src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwjidx0A,vdwjidx0B;
  84     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  85     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  87     real             *charge;
  88     int              nvdwtype;
  89     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  90     int              *vdwtype;
  91     real             *vdwparam;
  92     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  93     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  94     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 114     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 115     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 116     nvdwtype         = fr->ntype;
 117     vdwparam         = fr->nbfp;
 118     vdwtype          = mdatoms->typeA;
 119
 120     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 121     rcutoff_scalar   = fr->rcoulomb;
 122     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 123     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 124
 125     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 126     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 127
 128     /* Avoid stupid compiler warnings */
 129     jnrA = jnrB = 0;
 130     j_coord_offsetA = 0;
 131     j_coord_offsetB = 0;
 132
 133     outeriter        = 0;
 134     inneriter        = 0;
 135
 136     /* Start outer loop over neighborlists */
 137     for(iidx=0; iidx<nri; iidx++)
 138     {
 139         /* Load shift vector for this list */
 140         i_shift_offset   = DIM*shiftidx[iidx];
 141
 142         /* Load limits for loop over neighbors */
 143         j_index_start    = jindex[iidx];
 144         j_index_end      = jindex[iidx+1];
 145
 146         /* Get outer coordinate index */
 147         inr              = iinr[iidx];
 148         i_coord_offset   = DIM*inr;
 149
 150         /* Load i particle coords and add shift vector */
 151         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 152
 153         fix0             = _fjsp_setzero_v2r8();
 154         fiy0             = _fjsp_setzero_v2r8();
 155         fiz0             = _fjsp_setzero_v2r8();
 156
 157         /* Load parameters for i particles */
 158         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 159         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 160
 161         /* Reset potential sums */
 162         velecsum         = _fjsp_setzero_v2r8();
 163         vvdwsum          = _fjsp_setzero_v2r8();
 164
 165         /* Start inner kernel loop */
 166         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 167         {
 168
 169             /* Get j neighbor index, and coordinate index */
 170             jnrA             = jjnr[jidx];
 171             jnrB             = jjnr[jidx+1];
 172             j_coord_offsetA  = DIM*jnrA;
 173             j_coord_offsetB  = DIM*jnrB;
 174
 175             /* load j atom coordinates */
 176             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 177                                               &jx0,&jy0,&jz0);
 178
 179             /* Calculate displacement vector */
 180             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 181             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 182             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 183
 184             /* Calculate squared distance and things based on it */
 185             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 186
 187             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 188
 189             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 190
 191             /* Load parameters for j particles */
 192             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 193             vdwjidx0A        = 2*vdwtype[jnrA+0];
 194             vdwjidx0B        = 2*vdwtype[jnrB+0];
 195
 196             /**************************
 197              * CALCULATE INTERACTIONS *
 198              **************************/
 199
 200             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 201             {
 202
 203             /* Compute parameters for interactions between i and j atoms */
 204             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 205             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 206                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 207
 208             /* REACTION-FIELD ELECTROSTATICS */
 209             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 210             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 211
 212             /* LENNARD-JONES DISPERSION/REPULSION */
 213
 214             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 215             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 216             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 217             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 218                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 219             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 220
 221             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 222
 223             /* Update potential sum for this i atom from the interaction with this j atom. */
 224             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 225             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 226             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 227             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 228
 229             fscal            = _fjsp_add_v2r8(felec,fvdw);
 230
 231             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 232
 233             /* Update vectorial force */
 234             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 235             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 236             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 237
 238             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 239
 240             }
 241
 242             /* Inner loop uses 57 flops */
 243         }
 244
 245         if(jidx<j_index_end)
 246         {
 247
 248             jnrA             = jjnr[jidx];
 249             j_coord_offsetA  = DIM*jnrA;
 250
 251             /* load j atom coordinates */
 252             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 253                                               &jx0,&jy0,&jz0);
 254
 255             /* Calculate displacement vector */
 256             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 257             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 258             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 259
 260             /* Calculate squared distance and things based on it */
 261             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 262
 263             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 264
 265             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 266
 267             /* Load parameters for j particles */
