src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_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_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_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              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  99     real             rswitch_scalar,d_scalar;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 118     charge           = mdatoms->chargeA;
 119     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 120     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 121     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 134     rcutoff_scalar   = fr->ic->rcoulomb;
 135     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 136     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 137
 138     rswitch_scalar   = fr->ic->rvdw_switch;
 139     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 140     /* Setup switch parameters */
 141     d_scalar         = rcutoff_scalar-rswitch_scalar;
 142     d                = gmx_fjsp_set1_v2r8(d_scalar);
 143     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 144     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 145     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 146     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 147     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 148     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 149
 150     /* Avoid stupid compiler warnings */
 151     jnrA = jnrB = 0;
 152     j_coord_offsetA = 0;
 153     j_coord_offsetB = 0;
 154
 155     outeriter        = 0;
 156     inneriter        = 0;
 157
 158     /* Start outer loop over neighborlists */
 159     for(iidx=0; iidx<nri; iidx++)
 160     {
 161         /* Load shift vector for this list */
 162         i_shift_offset   = DIM*shiftidx[iidx];
 163
 164         /* Load limits for loop over neighbors */
 165         j_index_start    = jindex[iidx];
 166         j_index_end      = jindex[iidx+1];
 167
 168         /* Get outer coordinate index */
 169         inr              = iinr[iidx];
 170         i_coord_offset   = DIM*inr;
 171
 172         /* Load i particle coords and add shift vector */
 173         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 174                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 175
 176         fix0             = _fjsp_setzero_v2r8();
 177         fiy0             = _fjsp_setzero_v2r8();
 178         fiz0             = _fjsp_setzero_v2r8();
 179         fix1             = _fjsp_setzero_v2r8();
 180         fiy1             = _fjsp_setzero_v2r8();
 181         fiz1             = _fjsp_setzero_v2r8();
 182         fix2             = _fjsp_setzero_v2r8();
 183         fiy2             = _fjsp_setzero_v2r8();
 184         fiz2             = _fjsp_setzero_v2r8();
 185
 186         /* Reset potential sums */
 187         velecsum         = _fjsp_setzero_v2r8();
 188         vvdwsum          = _fjsp_setzero_v2r8();
 189
 190         /* Start inner kernel loop */
 191         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 192         {
 193
 194             /* Get j neighbor index, and coordinate index */
 195             jnrA             = jjnr[jidx];
 196             jnrB             = jjnr[jidx+1];
 197             j_coord_offsetA  = DIM*jnrA;
 198             j_coord_offsetB  = DIM*jnrB;
 199
 200             /* load j atom coordinates */
 201             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 202                                               &jx0,&jy0,&jz0);
 203
 204             /* Calculate displacement vector */
 205             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 206             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 207             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 208             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 209             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 210             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 211             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 212             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 213             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 214
 215             /* Calculate squared distance and things based on it */
 216             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 217             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 218             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 219
 220             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 221             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 222             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 223
 224             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 225             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 226             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 227
 228             /* Load parameters for j particles */
 229             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 230             vdwjidx0A        = 2*vdwtype[jnrA+0];
 231             vdwjidx0B        = 2*vdwtype[jnrB+0];
 232
 233             fjx0             = _fjsp_setzero_v2r8();
 234             fjy0             = _fjsp_setzero_v2r8();
 235             fjz0             = _fjsp_setzero_v2r8();
 236
 237             /**************************
 238              * CALCULATE INTERACTIONS *
 239              **************************/
 240
 241             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 242             {
 243
 244             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 245
 246             /* Compute parameters for interactions between i and j atoms */
 247             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 248             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 249                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 250
 251             /* REACTION-FIELD ELECTROSTATICS */
 252             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 253             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 254
 255             /* LENNARD-JONES DISPERSION/REPULSION */
 256
 257             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 258             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 259             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 260             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 261             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 262
 263             d                = _fjsp_sub_v2r8(r00,rswitch);
 264             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 265             d2               = _fjsp_mul_v2r8(d,d);
 266             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 267
 268             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 269
 270             /* Evaluate switch function */
 271             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 272             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 273             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 274             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 275
 276             /* Update potential sum for this i atom from the interaction with this j atom. */
 277             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 278             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 279             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = _fjsp_add_v2r8(felec,fvdw);
 283
 284             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 285
 286             /* Update vectorial force */
 287             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 288             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 289             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 294
 295             }
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 302             {
 303
 304             /* Compute parameters for interactions between i and j atoms */
