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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 104     real             rswitch_scalar,d_scalar;
 105     _fjsp_v2r8       itab_tmp;
 106     _fjsp_v2r8       dummy_mask,cutoff_mask;
 107     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 108     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 109     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 110
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 125     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 126     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 127     nvdwtype         = fr->ntype;
 128     vdwparam         = fr->nbfp;
 129     vdwtype          = mdatoms->typeA;
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 139     rcutoff_scalar   = fr->rcoulomb;
 140     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 141     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 142
 143     rswitch_scalar   = fr->rvdw_switch;
 144     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 145     /* Setup switch parameters */
 146     d_scalar         = rcutoff_scalar-rswitch_scalar;
 147     d                = gmx_fjsp_set1_v2r8(d_scalar);
 148     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 149     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 150     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 151     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 152     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 153     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 154
 155     /* Avoid stupid compiler warnings */
 156     jnrA = jnrB = 0;
 157     j_coord_offsetA = 0;
 158     j_coord_offsetB = 0;
 159
 160     outeriter        = 0;
 161     inneriter        = 0;
 162
 163     /* Start outer loop over neighborlists */
 164     for(iidx=0; iidx<nri; iidx++)
 165     {
 166         /* Load shift vector for this list */
 167         i_shift_offset   = DIM*shiftidx[iidx];
 168
 169         /* Load limits for loop over neighbors */
 170         j_index_start    = jindex[iidx];
 171         j_index_end      = jindex[iidx+1];
 172
 173         /* Get outer coordinate index */
 174         inr              = iinr[iidx];
 175         i_coord_offset   = DIM*inr;
 176
 177         /* Load i particle coords and add shift vector */
 178         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 179                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 180
 181         fix0             = _fjsp_setzero_v2r8();
 182         fiy0             = _fjsp_setzero_v2r8();
 183         fiz0             = _fjsp_setzero_v2r8();
 184         fix1             = _fjsp_setzero_v2r8();
 185         fiy1             = _fjsp_setzero_v2r8();
 186         fiz1             = _fjsp_setzero_v2r8();
 187         fix2             = _fjsp_setzero_v2r8();
 188         fiy2             = _fjsp_setzero_v2r8();
 189         fiz2             = _fjsp_setzero_v2r8();
 190         fix3             = _fjsp_setzero_v2r8();
 191         fiy3             = _fjsp_setzero_v2r8();
 192         fiz3             = _fjsp_setzero_v2r8();
 193
 194         /* Reset potential sums */
 195         velecsum         = _fjsp_setzero_v2r8();
 196         vvdwsum          = _fjsp_setzero_v2r8();
 197
 198         /* Start inner kernel loop */
 199         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 200         {
 201
 202             /* Get j neighbor index, and coordinate index */
 203             jnrA             = jjnr[jidx];
 204             jnrB             = jjnr[jidx+1];
 205             j_coord_offsetA  = DIM*jnrA;
 206             j_coord_offsetB  = DIM*jnrB;
 207
 208             /* load j atom coordinates */
 209             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 210                                               &jx0,&jy0,&jz0);
 211
 212             /* Calculate displacement vector */
 213             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 214             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 215             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 216             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 217             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 218             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 219             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 220             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 221             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 222             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 223             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 224             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 225
 226             /* Calculate squared distance and things based on it */
 227             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 228             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 229             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 230             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 231
 232             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 233             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 234             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 235             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 236
 237             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 238             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 239             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 240             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 241
 242             /* Load parameters for j particles */
 243             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 244             vdwjidx0A        = 2*vdwtype[jnrA+0];
 245             vdwjidx0B        = 2*vdwtype[jnrB+0];
 246
 247             fjx0             = _fjsp_setzero_v2r8();
 248             fjy0             = _fjsp_setzero_v2r8();
 249             fjz0             = _fjsp_setzero_v2r8();
 250
 251             /**************************
 252              * CALCULATE INTERACTIONS *
 253              **************************/
 254
 255             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 256             {
 257
 258             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 259
 260             /* Compute parameters for interactions between i and j atoms */
 261             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 262                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 263
 264             /* LENNARD-JONES DISPERSION/REPULSION */
 265
 266             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 267             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 268             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 269             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 270             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 271
 272             d                = _fjsp_sub_v2r8(r00,rswitch);
 273             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 274             d2               = _fjsp_mul_v2r8(d,d);
 275             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 276
 277             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 278
 279             /* Evaluate switch function */
 280             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 281             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 282             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 283             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 284
 285             /* Update potential sum for this i atom from the interaction with this j atom. */
 286             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 287             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 288
 289             fscal            = fvdw;
 290
 291             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 292
 293             /* Update vectorial force */
 294             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 295             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 296             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 297
 298             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 299             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 300             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 301
 302             }
 303
 304             /**************************
 305              * CALCULATE INTERACTIONS *
 306              **************************/
 307
 308             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 309             {
 310
 311             /* Compute parameters for interactions between i and j atoms */
 312             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 313
 314             /* REACTION-FIELD ELECTROSTATICS */
 315             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 316             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 317
 318             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 319
 320             /* Update potential sum for this i atom from the interaction with this j atom. */
 321             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 322             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 323
 324             fscal            = felec;
 325
 326             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 327
 328             /* Update vectorial force */
