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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/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                      struct 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->ic->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->ic->rcoulomb;
 138     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 139     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 140
 141     rswitch_scalar   = fr->ic->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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 471                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 472
 473             /* LENNARD-JONES DISPERSION/REPULSION */
 474
 475             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 476             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 477             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 478             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 479             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 480
 481             d                = _fjsp_sub_v2r8(r00,rswitch);
 482             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 483             d2               = _fjsp_mul_v2r8(d,d);
 484             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 485
 486             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 487
 488             /* Evaluate switch function */
 489             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 490             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 491             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 492             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 496             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 497             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 498
 499             fscal            = fvdw;
 500
 501             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 502
 503             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 504
 505             /* Update vectorial force */
 506             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 507             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 508             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 509
 510             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 511             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 512             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 513
 514             }
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 521             {
 522
 523             /* Compute parameters for interactions between i and j atoms */
 524             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 525
 526             /* REACTION-FIELD ELECTROSTATICS */
 527             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 528             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 529
 530             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 531
 532             /* Update potential sum for this i atom from the interaction with this j atom. */
 533             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 534             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 535             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 536
 537             fscal            = felec;
 538
 539             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 540
 541             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 542
 543             /* Update vectorial force */
 544             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 545             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 546             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 547
 548             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 549             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 550             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 551
 552             }
 553
 554             /**************************
 555              * CALCULATE INTERACTIONS *
 556              **************************/
 557
 558             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 559             {
 560
 561             /* Compute parameters for interactions between i and j atoms */
 562             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 563
 564             /* REACTION-FIELD ELECTROSTATICS */
 565             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 566             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 567
 568             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 569
 570             /* Update potential sum for this i atom from the interaction with this j atom. */
 571             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 572             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 573             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 574
 575             fscal            = felec;
 576
 577             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 578
 579             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 580
 581             /* Update vectorial force */
 582             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 583             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 584             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 585
 586             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 587             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 588             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 589
 590             }
 591
 592             /**************************
 593              * CALCULATE INTERACTIONS *
 594              **************************/
 595
 596             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 597             {
 598
 599             /* Compute parameters for interactions between i and j atoms */
 600             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 601
 602             /* REACTION-FIELD ELECTROSTATICS */
 603             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 604             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 605
 606             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 607
 608             /* Update potential sum for this i atom from the interaction with this j atom. */
 609             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 610             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 611             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 612
 613             fscal            = felec;
 614
 615             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 616
 617             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 618
 619             /* Update vectorial force */
 620             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 621             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 622             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 623
 624             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 625             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 626             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 627
 628             }
 629
 630             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 631
 632             /* Inner loop uses 182 flops */
 633         }
 634
 635         /* End of innermost loop */
 636
 637         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 638                                               f+i_coord_offset,fshift+i_shift_offset);
 639
 640         ggid                        = gid[iidx];
 641         /* Update potential energies */
 642         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 643         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 644
 645         /* Increment number of inner iterations */
 646         inneriter                  += j_index_end - j_index_start;
 647
 648         /* Outer loop uses 26 flops */
 649     }
 650
 651     /* Increment number of outer iterations */
 652     outeriter        += nri;
 653
 654     /* Update outer/inner flops */
 655
 656     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*182);
 657 }
 658 /*
 659  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 660  * Electrostatics interaction: ReactionField
 661  * VdW interaction:            LennardJones
 662  * Geometry:                   Water4-Particle
 663  * Calculate force/pot:        Force
 664  */
 665 void
 666 nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 667                     (t_nblist                    * gmx_restrict       nlist,
 668                      rvec                        * gmx_restrict          xx,
 669                      rvec                        * gmx_restrict          ff,
 670                      struct t_forcerec           * gmx_restrict          fr,
 671                      t_mdatoms                   * gmx_restrict     mdatoms,
 672                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 673                      t_nrnb                      * gmx_restrict        nrnb)
 674 {
 675     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 676      * just 0 for non-waters.
 677      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 678      * jnr indices corresponding to data put in the four positions in the SIMD register.