 305             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 306
 307             /* REACTION-FIELD ELECTROSTATICS */
 308             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 309             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 310
 311             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 312
 313             /* Update potential sum for this i atom from the interaction with this j atom. */
 314             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 315             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 316
 317             fscal            = felec;
 318
 319             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 320
 321             /* Update vectorial force */
 322             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 323             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 324             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 325
 326             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 327             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 328             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 329
 330             }
 331
 332             /**************************
 333              * CALCULATE INTERACTIONS *
 334              **************************/
 335
 336             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 337             {
 338
 339             /* Compute parameters for interactions between i and j atoms */
 340             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 341
 342             /* REACTION-FIELD ELECTROSTATICS */
 343             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 344             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 345
 346             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 347
 348             /* Update potential sum for this i atom from the interaction with this j atom. */
 349             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 350             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 351
 352             fscal            = felec;
 353
 354             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 355
 356             /* Update vectorial force */
 357             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 358             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 359             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 360
 361             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 362             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 363             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 364
 365             }
 366
 367             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 368
 369             /* Inner loop uses 154 flops */
 370         }
 371
 372         if(jidx<j_index_end)
 373         {
 374
 375             jnrA             = jjnr[jidx];
 376             j_coord_offsetA  = DIM*jnrA;
 377
 378             /* load j atom coordinates */
 379             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 380                                               &jx0,&jy0,&jz0);
 381
 382             /* Calculate displacement vector */
 383             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 384             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 385             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 386             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 387             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 388             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 389             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 390             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 391             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 392
 393             /* Calculate squared distance and things based on it */
 394             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 395             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 396             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 397
 398             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 399             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 400             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 401
 402             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 403             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 404             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 405
 406             /* Load parameters for j particles */
 407             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 408             vdwjidx0A        = 2*vdwtype[jnrA+0];
 409
 410             fjx0             = _fjsp_setzero_v2r8();
 411             fjy0             = _fjsp_setzero_v2r8();
 412             fjz0             = _fjsp_setzero_v2r8();
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 419             {
 420
 421             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 422
 423             /* Compute parameters for interactions between i and j atoms */
 424             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 425             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 426                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 427
 428             /* REACTION-FIELD ELECTROSTATICS */
 429             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 430             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 431
 432             /* LENNARD-JONES DISPERSION/REPULSION */
 433
 434             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 435             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 436             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 437             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 438             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 439
 440             d                = _fjsp_sub_v2r8(r00,rswitch);
 441             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 442             d2               = _fjsp_mul_v2r8(d,d);
 443             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 444
 445             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 446
 447             /* Evaluate switch function */
 448             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 449             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 450             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 451             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 452
 453             /* Update potential sum for this i atom from the interaction with this j atom. */
 454             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 455             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 456             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 457             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 458             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 459             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 460
 461             fscal            = _fjsp_add_v2r8(felec,fvdw);
 462
 463             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 464
 465             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 466
 467             /* Update vectorial force */
 468             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 469             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 470             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 471
 472             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 473             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 474             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 475
 476             }
 477
 478             /**************************
 479              * CALCULATE INTERACTIONS *
 480              **************************/
 481