 329             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 330             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 331             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 332
 333             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 334             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 335             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 336
 337             }
 338
 339             /**************************
 340              * CALCULATE INTERACTIONS *
 341              **************************/
 342
 343             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 344             {
 345
 346             /* Compute parameters for interactions between i and j atoms */
 347             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 348
 349             /* REACTION-FIELD ELECTROSTATICS */
 350             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 351             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 352
 353             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 357             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 358
 359             fscal            = felec;
 360
 361             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 362
 363             /* Update vectorial force */
 364             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 365             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 366             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 367
 368             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 369             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 370             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 371
 372             }
 373
 374             /**************************
 375              * CALCULATE INTERACTIONS *
 376              **************************/
 377
 378             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 379             {
 380
 381             /* Compute parameters for interactions between i and j atoms */
 382             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 383
 384             /* REACTION-FIELD ELECTROSTATICS */
 385             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 386             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 387
 388             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 392             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 393
 394             fscal            = felec;
 395
 396             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 397
 398             /* Update vectorial force */
 399             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 400             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 401             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 402
 403             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 404             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 405             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 406
 407             }
 408
 409             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 410
 411             /* Inner loop uses 182 flops */
 412         }
 413
 414         if(jidx<j_index_end)
 415         {
 416
 417             jnrA             = jjnr[jidx];
 418             j_coord_offsetA  = DIM*jnrA;
 419
 420             /* load j atom coordinates */
 421             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 422                                               &jx0,&jy0,&jz0);
 423
 424             /* Calculate displacement vector */
 425             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 426             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 427             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 428             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 429             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 430             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 431             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 432             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 433             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 434             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 435             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 436             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 437
 438             /* Calculate squared distance and things based on it */
 439             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 440             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 441             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 442             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 443
 444             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 445             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 446             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 447             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 448
 449             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 450             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 451             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 452             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 453
 454             /* Load parameters for j particles */
 455             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 456             vdwjidx0A        = 2*vdwtype[jnrA+0];
 457
 458             fjx0             = _fjsp_setzero_v2r8();
 459             fjy0             = _fjsp_setzero_v2r8();
 460             fjz0             = _fjsp_setzero_v2r8();
 461
 462             /**************************
 463              * CALCULATE INTERACTIONS *
 464              **************************/
 465
 466             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 467             {
 468
 469             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 470
 471             /* Compute parameters for interactions between i and j atoms */
 472             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 473                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 474
 475             /* LENNARD-JONES DISPERSION/REPULSION */
 476
 477             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 478             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 479             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 480             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 481             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 482
 483             d                = _fjsp_sub_v2r8(r00,rswitch);
 484             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 485             d2               = _fjsp_mul_v2r8(d,d);
 486             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 487
 488             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 489
 490             /* Evaluate switch function */
 491             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 492             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 493             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 494             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 498             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 499             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 500
 501             fscal            = fvdw;
 502
 503             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 504
 505             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 506
 507             /* Update vectorial force */
 508             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 509             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 510             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 511
 512             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 513             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 514             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 515
 516             }
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 523             {
 524
 525             /* Compute parameters for interactions between i and j atoms */
 526             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 527
 528             /* REACTION-FIELD ELECTROSTATICS */
 529             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 530             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 531
 532             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 533
 534             /* Update potential sum for this i atom from the interaction with this j atom. */
 535             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 536             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 537             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 538
 539             fscal            = felec;
 540
 541             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 542
 543             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 544
 545             /* Update vectorial force */
 546             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 547             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 548             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 549
 550             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 551             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 552             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 553
