src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSh_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, 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_VdwLJSh_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwLJSh_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       itab_tmp;
 102     _fjsp_v2r8       dummy_mask,cutoff_mask;
 103     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 104     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 105     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 106
 107     x                = xx[0];
 108     f                = ff[0];
 109
 110     nri              = nlist->nri;
 111     iinr             = nlist->iinr;
 112     jindex           = nlist->jindex;
 113     jjnr             = nlist->jjnr;
 114     shiftidx         = nlist->shift;
 115     gid              = nlist->gid;
 116     shiftvec         = fr->shift_vec[0];
 117     fshift           = fr->fshift[0];
 118     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 119     charge           = mdatoms->chargeA;
 120     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 121     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 122     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     /* Setup water-specific parameters */
 128     inr              = nlist->iinr[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     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 132     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 133
 134     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 135     rcutoff_scalar   = fr->ic->rcoulomb;
 136     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 137     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 138
 139     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 140     rvdw             = gmx_fjsp_set1_v2r8(fr->ic->rvdw);
 141
 142     /* Avoid stupid compiler warnings */
 143     jnrA = jnrB = 0;
 144     j_coord_offsetA = 0;
 145     j_coord_offsetB = 0;
 146
 147     outeriter        = 0;
 148     inneriter        = 0;
 149
 150     /* Start outer loop over neighborlists */
 151     for(iidx=0; iidx<nri; iidx++)
 152     {
 153         /* Load shift vector for this list */
 154         i_shift_offset   = DIM*shiftidx[iidx];
 155
 156         /* Load limits for loop over neighbors */
 157         j_index_start    = jindex[iidx];
 158         j_index_end      = jindex[iidx+1];
 159
 160         /* Get outer coordinate index */
 161         inr              = iinr[iidx];
 162         i_coord_offset   = DIM*inr;
 163
 164         /* Load i particle coords and add shift vector */
 165         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 166                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 167
 168         fix0             = _fjsp_setzero_v2r8();
 169         fiy0             = _fjsp_setzero_v2r8();
 170         fiz0             = _fjsp_setzero_v2r8();
 171         fix1             = _fjsp_setzero_v2r8();
 172         fiy1             = _fjsp_setzero_v2r8();
 173         fiz1             = _fjsp_setzero_v2r8();
 174         fix2             = _fjsp_setzero_v2r8();
 175         fiy2             = _fjsp_setzero_v2r8();
 176         fiz2             = _fjsp_setzero_v2r8();
 177         fix3             = _fjsp_setzero_v2r8();
 178         fiy3             = _fjsp_setzero_v2r8();
 179         fiz3             = _fjsp_setzero_v2r8();
 180
 181         /* Reset potential sums */
 182         velecsum         = _fjsp_setzero_v2r8();
 183         vvdwsum          = _fjsp_setzero_v2r8();
 184
 185         /* Start inner kernel loop */
 186         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 187         {
 188
 189             /* Get j neighbor index, and coordinate index */
 190             jnrA             = jjnr[jidx];
 191             jnrB             = jjnr[jidx+1];
 192             j_coord_offsetA  = DIM*jnrA;
 193             j_coord_offsetB  = DIM*jnrB;
 194
 195             /* load j atom coordinates */
 196             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 197                                               &jx0,&jy0,&jz0);
 198
 199             /* Calculate displacement vector */
 200             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 201             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 202             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 203             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 204             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 205             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 206             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 207             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 208             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 209             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 210             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 211             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 212
 213             /* Calculate squared distance and things based on it */
 214             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 215             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 216             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 217             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 218
 219             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 220             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 221             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 222
 223             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 224             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 225             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 226             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 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             /* Compute parameters for interactions between i and j atoms */
 245             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 246                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 247
 248             /* LENNARD-JONES DISPERSION/REPULSION */
 249
 250             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 251             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 252             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 253             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 254                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 255             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 256
 257             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 258
 259             /* Update potential sum for this i atom from the interaction with this j atom. */
 260             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 261             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 262
 263             fscal            = fvdw;
 264
 265             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 266
 267             /* Update vectorial force */
 268             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 269             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 270             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 271
 272             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 273             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 274             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 275
 276             }
 277
 278             /**************************
 279              * CALCULATE INTERACTIONS *
 280              **************************/
 281
 282             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 283             {
 284
 285             /* Compute parameters for interactions between i and j atoms */
 286             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 287
 288             /* REACTION-FIELD ELECTROSTATICS */
 289             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 290             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 291
 292             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 293
 294             /* Update potential sum for this i atom from the interaction with this j atom. */
 295             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 296             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 297
 298             fscal            = felec;
 299
 300             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 301
 302             /* Update vectorial force */
 303             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 304             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 305             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 306