 482             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 483             {
 484
 485             /* Compute parameters for interactions between i and j atoms */
 486             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 487
 488             /* REACTION-FIELD ELECTROSTATICS */
 489             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 490             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 491
 492             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 496             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 497             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 498
 499             fscal            = felec;
 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             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 507             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 508             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 509
 510             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 511             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 512             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 513
 514             }
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 521             {
 522
 523             /* Compute parameters for interactions between i and j atoms */
 524             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 525
 526             /* REACTION-FIELD ELECTROSTATICS */
 527             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 528             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 529
 530             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,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             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 545             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 546             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 547
 548             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 549             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 550             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 551
 552             }
 553
 554             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 555
 556             /* Inner loop uses 154 flops */
 557         }
 558
 559         /* End of innermost loop */
 560
 561         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 562                                               f+i_coord_offset,fshift+i_shift_offset);
 563
 564         ggid                        = gid[iidx];
 565         /* Update potential energies */
 566         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 567         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 568
 569         /* Increment number of inner iterations */
 570         inneriter                  += j_index_end - j_index_start;
 571
 572         /* Outer loop uses 20 flops */
 573     }
 574
 575     /* Increment number of outer iterations */
 576     outeriter        += nri;
 577
 578     /* Update outer/inner flops */
 579
 580     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 581 }
 582 /*
 583  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 584  * Electrostatics interaction: ReactionField
 585  * VdW interaction:            LennardJones
 586  * Geometry:                   Water3-Particle
 587  * Calculate force/pot:        Force
 588  */
 589 void
 590 nb_kernel_ElecRFCut_VdwLJSw_GeomW3P1_F_sparc64_hpc_ace_double
 591                     (t_nblist                    * gmx_restrict       nlist,
 592                      rvec                        * gmx_restrict          xx,
 593                      rvec                        * gmx_restrict          ff,
 594                      struct t_forcerec           * gmx_restrict          fr,
 595                      t_mdatoms                   * gmx_restrict     mdatoms,
 596                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 597                      t_nrnb                      * gmx_restrict        nrnb)
 598 {
 599     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 600      * just 0 for non-waters.
 601      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 602      * jnr indices corresponding to data put in the four positions in the SIMD register.
 603      */
 604     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 605     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 606     int              jnrA,jnrB;
 607     int              j_coord_offsetA,j_coord_offsetB;
 608     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 609     real             rcutoff_scalar;
 610     real             *shiftvec,*fshift,*x,*f;
 611     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 612     int              vdwioffset0;
 613     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 614     int              vdwioffset1;
 615     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 616     int              vdwioffset2;
 617     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 618     int              vdwjidx0A,vdwjidx0B;
 619     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 620     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 621     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 622     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 623     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 624     real             *charge;
 625     int              nvdwtype;
 626     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 627     int              *vdwtype;
 628     real             *vdwparam;
 629     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 630     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 631     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 632     real             rswitch_scalar,d_scalar;
 633     _fjsp_v2r8       itab_tmp;
 634     _fjsp_v2r8       dummy_mask,cutoff_mask;
 635     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 636     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 637     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 638
 639     x                = xx[0];
 640     f                = ff[0];
 641
 642     nri              = nlist->nri;
 643     iinr             = nlist->iinr;
 644     jindex           = nlist->jindex;
 645     jjnr             = nlist->jjnr;
 646     shiftidx         = nlist->shift;
 647     gid              = nlist->gid;
 648     shiftvec         = fr->shift_vec[0];
 649     fshift           = fr->fshift[0];
 650     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 651     charge           = mdatoms->chargeA;
 652     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 653     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 654     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 655     nvdwtype         = fr->ntype;
 656     vdwparam         = fr->nbfp;
 657     vdwtype          = mdatoms->typeA;
 658
 659     /* Setup water-specific parameters */
 660     inr              = nlist->iinr[0];
 661     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 662     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 663     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 664     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 665
 666     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 667     rcutoff_scalar   = fr->ic->rcoulomb;
 668     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 669     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 670
 671     rswitch_scalar   = fr->ic->rvdw_switch;
 672     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 673     /* Setup switch parameters */
 674     d_scalar         = rcutoff_scalar-rswitch_scalar;
 675     d                = gmx_fjsp_set1_v2r8(d_scalar);
 676     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 677     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 678     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 679     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 680     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 681     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 682