 554             }
 555
 556             /**************************
 557              * CALCULATE INTERACTIONS *
 558              **************************/
 559
 560             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 561             {
 562
 563             /* Compute parameters for interactions between i and j atoms */
 564             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 565
 566             /* REACTION-FIELD ELECTROSTATICS */
 567             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 568             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 569
 570             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 571
 572             /* Update potential sum for this i atom from the interaction with this j atom. */
 573             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 574             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 575             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 576
 577             fscal            = felec;
 578
 579             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 580
 581             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 582
 583             /* Update vectorial force */
 584             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 585             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 586             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 587
 588             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 589             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 590             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 591
 592             }
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 599             {
 600
 601             /* Compute parameters for interactions between i and j atoms */
 602             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 603
 604             /* REACTION-FIELD ELECTROSTATICS */
 605             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 606             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 607
 608             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 609
 610             /* Update potential sum for this i atom from the interaction with this j atom. */
 611             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 612             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 613             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 614
 615             fscal            = felec;
 616
 617             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 618
 619             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 620
 621             /* Update vectorial force */
 622             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 623             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 624             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 625
 626             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 627             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 628             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 629
 630             }
 631
 632             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 633
 634             /* Inner loop uses 182 flops */
 635         }
 636
 637         /* End of innermost loop */
 638
 639         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 640                                               f+i_coord_offset,fshift+i_shift_offset);
 641
 642         ggid                        = gid[iidx];
 643         /* Update potential energies */
 644         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 645         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 646
 647         /* Increment number of inner iterations */
 648         inneriter                  += j_index_end - j_index_start;
 649
 650         /* Outer loop uses 26 flops */
 651     }
 652
 653     /* Increment number of outer iterations */
 654     outeriter        += nri;
 655
 656     /* Update outer/inner flops */
 657
 658     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*182);
 659 }
 660 /*
 661  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 662  * Electrostatics interaction: ReactionField
 663  * VdW interaction:            LennardJones
 664  * Geometry:                   Water4-Particle
 665  * Calculate force/pot:        Force
 666  */
 667 void
 668 nb_kernel_ElecRFCut_VdwLJSw_GeomW4P1_F_sparc64_hpc_ace_double
 669                     (t_nblist                    * gmx_restrict       nlist,
 670                      rvec                        * gmx_restrict          xx,
 671                      rvec                        * gmx_restrict          ff,
 672                      t_forcerec                  * gmx_restrict          fr,
 673                      t_mdatoms                   * gmx_restrict     mdatoms,
 674                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 675                      t_nrnb                      * gmx_restrict        nrnb)
 676 {
 677     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 678      * just 0 for non-waters.
 679      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 680      * jnr indices corresponding to data put in the four positions in the SIMD register.
 681      */
 682     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 683     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 684     int              jnrA,jnrB;
 685     int              j_coord_offsetA,j_coord_offsetB;
 686     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 687     real             rcutoff_scalar;
 688     real             *shiftvec,*fshift,*x,*f;
 689     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 690     int              vdwioffset0;
 691     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 692     int              vdwioffset1;
 693     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 694     int              vdwioffset2;
 695     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 696     int              vdwioffset3;
 697     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 698     int              vdwjidx0A,vdwjidx0B;
 699     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 700     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 701     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 702     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 703     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 704     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 705     real             *charge;
 706     int              nvdwtype;
 707     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 708     int              *vdwtype;
 709     real             *vdwparam;
 710     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 711     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 712     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 713     real             rswitch_scalar,d_scalar;
 714     _fjsp_v2r8       itab_tmp;
 715     _fjsp_v2r8       dummy_mask,cutoff_mask;
 716     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 717     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 718     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 719
 720     x                = xx[0];
 721     f                = ff[0];
 722
 723     nri              = nlist->nri;
 724     iinr             = nlist->iinr;
 725     jindex           = nlist->jindex;
 726     jjnr             = nlist->jjnr;
 727     shiftidx         = nlist->shift;
 728     gid              = nlist->gid;
 729     shiftvec         = fr->shift_vec[0];
 730     fshift           = fr->fshift[0];
 731     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 732     charge           = mdatoms->chargeA;
 733     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 734     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 735     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 736     nvdwtype         = fr->ntype;
 737     vdwparam         = fr->nbfp;
 738     vdwtype          = mdatoms->typeA;
 739
 740     /* Setup water-specific parameters */
 741     inr              = nlist->iinr[0];
 742     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 743     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 744     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 745     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 746
 747     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 748     rcutoff_scalar   = fr->rcoulomb;
 749     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 750     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 751
 752     rswitch_scalar   = fr->rvdw_switch;
 753     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 754     /* Setup switch parameters */
 755     d_scalar         = rcutoff_scalar-rswitch_scalar;
 756     d                = gmx_fjsp_set1_v2r8(d_scalar);
 757     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 758     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 759     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 760     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 761     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 762     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 763
 764     /* Avoid stupid compiler warnings */
 765     jnrA = jnrB = 0;
 766     j_coord_offsetA = 0;
 767     j_coord_offsetB = 0;
 768
 769     outeriter        = 0;
 770     inneriter        = 0;
 771
 772     /* Start outer loop over neighborlists */
 773     for(iidx=0; iidx<nri; iidx++)
 774     {