 307             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 308             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 309             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 310
 311             }
 312
 313             /**************************
 314              * CALCULATE INTERACTIONS *
 315              **************************/
 316
 317             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 318             {
 319
 320             /* Compute parameters for interactions between i and j atoms */
 321             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 322
 323             /* REACTION-FIELD ELECTROSTATICS */
 324             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 325             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 326
 327             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 331             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 332
 333             fscal            = felec;
 334
 335             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 336
 337             /* Update vectorial force */
 338             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 339             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 340             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 341
 342             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 343             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 344             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 345
 346             }
 347
 348             /**************************
 349              * CALCULATE INTERACTIONS *
 350              **************************/
 351
 352             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 353             {
 354
 355             /* Compute parameters for interactions between i and j atoms */
 356             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 357
 358             /* REACTION-FIELD ELECTROSTATICS */
 359             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 360             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 361
 362             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 363
 364             /* Update potential sum for this i atom from the interaction with this j atom. */
 365             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 366             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 371
 372             /* Update vectorial force */
 373             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 374             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 375             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 376
 377             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 378             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 379             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 380
 381             }
 382
 383             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 384
 385             /* Inner loop uses 164 flops */
 386         }
 387
 388         if(jidx<j_index_end)
 389         {
 390
 391             jnrA             = jjnr[jidx];
 392             j_coord_offsetA  = DIM*jnrA;
 393
 394             /* load j atom coordinates */
 395             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 396                                               &jx0,&jy0,&jz0);
 397
 398             /* Calculate displacement vector */
 399             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 400             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 401             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 402             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 403             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 404             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 405             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 406             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 407             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 408             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 409             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 410             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 411
 412             /* Calculate squared distance and things based on it */
 413             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 414             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 415             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 416             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 417
 418             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 419             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 420             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 421
 422             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 423             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 424             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 425             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 426
 427             /* Load parameters for j particles */
 428             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 429             vdwjidx0A        = 2*vdwtype[jnrA+0];
 430
 431             fjx0             = _fjsp_setzero_v2r8();
 432             fjy0             = _fjsp_setzero_v2r8();
 433             fjz0             = _fjsp_setzero_v2r8();
 434
 435             /**************************
 436              * CALCULATE INTERACTIONS *
 437              **************************/
 438
 439             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 440             {
 441
 442             /* Compute parameters for interactions between i and j atoms */
 443             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 444                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 445
 446             /* LENNARD-JONES DISPERSION/REPULSION */
 447
 448             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 449             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 450             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 451             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 452                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 453             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 454
 455             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 456
 457             /* Update potential sum for this i atom from the interaction with this j atom. */
 458             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 459             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 460             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 461
 462             fscal            = fvdw;
 463
 464             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 465
 466             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 467
 468             /* Update vectorial force */
 469             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 470             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 471             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 472
 473             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 474             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 475             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 476
 477             }
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 484             {
 485
 486             /* Compute parameters for interactions between i and j atoms */
 487             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 488
 489             /* REACTION-FIELD ELECTROSTATICS */
 490             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 491             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 492
 493             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 497             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 498             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 499
 500             fscal            = felec;
 501
 502             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 508             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 509             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 514
 515             }
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 522             {
 523
 524             /* Compute parameters for interactions between i and j atoms */
 525             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 526