 683     /* Avoid stupid compiler warnings */
 684     jnrA = jnrB = 0;
 685     j_coord_offsetA = 0;
 686     j_coord_offsetB = 0;
 687
 688     outeriter        = 0;
 689     inneriter        = 0;
 690
 691     /* Start outer loop over neighborlists */
 692     for(iidx=0; iidx<nri; iidx++)
 693     {
 694         /* Load shift vector for this list */
 695         i_shift_offset   = DIM*shiftidx[iidx];
 696
 697         /* Load limits for loop over neighbors */
 698         j_index_start    = jindex[iidx];
 699         j_index_end      = jindex[iidx+1];
 700
 701         /* Get outer coordinate index */
 702         inr              = iinr[iidx];
 703         i_coord_offset   = DIM*inr;
 704
 705         /* Load i particle coords and add shift vector */
 706         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 707                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 708
 709         fix0             = _fjsp_setzero_v2r8();
 710         fiy0             = _fjsp_setzero_v2r8();
 711         fiz0             = _fjsp_setzero_v2r8();
 712         fix1             = _fjsp_setzero_v2r8();
 713         fiy1             = _fjsp_setzero_v2r8();
 714         fiz1             = _fjsp_setzero_v2r8();
 715         fix2             = _fjsp_setzero_v2r8();
 716         fiy2             = _fjsp_setzero_v2r8();
 717         fiz2             = _fjsp_setzero_v2r8();
 718
 719         /* Start inner kernel loop */
 720         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 721         {
 722
 723             /* Get j neighbor index, and coordinate index */
 724             jnrA             = jjnr[jidx];
 725             jnrB             = jjnr[jidx+1];
 726             j_coord_offsetA  = DIM*jnrA;
 727             j_coord_offsetB  = DIM*jnrB;
 728
 729             /* load j atom coordinates */
 730             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 731                                               &jx0,&jy0,&jz0);
 732
 733             /* Calculate displacement vector */
 734             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 735             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 736             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 737             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 738             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 739             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 740             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 741             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 742             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 743
 744             /* Calculate squared distance and things based on it */
 745             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 746             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 747             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 748
 749             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 750             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 751             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 752
 753             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 754             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 755             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 756
 757             /* Load parameters for j particles */
 758             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 759             vdwjidx0A        = 2*vdwtype[jnrA+0];
 760             vdwjidx0B        = 2*vdwtype[jnrB+0];
 761
 762             fjx0             = _fjsp_setzero_v2r8();
 763             fjy0             = _fjsp_setzero_v2r8();
 764             fjz0             = _fjsp_setzero_v2r8();
 765
 766             /**************************
 767              * CALCULATE INTERACTIONS *
 768              **************************/
 769
 770             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 771             {
 772
 773             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 774
 775             /* Compute parameters for interactions between i and j atoms */
 776             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 777             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 778                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 779
 780             /* REACTION-FIELD ELECTROSTATICS */
 781             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 782
 783             /* LENNARD-JONES DISPERSION/REPULSION */
 784
 785             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 786             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 787             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 788             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 789             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 790
 791             d                = _fjsp_sub_v2r8(r00,rswitch);
 792             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 793             d2               = _fjsp_mul_v2r8(d,d);
 794             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 795
 796             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 797
 798             /* Evaluate switch function */
 799             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 800             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 801             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 802
 803             fscal            = _fjsp_add_v2r8(felec,fvdw);
 804
 805             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 806
 807             /* Update vectorial force */
 808             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 809             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 810             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 811
 812             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 813             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 814             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 815
 816             }
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 823             {
 824
 825             /* Compute parameters for interactions between i and j atoms */
 826             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 827
 828             /* REACTION-FIELD ELECTROSTATICS */
 829             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 830
 831             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 832
 833             fscal            = felec;
 834
 835             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 836
 837             /* Update vectorial force */
 838             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 839             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 840             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 841
 842             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 843             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 844             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 845
 846             }
 847
 848             /**************************
 849              * CALCULATE INTERACTIONS *
 850              **************************/
 851
 852             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 853             {
 854
 855             /* Compute parameters for interactions between i and j atoms */
 856             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 857
 858             /* REACTION-FIELD ELECTROSTATICS */
 859             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 860
 861             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 862
 863             fscal            = felec;
 864
 865             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 866
 867             /* Update vectorial force */