 268             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 269             vdwjidx0A        = 2*vdwtype[jnrA+0];
 270
 271             /**************************
 272              * CALCULATE INTERACTIONS *
 273              **************************/
 274
 275             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 276             {
 277
 278             /* Compute parameters for interactions between i and j atoms */
 279             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 280             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 281
 282             /* REACTION-FIELD ELECTROSTATICS */
 283             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 284             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 285
 286             /* LENNARD-JONES DISPERSION/REPULSION */
 287
 288             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 289             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 290             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 291             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 292                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 293             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 294
 295             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 296
 297             /* Update potential sum for this i atom from the interaction with this j atom. */
 298             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 299             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 300             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 301             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 302             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 303             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 304
 305             fscal            = _fjsp_add_v2r8(felec,fvdw);
 306
 307             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 308
 309             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 310
 311             /* Update vectorial force */
 312             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 313             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 314             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 315
 316             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 317
 318             }
 319
 320             /* Inner loop uses 57 flops */
 321         }
 322
 323         /* End of innermost loop */
 324
 325         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 326                                               f+i_coord_offset,fshift+i_shift_offset);
 327
 328         ggid                        = gid[iidx];
 329         /* Update potential energies */
 330         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 331         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 332
 333         /* Increment number of inner iterations */
 334         inneriter                  += j_index_end - j_index_start;
 335
 336         /* Outer loop uses 9 flops */
 337     }
 338
 339     /* Increment number of outer iterations */
 340     outeriter        += nri;
 341
 342     /* Update outer/inner flops */
 343
 344     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*57);
 345 }
 346 /*
 347  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 348  * Electrostatics interaction: ReactionField
 349  * VdW interaction:            LennardJones
 350  * Geometry:                   Particle-Particle
 351  * Calculate force/pot:        Force
 352  */
 353 void
 354 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 355                     (t_nblist                    * gmx_restrict       nlist,
 356                      rvec                        * gmx_restrict          xx,
 357                      rvec                        * gmx_restrict          ff,
 358                      t_forcerec                  * gmx_restrict          fr,
 359                      t_mdatoms                   * gmx_restrict     mdatoms,
 360                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 361                      t_nrnb                      * gmx_restrict        nrnb)
 362 {
 363     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 364      * just 0 for non-waters.
 365      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 366      * jnr indices corresponding to data put in the four positions in the SIMD register.
 367      */
 368     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 369     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 370     int              jnrA,jnrB;
 371     int              j_coord_offsetA,j_coord_offsetB;
 372     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 373     real             rcutoff_scalar;
 374     real             *shiftvec,*fshift,*x,*f;
 375     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 376     int              vdwioffset0;
 377     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 378     int              vdwjidx0A,vdwjidx0B;
 379     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 380     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 381     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 382     real             *charge;
 383     int              nvdwtype;
 384     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 385     int              *vdwtype;
 386     real             *vdwparam;
 387     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 388     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 389     _fjsp_v2r8       itab_tmp;
 390     _fjsp_v2r8       dummy_mask,cutoff_mask;
 391     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 392     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 393     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 394
 395     x                = xx[0];
 396     f                = ff[0];
 397
 398     nri              = nlist->nri;
 399     iinr             = nlist->iinr;
 400     jindex           = nlist->jindex;
 401     jjnr             = nlist->jjnr;
 402     shiftidx         = nlist->shift;
 403     gid              = nlist->gid;
 404     shiftvec         = fr->shift_vec[0];
 405     fshift           = fr->fshift[0];
 406     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 407     charge           = mdatoms->chargeA;
 408     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 409     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 410     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 411     nvdwtype         = fr->ntype;
 412     vdwparam         = fr->nbfp;
 413     vdwtype          = mdatoms->typeA;
 414
 415     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 416     rcutoff_scalar   = fr->rcoulomb;
 417     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 418     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 419
 420     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 421     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 422