 679      */
 680     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 681     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 682     int              jnrA,jnrB;
 683     int              j_coord_offsetA,j_coord_offsetB;
 684     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 685     real             rcutoff_scalar;
 686     real             *shiftvec,*fshift,*x,*f;
 687     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 688     int              vdwioffset0;
 689     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 690     int              vdwioffset1;
 691     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 692     int              vdwioffset2;
 693     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 694     int              vdwioffset3;
 695     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 696     int              vdwjidx0A,vdwjidx0B;
 697     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 698     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 699     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 700     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 701     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 702     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 703     real             *charge;
 704     int              nvdwtype;
 705     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 706     int              *vdwtype;
 707     real             *vdwparam;
 708     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 709     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 710     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 711     real             rswitch_scalar,d_scalar;
 712     _fjsp_v2r8       itab_tmp;
 713     _fjsp_v2r8       dummy_mask,cutoff_mask;
 714     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 715     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 716     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 717
 718     x                = xx[0];
 719     f                = ff[0];
 720
 721     nri              = nlist->nri;
 722     iinr             = nlist->iinr;
 723     jindex           = nlist->jindex;
 724     jjnr             = nlist->jjnr;
 725     shiftidx         = nlist->shift;
 726     gid              = nlist->gid;
 727     shiftvec         = fr->shift_vec[0];
 728     fshift           = fr->fshift[0];
 729     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 730     charge           = mdatoms->chargeA;
 731     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 732     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 733     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 734     nvdwtype         = fr->ntype;
 735     vdwparam         = fr->nbfp;
 736     vdwtype          = mdatoms->typeA;
 737
 738     /* Setup water-specific parameters */
 739     inr              = nlist->iinr[0];
 740     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 741     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 742     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 743     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 744
 745     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 746     rcutoff_scalar   = fr->ic->rcoulomb;
 747     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 748     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 749
 750     rswitch_scalar   = fr->ic->rvdw_switch;
 751     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 752     /* Setup switch parameters */
 753     d_scalar         = rcutoff_scalar-rswitch_scalar;
 754     d                = gmx_fjsp_set1_v2r8(d_scalar);
 755     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 756     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 757     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 758     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 759     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 760     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 761
 762     /* Avoid stupid compiler warnings */
 763     jnrA = jnrB = 0;
 764     j_coord_offsetA = 0;
 765     j_coord_offsetB = 0;
 766
 767     outeriter        = 0;
 768     inneriter        = 0;
 769
 770     /* Start outer loop over neighborlists */
 771     for(iidx=0; iidx<nri; iidx++)
 772     {
 773         /* Load shift vector for this list */
 774         i_shift_offset   = DIM*shiftidx[iidx];
 775
 776         /* Load limits for loop over neighbors */
 777         j_index_start    = jindex[iidx];
 778         j_index_end      = jindex[iidx+1];
 779
 780         /* Get outer coordinate index */
 781         inr              = iinr[iidx];
 782         i_coord_offset   = DIM*inr;
 783
 784         /* Load i particle coords and add shift vector */
 785         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 786                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 787
 788         fix0             = _fjsp_setzero_v2r8();
 789         fiy0             = _fjsp_setzero_v2r8();
 790         fiz0             = _fjsp_setzero_v2r8();
 791         fix1             = _fjsp_setzero_v2r8();
 792         fiy1             = _fjsp_setzero_v2r8();
 793         fiz1             = _fjsp_setzero_v2r8();
 794         fix2             = _fjsp_setzero_v2r8();
 795         fiy2             = _fjsp_setzero_v2r8();
 796         fiz2             = _fjsp_setzero_v2r8();
 797         fix3             = _fjsp_setzero_v2r8();
 798         fiy3             = _fjsp_setzero_v2r8();
 799         fiz3             = _fjsp_setzero_v2r8();
 800
 801         /* Start inner kernel loop */
 802         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 803         {
 804
 805             /* Get j neighbor index, and coordinate index */
 806             jnrA             = jjnr[jidx];
 807             jnrB             = jjnr[jidx+1];
 808             j_coord_offsetA  = DIM*jnrA;
 809             j_coord_offsetB  = DIM*jnrB;
 810
 811             /* load j atom coordinates */
 812             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 813                                               &jx0,&jy0,&jz0);
 814
 815             /* Calculate displacement vector */
 816             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 817             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 818             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 819             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 820             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 821             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 822             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 823             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 824             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 825             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 826             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 827             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 828
 829             /* Calculate squared distance and things based on it */
 830             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 831             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 832             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 833             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 834
 835             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 836             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 837             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 838             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 839
 840             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 841             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 842             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 843             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 844
 845             /* Load parameters for j particles */
 846             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 847             vdwjidx0A        = 2*vdwtype[jnrA+0];
 848             vdwjidx0B        = 2*vdwtype[jnrB+0];
 849
 850             fjx0             = _fjsp_setzero_v2r8();
 851             fjy0             = _fjsp_setzero_v2r8();
 852             fjz0             = _fjsp_setzero_v2r8();
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 859             {
 860
 861             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 862
 863             /* Compute parameters for interactions between i and j atoms */
 864             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 865                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 866
 867             /* LENNARD-JONES DISPERSION/REPULSION */
 868
 869             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 870             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 871             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 872             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 873             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 874
 875             d                = _fjsp_sub_v2r8(r00,rswitch);
 876             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 877             d2               = _fjsp_mul_v2r8(d,d);
 878             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 879
 880             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 881
 882             /* Evaluate switch function */
 883             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 884             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 885             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 886
 887             fscal            = fvdw;
 888
 889             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 890
 891             /* Update vectorial force */
 892             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 893             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 894             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 895
 896             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 897             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 898             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 899
 900             }
 901
 902             /**************************
 903              * CALCULATE INTERACTIONS *
 904              **************************/
 905
 906             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 907             {
 908
 909             /* Compute parameters for interactions between i and j atoms */
 910             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 911
 912             /* REACTION-FIELD ELECTROSTATICS */
 913             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 914
 915             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 916
 917             fscal            = felec;
 918
 919             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 920
 921             /* Update vectorial force */
 922             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 923             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 924             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 925
 926             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 927             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 928             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 929
 930             }
 931
 932             /**************************
 933              * CALCULATE INTERACTIONS *
 934              **************************/
 935