 527             /* REACTION-FIELD ELECTROSTATICS */
 528             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 529             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 530
 531             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 532
 533             /* Update potential sum for this i atom from the interaction with this j atom. */
 534             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 535             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 536             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 537
 538             fscal            = felec;
 539
 540             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 541
 542             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 543
 544             /* Update vectorial force */
 545             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 546             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 547             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 548
 549             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 550             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 551             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 552
 553             }
 554
 555             /**************************
 556              * CALCULATE INTERACTIONS *
 557              **************************/
 558
 559             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 560             {
 561
 562             /* Compute parameters for interactions between i and j atoms */
 563             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 564
 565             /* REACTION-FIELD ELECTROSTATICS */
 566             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 567             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 568
 569             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 570
 571             /* Update potential sum for this i atom from the interaction with this j atom. */
 572             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 573             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 574             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 575
 576             fscal            = felec;
 577
 578             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 579
 580             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 581
 582             /* Update vectorial force */
 583             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 584             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 585             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 586
 587             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 588             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 589             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 590
 591             }
 592
 593             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 594
 595             /* Inner loop uses 164 flops */
 596         }
 597
 598         /* End of innermost loop */
 599
 600         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 601                                               f+i_coord_offset,fshift+i_shift_offset);
 602
 603         ggid                        = gid[iidx];
 604         /* Update potential energies */
 605         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 606         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 607
 608         /* Increment number of inner iterations */
 609         inneriter                  += j_index_end - j_index_start;
 610
 611         /* Outer loop uses 26 flops */
 612     }
 613
 614     /* Increment number of outer iterations */
 615     outeriter        += nri;
 616
 617     /* Update outer/inner flops */
 618
 619     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*164);
 620 }
 621 /*
 622  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomW4P1_F_sparc64_hpc_ace_double
 623  * Electrostatics interaction: ReactionField
 624  * VdW interaction:            LennardJones
 625  * Geometry:                   Water4-Particle
 626  * Calculate force/pot:        Force
 627  */
 628 void
 629 nb_kernel_ElecRFCut_VdwLJSh_GeomW4P1_F_sparc64_hpc_ace_double
 630                     (t_nblist                    * gmx_restrict       nlist,
 631                      rvec                        * gmx_restrict          xx,
 632                      rvec                        * gmx_restrict          ff,
 633                      struct t_forcerec           * gmx_restrict          fr,
 634                      t_mdatoms                   * gmx_restrict     mdatoms,
 635                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 636                      t_nrnb                      * gmx_restrict        nrnb)
 637 {
 638     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 639      * just 0 for non-waters.
 640      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 641      * jnr indices corresponding to data put in the four positions in the SIMD register.
 642      */
 643     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 644     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 645     int              jnrA,jnrB;
 646     int              j_coord_offsetA,j_coord_offsetB;
 647     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 648     real             rcutoff_scalar;
 649     real             *shiftvec,*fshift,*x,*f;
 650     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 651     int              vdwioffset0;
 652     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 653     int              vdwioffset1;
 654     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 655     int              vdwioffset2;
 656     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 657     int              vdwioffset3;
 658     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 659     int              vdwjidx0A,vdwjidx0B;
 660     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 661     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 662     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 663     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 664     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 665     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 666     real             *charge;
 667     int              nvdwtype;
 668     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 669     int              *vdwtype;
 670     real             *vdwparam;
 671     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 672     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 673     _fjsp_v2r8       itab_tmp;
 674     _fjsp_v2r8       dummy_mask,cutoff_mask;
 675     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 676     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 677     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 678
 679     x                = xx[0];
 680     f                = ff[0];
 681
 682     nri              = nlist->nri;
 683     iinr             = nlist->iinr;
 684     jindex           = nlist->jindex;
 685     jjnr             = nlist->jjnr;
 686     shiftidx         = nlist->shift;
 687     gid              = nlist->gid;
 688     shiftvec         = fr->shift_vec[0];
 689     fshift           = fr->fshift[0];
 690     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 691     charge           = mdatoms->chargeA;
 692     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 693     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 694     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 695     nvdwtype         = fr->ntype;
 696     vdwparam         = fr->nbfp;
 697     vdwtype          = mdatoms->typeA;
 698
 699     /* Setup water-specific parameters */
 700     inr              = nlist->iinr[0];
 701     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 702     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 703     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 704     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 705
 706     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 707     rcutoff_scalar   = fr->ic->rcoulomb;
 708     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 709     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 710
 711     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 712     rvdw             = gmx_fjsp_set1_v2r8(fr->ic->rvdw);
 713
 714     /* Avoid stupid compiler warnings */
 715     jnrA = jnrB = 0;
 716     j_coord_offsetA = 0;
 717     j_coord_offsetB = 0;
 718
 719     outeriter        = 0;
 720     inneriter        = 0;
 721
 722     /* Start outer loop over neighborlists */
 723     for(iidx=0; iidx<nri; iidx++)