 775         /* Load shift vector for this list */
 776         i_shift_offset   = DIM*shiftidx[iidx];
 777
 778         /* Load limits for loop over neighbors */
 779         j_index_start    = jindex[iidx];
 780         j_index_end      = jindex[iidx+1];
 781
 782         /* Get outer coordinate index */
 783         inr              = iinr[iidx];
 784         i_coord_offset   = DIM*inr;
 785
 786         /* Load i particle coords and add shift vector */
 787         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 788                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 789
 790         fix0             = _fjsp_setzero_v2r8();
 791         fiy0             = _fjsp_setzero_v2r8();
 792         fiz0             = _fjsp_setzero_v2r8();
 793         fix1             = _fjsp_setzero_v2r8();
 794         fiy1             = _fjsp_setzero_v2r8();
 795         fiz1             = _fjsp_setzero_v2r8();
 796         fix2             = _fjsp_setzero_v2r8();
 797         fiy2             = _fjsp_setzero_v2r8();
 798         fiz2             = _fjsp_setzero_v2r8();
 799         fix3             = _fjsp_setzero_v2r8();
 800         fiy3             = _fjsp_setzero_v2r8();
 801         fiz3             = _fjsp_setzero_v2r8();
 802
 803         /* Start inner kernel loop */
 804         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 805         {
 806
 807             /* Get j neighbor index, and coordinate index */
 808             jnrA             = jjnr[jidx];
 809             jnrB             = jjnr[jidx+1];
 810             j_coord_offsetA  = DIM*jnrA;
 811             j_coord_offsetB  = DIM*jnrB;
 812
 813             /* load j atom coordinates */
 814             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 815                                               &jx0,&jy0,&jz0);
 816
 817             /* Calculate displacement vector */
 818             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 819             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 820             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 821             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 822             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 823             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 824             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 825             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 826             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 827             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 828             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 829             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 830
 831             /* Calculate squared distance and things based on it */
 832             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 833             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 834             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 835             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 836
 837             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 838             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 839             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 840             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 841
 842             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 843             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 844             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 845             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 846
 847             /* Load parameters for j particles */
 848             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 849             vdwjidx0A        = 2*vdwtype[jnrA+0];
 850             vdwjidx0B        = 2*vdwtype[jnrB+0];
 851
 852             fjx0             = _fjsp_setzero_v2r8();
 853             fjy0             = _fjsp_setzero_v2r8();
 854             fjz0             = _fjsp_setzero_v2r8();
 855
 856             /**************************
 857              * CALCULATE INTERACTIONS *
 858              **************************/
 859
 860             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 861             {
 862
 863             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 864
 865             /* Compute parameters for interactions between i and j atoms */
 866             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 867                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 868
 869             /* LENNARD-JONES DISPERSION/REPULSION */
 870
 871             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 872             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 873             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 874             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 875             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 876
 877             d                = _fjsp_sub_v2r8(r00,rswitch);
 878             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 879             d2               = _fjsp_mul_v2r8(d,d);
 880             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 881
 882             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 883
 884             /* Evaluate switch function */
 885             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 886             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 887             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 888
 889             fscal            = fvdw;
 890
 891             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 892
 893             /* Update vectorial force */
 894             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 895             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 896             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 897
 898             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 899             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 900             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 901
 902             }
 903
 904             /**************************
 905              * CALCULATE INTERACTIONS *
 906              **************************/
 907
 908             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 909             {
 910
 911             /* Compute parameters for interactions between i and j atoms */
 912             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 913
 914             /* REACTION-FIELD ELECTROSTATICS */
 915             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 916
 917             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 918
 919             fscal            = felec;
 920
 921             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 922
 923             /* Update vectorial force */
 924             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 925             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 926             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 927
 928             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 929             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 930             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 931
 932             }
 933
 934             /**************************
 935              * CALCULATE INTERACTIONS *
 936              **************************/
 937
 938             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 939             {
 940
 941             /* Compute parameters for interactions between i and j atoms */
 942             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 943
 944             /* REACTION-FIELD ELECTROSTATICS */
 945             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 946
 947             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 948
 949             fscal            = felec;
 950
 951             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 952
 953             /* Update vectorial force */
 954             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 955             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 956             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 957
 958             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 959             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 960             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 961
 962             }
 963
 964             /**************************
 965              * CALCULATE INTERACTIONS *
 966              **************************/
 967
 968             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 969             {
 970
 971             /* Compute parameters for interactions between i and j atoms */
 972             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 973
 974             /* REACTION-FIELD ELECTROSTATICS */
 975             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 976
 977             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 978
 979             fscal            = felec;
 980
 981             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 982
 983             /* Update vectorial force */
 984             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 985             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 986             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 987
