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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      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              vdwjidx0A,vdwjidx0B;
  82     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  83     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  84     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  85     real             *charge;
  86     int              nvdwtype;
  87     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  88     int              *vdwtype;
  89     real             *vdwparam;
  90     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  91     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  92     _fjsp_v2r8       itab_tmp;
  93     _fjsp_v2r8       dummy_mask,cutoff_mask;
  94     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  95     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  96     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  97
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 110     charge           = mdatoms->chargeA;
 111     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 112     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 113     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 114     nvdwtype         = fr->ntype;
 115     vdwparam         = fr->nbfp;
 116     vdwtype          = mdatoms->typeA;
 117
 118     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 119     rcutoff_scalar   = fr->rcoulomb;
 120     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 121     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 122
 123     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 124     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 125
 126     /* Avoid stupid compiler warnings */
 127     jnrA = jnrB = 0;
 128     j_coord_offsetA = 0;
 129     j_coord_offsetB = 0;
 130
 131     outeriter        = 0;
 132     inneriter        = 0;
 133
 134     /* Start outer loop over neighborlists */
 135     for(iidx=0; iidx<nri; iidx++)
 136     {
 137         /* Load shift vector for this list */
 138         i_shift_offset   = DIM*shiftidx[iidx];
 139
 140         /* Load limits for loop over neighbors */
 141         j_index_start    = jindex[iidx];
 142         j_index_end      = jindex[iidx+1];
 143
 144         /* Get outer coordinate index */
 145         inr              = iinr[iidx];
 146         i_coord_offset   = DIM*inr;
 147
 148         /* Load i particle coords and add shift vector */
 149         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 150
 151         fix0             = _fjsp_setzero_v2r8();
 152         fiy0             = _fjsp_setzero_v2r8();
 153         fiz0             = _fjsp_setzero_v2r8();
 154
 155         /* Load parameters for i particles */
 156         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 157         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 158
 159         /* Reset potential sums */
 160         velecsum         = _fjsp_setzero_v2r8();
 161         vvdwsum          = _fjsp_setzero_v2r8();
 162
 163         /* Start inner kernel loop */
 164         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 165         {
 166
 167             /* Get j neighbor index, and coordinate index */
 168             jnrA             = jjnr[jidx];
 169             jnrB             = jjnr[jidx+1];
 170             j_coord_offsetA  = DIM*jnrA;
 171             j_coord_offsetB  = DIM*jnrB;
 172
 173             /* load j atom coordinates */
 174             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 175                                               &jx0,&jy0,&jz0);
 176
 177             /* Calculate displacement vector */
 178             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 179             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 180             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 181
 182             /* Calculate squared distance and things based on it */
 183             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 184
 185             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 186
 187             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 188
 189             /* Load parameters for j particles */
 190             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 191             vdwjidx0A        = 2*vdwtype[jnrA+0];
 192             vdwjidx0B        = 2*vdwtype[jnrB+0];
 193
 194             /**************************
 195              * CALCULATE INTERACTIONS *
 196              **************************/
 197
 198             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 199             {
 200
 201             /* Compute parameters for interactions between i and j atoms */
 202             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 203             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 204                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 205
 206             /* REACTION-FIELD ELECTROSTATICS */
 207             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 208             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 209
 210             /* LENNARD-JONES DISPERSION/REPULSION */
 211
 212             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 213             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 214             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 215             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 216                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 217             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 218
 219             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 220
 221             /* Update potential sum for this i atom from the interaction with this j atom. */
 222             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 223             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 224             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 225             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 226
 227             fscal            = _fjsp_add_v2r8(felec,fvdw);
 228
 229             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 230
 231             /* Update vectorial force */
 232             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 233             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 234             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 235
 236             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 237
 238             }
 239
 240             /* Inner loop uses 57 flops */
 241         }
 242
 243         if(jidx<j_index_end)
 244         {
 245
 246             jnrA             = jjnr[jidx];
 247             j_coord_offsetA  = DIM*jnrA;
 248
 249             /* load j atom coordinates */
 250             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 251                                               &jx0,&jy0,&jz0);
 252
 253             /* Calculate displacement vector */
 254             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 255             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 256             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 257