 936             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 937             {
 938
 939             /* Compute parameters for interactions between i and j atoms */
 940             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 941
 942             /* REACTION-FIELD ELECTROSTATICS */
 943             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 944
 945             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 946
 947             fscal            = felec;
 948
 949             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 950
 951             /* Update vectorial force */
 952             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 953             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 954             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 955
 956             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 957             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 958             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 959
 960             }
 961
 962             /**************************
 963              * CALCULATE INTERACTIONS *
 964              **************************/
 965
 966             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 967             {
 968
 969             /* Compute parameters for interactions between i and j atoms */
 970             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 971
 972             /* REACTION-FIELD ELECTROSTATICS */
 973             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 974
 975             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 976
 977             fscal            = felec;
 978
 979             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 980
 981             /* Update vectorial force */
 982             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 983             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 984             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 985
 986             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 987             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 988             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 989
 990             }
 991
 992             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 993
 994             /* Inner loop uses 161 flops */
 995         }
 996
 997         if(jidx<j_index_end)
 998         {
 999
1000             jnrA             = jjnr[jidx];
1001             j_coord_offsetA  = DIM*jnrA;
1002
1003             /* load j atom coordinates */
1004             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1005                                               &jx0,&jy0,&jz0);
1006
1007             /* Calculate displacement vector */
1008             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1009             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1010             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1011             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1012             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1013             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1014             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1015             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1016             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1017             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1018             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1019             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1020
1021             /* Calculate squared distance and things based on it */
1022             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1023             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1024             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1025             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1026
1027             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1028             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1029             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1030             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1031
1032             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1033             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1034             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1035             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1036
1037             /* Load parameters for j particles */
1038             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1039             vdwjidx0A        = 2*vdwtype[jnrA+0];
1040
1041             fjx0             = _fjsp_setzero_v2r8();
1042             fjy0             = _fjsp_setzero_v2r8();
1043             fjz0             = _fjsp_setzero_v2r8();
1044
1045             /**************************
1046              * CALCULATE INTERACTIONS *
1047              **************************/
1048
1049             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1050             {
1051
1052             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1053
1054             /* Compute parameters for interactions between i and j atoms */
1055             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1056                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1057
1058             /* LENNARD-JONES DISPERSION/REPULSION */
1059
1060             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1061             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1062             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1063             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1064             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1065
1066             d                = _fjsp_sub_v2r8(r00,rswitch);
1067             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1068             d2               = _fjsp_mul_v2r8(d,d);
1069             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1070
1071             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1072
1073             /* Evaluate switch function */
1074             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1075             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1076             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1077
1078             fscal            = fvdw;
1079
1080             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1081
1082             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1083
1084             /* Update vectorial force */
1085             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1086             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1087             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1088
1089             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1090             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1091             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1092
1093             }
1094
1095             /**************************
1096              * CALCULATE INTERACTIONS *
1097              **************************/
1098
1099             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1100             {
1101
1102             /* Compute parameters for interactions between i and j atoms */
1103             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1104
1105             /* REACTION-FIELD ELECTROSTATICS */
1106             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1107
1108             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1109
1110             fscal            = felec;
1111
1112             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1113
1114             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1115
1116             /* Update vectorial force */
1117             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1118             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1119             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1120
1121             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1122             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1123             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1124
1125             }
1126
1127             /**************************
1128              * CALCULATE INTERACTIONS *
1129              **************************/
1130
1131             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1132             {
1133
1134             /* Compute parameters for interactions between i and j atoms */
1135             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1136
1137             /* REACTION-FIELD ELECTROSTATICS */
1138             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1139
1140             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1141
1142             fscal            = felec;
1143
1144             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1145
1146             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1147
1148             /* Update vectorial force */
1149             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1150             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1151             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1152
1153             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1154             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1155             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1156
1157             }
1158
1159             /**************************
1160              * CALCULATE INTERACTIONS *
1161              **************************/
1162
1163             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1164             {
1165
1166             /* Compute parameters for interactions between i and j atoms */
1167             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1168
1169             /* REACTION-FIELD ELECTROSTATICS */
1170             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1171
1172             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1173
1174             fscal            = felec;
1175
1176             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1177
1178             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1179
1180             /* Update vectorial force */
1181             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1182             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1183             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1184
1185             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1186             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1187             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1188
1189             }
1190
1191             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1192
1193             /* Inner loop uses 161 flops */
1194         }
1195
1196         /* End of innermost loop */
1197
1198         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1199                                               f+i_coord_offset,fshift+i_shift_offset);
1200
1201         /* Increment number of inner iterations */
1202         inneriter                  += j_index_end - j_index_start;
1203
1204         /* Outer loop uses 24 flops */
1205     }
1206
1207     /* Increment number of outer iterations */
1208     outeriter        += nri;
1209
1210     /* Update outer/inner flops */
1211
1212     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*161);
1213 }