 988             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 989             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 990             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 991
 992             }
 993
 994             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 995
 996             /* Inner loop uses 161 flops */
 997         }
 998
 999         if(jidx<j_index_end)
1000         {
1001
1002             jnrA             = jjnr[jidx];
1003             j_coord_offsetA  = DIM*jnrA;
1004
1005             /* load j atom coordinates */
1006             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1007                                               &jx0,&jy0,&jz0);
1008
1009             /* Calculate displacement vector */
1010             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1011             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1012             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1013             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1014             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1015             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1016             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1017             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1018             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1019             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1020             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1021             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1022
1023             /* Calculate squared distance and things based on it */
1024             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1025             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1026             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1027             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1028
1029             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1030             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1031             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1032             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1033
1034             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1035             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1036             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1037             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1038
1039             /* Load parameters for j particles */
1040             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1041             vdwjidx0A        = 2*vdwtype[jnrA+0];
1042
1043             fjx0             = _fjsp_setzero_v2r8();
1044             fjy0             = _fjsp_setzero_v2r8();
1045             fjz0             = _fjsp_setzero_v2r8();
1046
1047             /**************************
1048              * CALCULATE INTERACTIONS *
1049              **************************/
1050
1051             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1052             {
1053
1054             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1055
1056             /* Compute parameters for interactions between i and j atoms */
1057             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1058                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1059
1060             /* LENNARD-JONES DISPERSION/REPULSION */
1061
1062             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1063             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
1064             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
1065             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
1066             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
1067
1068             d                = _fjsp_sub_v2r8(r00,rswitch);
1069             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1070             d2               = _fjsp_mul_v2r8(d,d);
1071             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1072
1073             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1074
1075             /* Evaluate switch function */
1076             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1077             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
1078             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1079
1080             fscal            = fvdw;
1081
1082             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1083
1084             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1085
1086             /* Update vectorial force */
1087             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1088             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1089             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1090
1091             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1092             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1093             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1094
1095             }
1096
1097             /**************************
1098              * CALCULATE INTERACTIONS *
1099              **************************/
1100
1101             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1102             {
1103
1104             /* Compute parameters for interactions between i and j atoms */
1105             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1106
1107             /* REACTION-FIELD ELECTROSTATICS */
1108             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1109
1110             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1111
1112             fscal            = felec;
1113
1114             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1115
1116             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1117
1118             /* Update vectorial force */
1119             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1120             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1121             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1122
1123             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1124             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1125             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1126
1127             }
1128
1129             /**************************
1130              * CALCULATE INTERACTIONS *
1131              **************************/
1132
1133             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1134             {
1135
1136             /* Compute parameters for interactions between i and j atoms */
1137             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1138
1139             /* REACTION-FIELD ELECTROSTATICS */
1140             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1141
1142             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1143
1144             fscal            = felec;
1145
1146             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1147
1148             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1149
1150             /* Update vectorial force */
1151             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1152             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1153             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1154
1155             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1156             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1157             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1158
1159             }
1160
1161             /**************************
1162              * CALCULATE INTERACTIONS *
1163              **************************/
1164
1165             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1166             {
1167
1168             /* Compute parameters for interactions between i and j atoms */
1169             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1170
1171             /* REACTION-FIELD ELECTROSTATICS */
1172             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1173
1174             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1175
1176             fscal            = felec;
1177
1178             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1179
1180             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1181
1182             /* Update vectorial force */
1183             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1184             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1185             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1186
1187             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1188             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1189             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1190
1191             }
1192
1193             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1194
1195             /* Inner loop uses 161 flops */
1196         }
1197
1198         /* End of innermost loop */
1199
1200         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1201                                               f+i_coord_offset,fshift+i_shift_offset);
1202
1203         /* Increment number of inner iterations */
1204         inneriter                  += j_index_end - j_index_start;
1205
1206         /* Outer loop uses 24 flops */
1207     }
1208
1209     /* Increment number of outer iterations */
1210     outeriter        += nri;
1211
1212     /* Update outer/inner flops */
1213
1214     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*161);
1215 }