 258             /* Calculate squared distance and things based on it */
 259             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 260
 261             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 262
 263             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 264
 265             /* Load parameters for j particles */
 266             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 267             vdwjidx0A        = 2*vdwtype[jnrA+0];
 268
 269             /**************************
 270              * CALCULATE INTERACTIONS *
 271              **************************/
 272
 273             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 274             {
 275
 276             /* Compute parameters for interactions between i and j atoms */
 277             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 278             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 279
 280             /* REACTION-FIELD ELECTROSTATICS */
 281             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 282             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 283
 284             /* LENNARD-JONES DISPERSION/REPULSION */
 285
 286             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 287             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 288             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 289             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 290                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 291             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 292
 293             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 294
 295             /* Update potential sum for this i atom from the interaction with this j atom. */
 296             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 297             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 298             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 299             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 300             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 301             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 302
 303             fscal            = _fjsp_add_v2r8(felec,fvdw);
 304
 305             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 306
 307             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 308
 309             /* Update vectorial force */
 310             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 311             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 312             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 313
 314             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 315
 316             }
 317
 318             /* Inner loop uses 57 flops */
 319         }
 320
 321         /* End of innermost loop */
 322
 323         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 324                                               f+i_coord_offset,fshift+i_shift_offset);
 325
 326         ggid                        = gid[iidx];
 327         /* Update potential energies */
 328         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 329         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 330
 331         /* Increment number of inner iterations */
 332         inneriter                  += j_index_end - j_index_start;
 333
 334         /* Outer loop uses 9 flops */
 335     }
 336
 337     /* Increment number of outer iterations */
 338     outeriter        += nri;
 339
 340     /* Update outer/inner flops */
 341
 342     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*57);
 343 }
 344 /*
 345  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 346  * Electrostatics interaction: ReactionField
 347  * VdW interaction:            LennardJones
 348  * Geometry:                   Particle-Particle
 349  * Calculate force/pot:        Force
 350  */
 351 void
 352 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 353                     (t_nblist                    * gmx_restrict       nlist,
 354                      rvec                        * gmx_restrict          xx,
 355                      rvec                        * gmx_restrict          ff,
 356                      t_forcerec                  * gmx_restrict          fr,
 357                      t_mdatoms                   * gmx_restrict     mdatoms,
 358                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 359                      t_nrnb                      * gmx_restrict        nrnb)
 360 {
 361     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 362      * just 0 for non-waters.
 363      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 364      * jnr indices corresponding to data put in the four positions in the SIMD register.
 365      */
 366     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 367     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 368     int              jnrA,jnrB;
 369     int              j_coord_offsetA,j_coord_offsetB;
 370     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 371     real             rcutoff_scalar;
 372     real             *shiftvec,*fshift,*x,*f;
 373     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 374     int              vdwioffset0;
 375     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 376     int              vdwjidx0A,vdwjidx0B;
 377     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 378     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 379     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 380     real             *charge;
 381     int              nvdwtype;
 382     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 383     int              *vdwtype;
 384     real             *vdwparam;
 385     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 386     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 387     _fjsp_v2r8       itab_tmp;
 388     _fjsp_v2r8       dummy_mask,cutoff_mask;
 389     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 390     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 391     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 392
 393     x                = xx[0];
 394     f                = ff[0];
 395
 396     nri              = nlist->nri;
 397     iinr             = nlist->iinr;
 398     jindex           = nlist->jindex;
 399     jjnr             = nlist->jjnr;
 400     shiftidx         = nlist->shift;
 401     gid              = nlist->gid;
 402     shiftvec         = fr->shift_vec[0];
 403     fshift           = fr->fshift[0];
 404     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 405     charge           = mdatoms->chargeA;
 406     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 407     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 408     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 409     nvdwtype         = fr->ntype;
 410     vdwparam         = fr->nbfp;
 411     vdwtype          = mdatoms->typeA;
 412
 413     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 414     rcutoff_scalar   = fr->rcoulomb;
 415     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 416     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 417