 423     /* Avoid stupid compiler warnings */
 424     jnrA = jnrB = 0;
 425     j_coord_offsetA = 0;
 426     j_coord_offsetB = 0;
 427
 428     outeriter        = 0;
 429     inneriter        = 0;
 430
 431     /* Start outer loop over neighborlists */
 432     for(iidx=0; iidx<nri; iidx++)
 433     {
 434         /* Load shift vector for this list */
 435         i_shift_offset   = DIM*shiftidx[iidx];
 436
 437         /* Load limits for loop over neighbors */
 438         j_index_start    = jindex[iidx];
 439         j_index_end      = jindex[iidx+1];
 440
 441         /* Get outer coordinate index */
 442         inr              = iinr[iidx];
 443         i_coord_offset   = DIM*inr;
 444
 445         /* Load i particle coords and add shift vector */
 446         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 447
 448         fix0             = _fjsp_setzero_v2r8();
 449         fiy0             = _fjsp_setzero_v2r8();
 450         fiz0             = _fjsp_setzero_v2r8();
 451
 452         /* Load parameters for i particles */
 453         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 454         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 455
 456         /* Start inner kernel loop */
 457         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 458         {
 459
 460             /* Get j neighbor index, and coordinate index */
 461             jnrA             = jjnr[jidx];
 462             jnrB             = jjnr[jidx+1];
 463             j_coord_offsetA  = DIM*jnrA;
 464             j_coord_offsetB  = DIM*jnrB;
 465
 466             /* load j atom coordinates */
 467             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 468                                               &jx0,&jy0,&jz0);
 469
 470             /* Calculate displacement vector */
 471             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 472             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 473             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 474
 475             /* Calculate squared distance and things based on it */
 476             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 477
 478             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 479
 480             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 481
 482             /* Load parameters for j particles */
 483             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 484             vdwjidx0A        = 2*vdwtype[jnrA+0];
 485             vdwjidx0B        = 2*vdwtype[jnrB+0];
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 492             {
 493
 494             /* Compute parameters for interactions between i and j atoms */
 495             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 496             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 497                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 498
 499             /* REACTION-FIELD ELECTROSTATICS */
 500             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 501
 502             /* LENNARD-JONES DISPERSION/REPULSION */
 503
 504             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 505             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 506
 507             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 508
 509             fscal            = _fjsp_add_v2r8(felec,fvdw);
 510
 511             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 512
 513             /* Update vectorial force */
 514             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 515             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 516             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 517
 518             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 519
 520             }
 521
 522             /* Inner loop uses 40 flops */
 523         }
 524
 525         if(jidx<j_index_end)
 526         {
 527
 528             jnrA             = jjnr[jidx];
 529             j_coord_offsetA  = DIM*jnrA;
 530
 531             /* load j atom coordinates */
 532             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 533                                               &jx0,&jy0,&jz0);
 534
 535             /* Calculate displacement vector */
 536             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 537             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 538             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 539
 540             /* Calculate squared distance and things based on it */
 541             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 542
 543             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 544
 545             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 546
 547             /* Load parameters for j particles */
 548             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 549             vdwjidx0A        = 2*vdwtype[jnrA+0];
 550
 551             /**************************
 552              * CALCULATE INTERACTIONS *
 553              **************************/
 554
 555             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 556             {
 557
 558             /* Compute parameters for interactions between i and j atoms */
 559             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 560             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 561
 562             /* REACTION-FIELD ELECTROSTATICS */
 563             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 564
 565             /* LENNARD-JONES DISPERSION/REPULSION */
 566
 567             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 568             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 569
 570             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 571
 572             fscal            = _fjsp_add_v2r8(felec,fvdw);
 573
 574             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 575
 576             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 577
 578             /* Update vectorial force */
 579             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 580             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 581             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 582
 583             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 584
 585             }
 586
 587             /* Inner loop uses 40 flops */
 588         }
 589
 590         /* End of innermost loop */
 591
 592         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 593                                               f+i_coord_offset,fshift+i_shift_offset);
 594
 595         /* Increment number of inner iterations */
 596         inneriter                  += j_index_end - j_index_start;
 597
 598         /* Outer loop uses 7 flops */
 599     }
 600
 601     /* Increment number of outer iterations */
 602     outeriter        += nri;
 603
 604     /* Update outer/inner flops */
 605
 606     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*40);
 607 }