 724     {
 725         /* Load shift vector for this list */
 726         i_shift_offset   = DIM*shiftidx[iidx];
 727
 728         /* Load limits for loop over neighbors */
 729         j_index_start    = jindex[iidx];
 730         j_index_end      = jindex[iidx+1];
 731
 732         /* Get outer coordinate index */
 733         inr              = iinr[iidx];
 734         i_coord_offset   = DIM*inr;
 735
 736         /* Load i particle coords and add shift vector */
 737         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 738                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 739
 740         fix0             = _fjsp_setzero_v2r8();
 741         fiy0             = _fjsp_setzero_v2r8();
 742         fiz0             = _fjsp_setzero_v2r8();
 743         fix1             = _fjsp_setzero_v2r8();
 744         fiy1             = _fjsp_setzero_v2r8();
 745         fiz1             = _fjsp_setzero_v2r8();
 746         fix2             = _fjsp_setzero_v2r8();
 747         fiy2             = _fjsp_setzero_v2r8();
 748         fiz2             = _fjsp_setzero_v2r8();
 749         fix3             = _fjsp_setzero_v2r8();
 750         fiy3             = _fjsp_setzero_v2r8();
 751         fiz3             = _fjsp_setzero_v2r8();
 752
 753         /* Start inner kernel loop */
 754         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 755         {
 756
 757             /* Get j neighbor index, and coordinate index */
 758             jnrA             = jjnr[jidx];
 759             jnrB             = jjnr[jidx+1];
 760             j_coord_offsetA  = DIM*jnrA;
 761             j_coord_offsetB  = DIM*jnrB;
 762
 763             /* load j atom coordinates */
 764             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 765                                               &jx0,&jy0,&jz0);
 766
 767             /* Calculate displacement vector */
 768             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 769             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 770             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 771             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 772             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 773             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 774             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 775             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 776             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 777             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 778             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 779             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 780
 781             /* Calculate squared distance and things based on it */
 782             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 783             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 784             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 785             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 786
 787             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 788             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 789             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 790
 791             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 792             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 793             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 794             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 795
 796             /* Load parameters for j particles */
 797             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 798             vdwjidx0A        = 2*vdwtype[jnrA+0];
 799             vdwjidx0B        = 2*vdwtype[jnrB+0];
 800
 801             fjx0             = _fjsp_setzero_v2r8();
 802             fjy0             = _fjsp_setzero_v2r8();
 803             fjz0             = _fjsp_setzero_v2r8();
 804
 805             /**************************
 806              * CALCULATE INTERACTIONS *
 807              **************************/
 808
 809             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 810             {
 811
 812             /* Compute parameters for interactions between i and j atoms */
 813             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 814                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 815
 816             /* LENNARD-JONES DISPERSION/REPULSION */
 817
 818             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 819             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 820
 821             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 822
 823             fscal            = fvdw;
 824
 825             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 826
 827             /* Update vectorial force */
 828             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 829             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 830             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 831
 832             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 833             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 834             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 835
 836             }
 837
 838             /**************************
 839              * CALCULATE INTERACTIONS *
 840              **************************/
 841
 842             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 843             {
 844
 845             /* Compute parameters for interactions between i and j atoms */
 846             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 847
 848             /* REACTION-FIELD ELECTROSTATICS */
 849             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 850
 851             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 852
 853             fscal            = felec;
 854
 855             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 856
 857             /* Update vectorial force */
 858             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 859             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 860             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 861
 862             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 863             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 864             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 865
 866             }
 867
 868             /**************************
 869              * CALCULATE INTERACTIONS *
 870              **************************/
 871
 872             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 873             {
 874
 875             /* Compute parameters for interactions between i and j atoms */
 876             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 877
 878             /* REACTION-FIELD ELECTROSTATICS */
 879             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 880
 881             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 882
 883             fscal            = felec;
 884
 885             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 886
 887             /* Update vectorial force */
 888             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 889             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 890             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 891
 892             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 893             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 894             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 895
 896             }
 897
 898             /**************************
 899              * CALCULATE INTERACTIONS *
 900              **************************/
 901
 902             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 903             {
 904
 905             /* Compute parameters for interactions between i and j atoms */
 906             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 907
 908             /* REACTION-FIELD ELECTROSTATICS */
 909             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 910
 911             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 912
 913             fscal            = felec;
 914
 915             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 916
 917             /* Update vectorial force */
 918             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 919             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 920             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 921