 418     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 419     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 420
 421     /* Avoid stupid compiler warnings */
 422     jnrA = jnrB = 0;
 423     j_coord_offsetA = 0;
 424     j_coord_offsetB = 0;
 425
 426     outeriter        = 0;
 427     inneriter        = 0;
 428
 429     /* Start outer loop over neighborlists */
 430     for(iidx=0; iidx<nri; iidx++)
 431     {
 432         /* Load shift vector for this list */
 433         i_shift_offset   = DIM*shiftidx[iidx];
 434
 435         /* Load limits for loop over neighbors */
 436         j_index_start    = jindex[iidx];
 437         j_index_end      = jindex[iidx+1];
 438
 439         /* Get outer coordinate index */
 440         inr              = iinr[iidx];
 441         i_coord_offset   = DIM*inr;
 442
 443         /* Load i particle coords and add shift vector */
 444         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 445
 446         fix0             = _fjsp_setzero_v2r8();
 447         fiy0             = _fjsp_setzero_v2r8();
 448         fiz0             = _fjsp_setzero_v2r8();
 449
 450         /* Load parameters for i particles */
 451         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 452         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 453
 454         /* Start inner kernel loop */
 455         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 456         {
 457
 458             /* Get j neighbor index, and coordinate index */
 459             jnrA             = jjnr[jidx];
 460             jnrB             = jjnr[jidx+1];
 461             j_coord_offsetA  = DIM*jnrA;
 462             j_coord_offsetB  = DIM*jnrB;
 463
 464             /* load j atom coordinates */
 465             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 466                                               &jx0,&jy0,&jz0);
 467
 468             /* Calculate displacement vector */
 469             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 470             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 471             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 472
 473             /* Calculate squared distance and things based on it */
 474             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 475
 476             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 477
 478             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 479
 480             /* Load parameters for j particles */
 481             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 482             vdwjidx0A        = 2*vdwtype[jnrA+0];
 483             vdwjidx0B        = 2*vdwtype[jnrB+0];
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 490             {
 491
 492             /* Compute parameters for interactions between i and j atoms */
 493             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 494             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 495                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 496
 497             /* REACTION-FIELD ELECTROSTATICS */
 498             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 499
 500             /* LENNARD-JONES DISPERSION/REPULSION */
 501
 502             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 503             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 504
 505             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 506
 507             fscal            = _fjsp_add_v2r8(felec,fvdw);
 508
 509             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 510
 511             /* Update vectorial force */
 512             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 513             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 514             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 515
 516             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 517
 518             }
 519
 520             /* Inner loop uses 40 flops */
 521         }
 522
 523         if(jidx<j_index_end)
 524         {
 525
 526             jnrA             = jjnr[jidx];
 527             j_coord_offsetA  = DIM*jnrA;
 528
 529             /* load j atom coordinates */
 530             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 531                                               &jx0,&jy0,&jz0);
 532
 533             /* Calculate displacement vector */
 534             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 535             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 536             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 537
 538             /* Calculate squared distance and things based on it */
 539             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 540
 541             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 542
 543             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 544
 545             /* Load parameters for j particles */
 546             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 547             vdwjidx0A        = 2*vdwtype[jnrA+0];
 548
 549             /**************************
 550              * CALCULATE INTERACTIONS *
 551              **************************/
 552
 553             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 554             {
 555
 556             /* Compute parameters for interactions between i and j atoms */
 557             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 558             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 559
 560             /* REACTION-FIELD ELECTROSTATICS */
 561             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 562
 563             /* LENNARD-JONES DISPERSION/REPULSION */
 564
 565             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 566             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 567
 568             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 569
 570             fscal            = _fjsp_add_v2r8(felec,fvdw);
 571
 572             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 573
 574             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 575
 576             /* Update vectorial force */
 577             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 578             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 579             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 580
 581             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 582
 583             }
 584
 585             /* Inner loop uses 40 flops */
 586         }
 587
 588         /* End of innermost loop */
 589
 590         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 591                                               f+i_coord_offset,fshift+i_shift_offset);
 592
 593         /* Increment number of inner iterations */
 594         inneriter                  += j_index_end - j_index_start;
 595
 596         /* Outer loop uses 7 flops */
 597     }
 598
 599     /* Increment number of outer iterations */
 600     outeriter        += nri;
 601
 602     /* Update outer/inner flops */
 603
 604     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*40);
 605 }