             fjy0             = _fjsp_setzero_v2r8();
 851             fjz0             = _fjsp_setzero_v2r8();
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 858             {
 859
 860             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 861
 862             /* Compute parameters for interactions between i and j atoms */
 863             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 864                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 865
 866             /* LENNARD-JONES DISPERSION/REPULSION */
 867
 868             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 869             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 870             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 871             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 872             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 873
 874             d                = _fjsp_sub_v2r8(r00,rswitch);
 875             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 876             d2               = _fjsp_mul_v2r8(d,d);
 877             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 878
 879             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 880
 881             /* Evaluate switch function */
 882             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 883             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 884             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 885
 886             fscal            = fvdw;
 887
 888             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 889
 890             /* Update vectorial force */
 891             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 892             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 893             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 894
 895             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 896             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 897             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 898
 899             }
 900
 901             /**************************
 902              * CALCULATE INTERACTIONS *
 903              **************************/
 904
 905             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 906             {
 907
 908             /* Compute parameters for interactions between i and j atoms */
 909             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 910
 911             /* REACTION-FIELD ELECTROSTATICS */
 912             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 913
 914             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 915
 916             fscal            = felec;
 917
 918             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 919
 920             /* Update vectorial force */
 921             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 922             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 923             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 924
 925             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 926             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 927             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 928
 929             }
 930
 931             /**************************
 932              * CALCULATE INTERACTIONS *
 933              **************************/
 934
 935             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 936             {
 937
 938             /* Compute parameters for interactions between i and j atoms */
 939             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 940
 941             /* REACTION-FIELD ELECTROSTATICS */
 942             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 943
 944             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 945
 946             fscal            = felec;
 947
 948             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 949
 950             /* Update vectorial force */
 951             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 952             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 953             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 954
 955             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 956             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 957             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 958
 959             }
 960
 961             /**************************
 962              * CALCULATE INTERACTIONS *
 963              **************************/
 964
 965             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 966             {
 967
 968             /* Compute parameters for interactions between i and j atoms */
 969             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 970
 971             /* REACTION-FIELD ELECTROSTATICS */
 972             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 973
 974             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 975
 976             fscal            = felec;
 977
 978             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 979
 980             /* Update vectorial force */
 981             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 982             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 983             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 984
 985             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 986             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 987             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 988
 989             }
 990
 991             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 992
 993             /* Inner loop uses 161 flops */
 994         }
 995
 996         if(jidx<j_index_end)
 997         {
 998
 999             jnrA             = jjnr[jidx];
1000             j_coord_offsetA  = DIM*jnrA;
1001
1002             /* load j atom coordinates */
1003             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1004                                               &jx0,&jy0,&jz0);
1005
1006             /* Calculate displacement vector */
1007             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1008             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1009             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1010             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1011             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1012             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1013             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1014             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1015             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1016             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1017             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1018             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1019
1020             /* Calculate squared distance and things based on it */
1021             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1022             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1023             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1024             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1025
1026             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1027             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1028             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1029             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1030
1031             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1032             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1033             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1034             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1035
1036             /* Load parameters for j particles */
1037             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1038             vdwjidx0A        = 2*vdwtype[jnrA+0];
1039
1040             fjx0             = _fjsp_setzero_v2r8();
1041             fjy0             = _fjsp_setzero_v2r8();
1042             fjz0             = _fjsp_setzero_v2r8();
1043
1044             /**************************
1045              * CALCULATE INTERACTIONS *
1046              **************************/
1047
1048             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1049             {
1050
1051             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1052
1053             /* Compute parameters for interactions between i and j atoms */
1054             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1055
1056             /* LENNARD-JONES DISPERSION/REPULSION */
1057
1058             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1059             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1060             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1061             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1062             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1063
1064             d                = _fjsp_sub_v2r8(r00,rswitch);
1065             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1066             d2               = _fjsp_mul_v2r8(d,d);
1067             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1068
1069             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1070
1071             /* Evaluate switch function */
1072             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1073             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1074             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1075
1076             fscal            = fvdw;
1077
1078             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1079
1080             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1081
1082             /* Update vectorial force */
1083             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1084             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1085             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1086
1087             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1088             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1089             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1090
1091             }
1092
1093             /**************************
1094              * CALCULATE INTERACTIONS *
1095              **************************/
1096
1097             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1098             {
1099
1100             /* Compute parameters for interactions between i and j atoms */
1101             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1102
1103             /* REACTION-FIELD ELECTROSTATICS */
1104             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1105
1106             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1107
1108             fscal            = felec;
1109
1110             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1111
1112             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1113
1114             /* Update vectorial force */
1115             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1116             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1117             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1118
1119             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1120             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1121             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1122
1123             }
1124
1125             /**************************
1126              * CALCULATE INTERACTIONS *
1127              **************************/
1128
1129             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1130             {
1131
1132             /* Compute parameters for interactions between i and j atoms */
1133             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1134
1135             /* REACTION-FIELD ELECTROSTATICS */
1136             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1137
1138             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1139
1140             fscal            = felec;
1141
1142             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1143
1144             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1145
1146             /* Update vectorial force */
1147             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1148             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1149             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1150
1151             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1152             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1153             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1154
1155             }
1156
1157             /**************************
1158              * CALCULATE INTERACTIONS *
1159              **************************/
1160
1161             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1162             {
1163
1164             /* Compute parameters for interactions between i and j atoms */
1165             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1166
1167             /* REACTION-FIELD ELECTROSTATICS */
1168             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1169
1170             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1171
1172             fscal            = felec;
1173
1174             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1175
1176             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1177
1178             /* Update vectorial force */
1179             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1180             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1181             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1182
1183             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1184             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1185             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1186
1187             }
1188
1189             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1190
1191             /* Inner loop uses 161 flops */
1192         }
1193
1194         /* End of innermost loop */
1195
1196         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1197                                               f+i_coord_offset,fshift+i_shift_offset);
1198
1199         /* Increment number of inner iterations */
1200         inneriter                  += j_index_end - j_index_start;
1201
1202         /* Outer loop uses 24 flops */
1203     }
1204
1205     /* Increment number of outer iterations */
1206     outeriter        += nri;
1207
1208     /* Update outer/inner flops */
1209
1210     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*161);
1211 }