 922             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 923             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 924             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 925
 926             }
 927
 928             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 929
 930             /* Inner loop uses 135 flops */
 931         }
 932
 933         if(jidx<j_index_end)
 934         {
 935
 936             jnrA             = jjnr[jidx];
 937             j_coord_offsetA  = DIM*jnrA;
 938
 939             /* load j atom coordinates */
 940             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 941                                               &jx0,&jy0,&jz0);
 942
 943             /* Calculate displacement vector */
 944             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 945             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 946             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 947             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 948             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 949             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 950             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 951             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 952             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 953             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 954             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 955             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 956
 957             /* Calculate squared distance and things based on it */
 958             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 959             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 960             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 961             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 962
 963             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 964             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 965             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 966
 967             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 968             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 969             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 970             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 971
 972             /* Load parameters for j particles */
 973             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 974             vdwjidx0A        = 2*vdwtype[jnrA+0];
 975
 976             fjx0             = _fjsp_setzero_v2r8();
 977             fjy0             = _fjsp_setzero_v2r8();
 978             fjz0             = _fjsp_setzero_v2r8();
 979
 980             /**************************
 981              * CALCULATE INTERACTIONS *
 982              **************************/
 983
 984             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 985             {
 986
 987             /* Compute parameters for interactions between i and j atoms */
 988             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 989                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 990
 991             /* LENNARD-JONES DISPERSION/REPULSION */
 992
 993             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 994             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 995
 996             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 997
 998             fscal            = fvdw;
 999
1000             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1001
1002             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1003
1004             /* Update vectorial force */
1005             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1006             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1007             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1008
1009             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1010             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1011             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1012
1013             }
1014
1015             /**************************
1016              * CALCULATE INTERACTIONS *
1017              **************************/
1018
1019             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1020             {
1021
1022             /* Compute parameters for interactions between i and j atoms */
1023             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1024
1025             /* REACTION-FIELD ELECTROSTATICS */
1026             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1027
1028             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1029
1030             fscal            = felec;
1031
1032             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1033
1034             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1035
1036             /* Update vectorial force */
1037             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1038             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1039             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1040
1041             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1042             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1043             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1044
1045             }
1046
1047             /**************************
1048              * CALCULATE INTERACTIONS *
1049              **************************/
1050
1051             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1052             {
1053
1054             /* Compute parameters for interactions between i and j atoms */
1055             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1056
1057             /* REACTION-FIELD ELECTROSTATICS */
1058             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1059
1060             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1061
1062             fscal            = felec;
1063
1064             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1065
1066             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1067
1068             /* Update vectorial force */
1069             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1070             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1071             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1072
1073             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1074             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1075             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1076
1077             }
1078
1079             /**************************
1080              * CALCULATE INTERACTIONS *
1081              **************************/
1082
1083             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1084             {
1085
1086             /* Compute parameters for interactions between i and j atoms */
1087             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1088
1089             /* REACTION-FIELD ELECTROSTATICS */
1090             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1091
1092             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1093
1094             fscal            = felec;
1095
1096             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1097
1098             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1099
1100             /* Update vectorial force */
1101             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1102             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1103             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1104
1105             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1106             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1107             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1108
1109             }
1110
1111             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1112
1113             /* Inner loop uses 135 flops */
1114         }
1115
1116         /* End of innermost loop */
1117
1118         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1119                                               f+i_coord_offset,fshift+i_shift_offset);
1120
1121         /* Increment number of inner iterations */
1122         inneriter                  += j_index_end - j_index_start;
1123
1124         /* Outer loop uses 24 flops */
1125     }
1126
1127     /* Increment number of outer iterations */
1128     outeriter        += nri;
1129
1130     /* Update outer/inner flops */
1131
1132     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*135);
1133 }