 868             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 869             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 870             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 871
 872             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 873             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 874             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 875
 876             }
 877
 878             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 879
 880             /* Inner loop uses 133 flops */
 881         }
 882
 883         if(jidx<j_index_end)
 884         {
 885
 886             jnrA             = jjnr[jidx];
 887             j_coord_offsetA  = DIM*jnrA;
 888
 889             /* load j atom coordinates */
 890             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 891                                               &jx0,&jy0,&jz0);
 892
 893             /* Calculate displacement vector */
 894             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 895             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 896             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 897             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 898             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 899             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 900             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 901             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 902             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 903
 904             /* Calculate squared distance and things based on it */
 905             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 906             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 907             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 908
 909             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 910             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 911             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 912
 913             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 914             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 915             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 916
 917             /* Load parameters for j particles */
 918             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 919             vdwjidx0A        = 2*vdwtype[jnrA+0];
 920
 921             fjx0             = _fjsp_setzero_v2r8();
 922             fjy0             = _fjsp_setzero_v2r8();
 923             fjz0             = _fjsp_setzero_v2r8();
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 930             {
 931
 932             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 933
 934             /* Compute parameters for interactions between i and j atoms */
 935             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 936             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 937                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 938
 939             /* REACTION-FIELD ELECTROSTATICS */
 940             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 941
 942             /* LENNARD-JONES DISPERSION/REPULSION */
 943
 944             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 945             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 946             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 947             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 948             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 949
 950             d                = _fjsp_sub_v2r8(r00,rswitch);
 951             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 952             d2               = _fjsp_mul_v2r8(d,d);
 953             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 954
 955             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 956
 957             /* Evaluate switch function */
 958             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 959             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 960             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 961
 962             fscal            = _fjsp_add_v2r8(felec,fvdw);
 963
 964             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 965
 966             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 967
 968             /* Update vectorial force */
 969             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 970             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 971             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 972
 973             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 974             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 975             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 976
 977             }
 978
 979             /**************************
 980              * CALCULATE INTERACTIONS *
 981              **************************/
 982
 983             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 984             {
 985
 986             /* Compute parameters for interactions between i and j atoms */
 987             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 988
 989             /* REACTION-FIELD ELECTROSTATICS */
 990             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 991
 992             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 993
 994             fscal            = felec;
 995
 996             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 997
 998             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 999
1000             /* Update vectorial force */
1001             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1002             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1003             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1004
1005             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1006             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1007             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1008
1009             }
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1016             {
1017
1018             /* Compute parameters for interactions between i and j atoms */
1019             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1020
1021             /* REACTION-FIELD ELECTROSTATICS */
1022             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1023
1024             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1025
1026             fscal            = felec;
1027
1028             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1029
1030             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1031
1032             /* Update vectorial force */
1033             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1034             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1035             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1036
1037             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1038             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1039             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1040
1041             }
1042
1043             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1044
1045             /* Inner loop uses 133 flops */
1046         }
1047
1048         /* End of innermost loop */
1049
1050         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1051                                               f+i_coord_offset,fshift+i_shift_offset);
1052
1053         /* Increment number of inner iterations */
1054         inneriter                  += j_index_end - j_index_start;
1055
1056         /* Outer loop uses 18 flops */
1057     }
1058
1059     /* Increment number of outer iterations */
1060     outeriter        += nri;
1061
1062     /* Update outer/inner flops */
1063
1064     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*133);
1065 }