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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 281                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 282
 283             /* REACTION-FIELD ELECTROSTATICS */
 284             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 285             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 286
 287             /* LENNARD-JONES DISPERSION/REPULSION */
 288
 289             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 290             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 291             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 292             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 293                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 294             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 295
 296             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 297
 298             /* Update potential sum for this i atom from the interaction with this j atom. */
 299             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 300             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 301             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 302             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 303             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 304             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 305
 306             fscal            = _fjsp_add_v2r8(felec,fvdw);
 307
 308             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 309
 310             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 311
 312             /* Update vectorial force */
 313             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 314             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 315             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 316
 317             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 318
 319             }
 320
 321             /* Inner loop uses 57 flops */
 322         }
 323
 324         /* End of innermost loop */
 325
 326         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 327                                               f+i_coord_offset,fshift+i_shift_offset);
 328
 329         ggid                        = gid[iidx];
 330         /* Update potential energies */
 331         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 332         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 333
 334         /* Increment number of inner iterations */
 335         inneriter                  += j_index_end - j_index_start;
 336
 337         /* Outer loop uses 9 flops */
 338     }
 339
 340     /* Increment number of outer iterations */
 341     outeriter        += nri;
 342
 343     /* Update outer/inner flops */
 344
 345     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*57);
 346 }
 347 /*
 348  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 349  * Electrostatics interaction: ReactionField
 350  * VdW interaction:            LennardJones
 351  * Geometry:                   Particle-Particle
 352  * Calculate force/pot:        Force
 353  */
 354 void
 355 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 356                     (t_nblist                    * gmx_restrict       nlist,
 357                      rvec                        * gmx_restrict          xx,
 358                      rvec                        * gmx_restrict          ff,
 359                      t_forcerec                  * gmx_restrict          fr,
 360                      t_mdatoms                   * gmx_restrict     mdatoms,
 361                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 362                      t_nrnb                      * gmx_restrict        nrnb)
 363 {
 364     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 365      * just 0 for non-waters.
 366      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 367      * jnr indices corresponding to data put in the four positions in the SIMD register.
 368      */
 369     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 370     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 371     int              jnrA,jnrB;
 372     int              j_coord_offsetA,j_coord_offsetB;
 373     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 374     real             rcutoff_scalar;
 375     real             *shiftvec,*fshift,*x,*f;
 376     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 377     int              vdwioffset0;
 378     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 379     int              vdwjidx0A,vdwjidx0B;
 380     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 381     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 382     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 383     real             *charge;
 384     int              nvdwtype;
 385     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 386     int              *vdwtype;
 387     real             *vdwparam;
 388     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 389     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 390     _fjsp_v2r8       itab_tmp;
 391     _fjsp_v2r8       dummy_mask,cutoff_mask;
 392     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 393     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 394     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 395
 396     x                = xx[0];
 397     f                = ff[0];
 398
 399     nri              = nlist->nri;
 400     iinr             = nlist->iinr;
 401     jindex           = nlist->jindex;
 402     jjnr             = nlist->jjnr;
 403     shiftidx         = nlist->shift;
 404     gid              = nlist->gid;
 405     shiftvec         = fr->shift_vec[0];
 406     fshift           = fr->fshift[0];
 407     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 408     charge           = mdatoms->chargeA;
 409     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 410     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 411     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 412     nvdwtype         = fr->ntype;
 413     vdwparam         = fr->nbfp;
 414     vdwtype          = mdatoms->typeA;
 415
 416     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 417     rcutoff_scalar   = fr->rcoulomb;
 418     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 419     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 420
 421     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 422     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 423
 424     /* Avoid stupid compiler warnings */
 425     jnrA = jnrB = 0;
 426     j_coord_offsetA = 0;
 427     j_coord_offsetB = 0;
 428
 429     outeriter        = 0;
 430     inneriter        = 0;
 431
 432     /* Start outer loop over neighborlists */
 433     for(iidx=0; iidx<nri; iidx++)
 434     {
 435         /* Load shift vector for this list */
 436         i_shift_offset   = DIM*shiftidx[iidx];
 437
 438         /* Load limits for loop over neighbors */
 439         j_index_start    = jindex[iidx];
 440         j_index_end      = jindex[iidx+1];
 441
 442         /* Get outer coordinate index */
 443         inr              = iinr[iidx];
 444         i_coord_offset   = DIM*inr;
 445
 446         /* Load i particle coords and add shift vector */
 447         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 448
 449         fix0             = _fjsp_setzero_v2r8();
 450         fiy0             = _fjsp_setzero_v2r8();
 451         fiz0             = _fjsp_setzero_v2r8();
 452
 453         /* Load parameters for i particles */
 454         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 455         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 456
 457         /* Start inner kernel loop */
 458         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 459         {
 460
 461             /* Get j neighbor index, and coordinate index */
 462             jnrA             = jjnr[jidx];
 463             jnrB             = jjnr[jidx+1];
 464             j_coord_offsetA  = DIM*jnrA;
 465             j_coord_offsetB  = DIM*jnrB;
 466
 467             /* load j atom coordinates */
 468             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 469                                               &jx0,&jy0,&jz0);
 470
 471             /* Calculate displacement vector */
 472             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 473             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 474             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 475
 476             /* Calculate squared distance and things based on it */
 477             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 478
 479             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 480
 481             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 482
 483             /* Load parameters for j particles */
 484             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 485             vdwjidx0A        = 2*vdwtype[jnrA+0];
 486             vdwjidx0B        = 2*vdwtype[jnrB+0];
 487
 488             /**************************
 489              * CALCULATE INTERACTIONS *
 490              **************************/
 491
 492             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 493             {
 494
 495             /* Compute parameters for interactions between i and j atoms */
 496             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 497             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 498                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 499
 500             /* REACTION-FIELD ELECTROSTATICS */
 501             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 502
 503             /* LENNARD-JONES DISPERSION/REPULSION */
 504
 505             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 506             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 507
 508             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 509
 510             fscal            = _fjsp_add_v2r8(felec,fvdw);
 511
 512             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 513
 514             /* Update vectorial force */
 515             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 516             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 517             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 518
 519             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 520
 521             }
 522
 523             /* Inner loop uses 40 flops */
 524         }
 525
 526         if(jidx<j_index_end)
 527         {
 528
 529             jnrA             = jjnr[jidx];
 530             j_coord_offsetA  = DIM*jnrA;
 531
 532             /* load j atom coordinates */
 533             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 534                                               &jx0,&jy0,&jz0);
 535
 536             /* Calculate displacement vector */
 537             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 538             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 539             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 540
 541             /* Calculate squared distance and things based on it */
 542             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 543
 544             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 545
 546             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 547
 548             /* Load parameters for j particles */
 549             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 550             vdwjidx0A        = 2*vdwtype[jnrA+0];
 551
 552             /**************************
 553              * CALCULATE INTERACTIONS *
 554              **************************/
 555
 556             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 557             {
 558
 559             /* Compute parameters for interactions between i and j atoms */
 560             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 561             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 562                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 563
 564             /* REACTION-FIELD ELECTROSTATICS */
 565             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 566
 567             /* LENNARD-JONES DISPERSION/REPULSION */
 568
 569             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 570             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 571
 572             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 573
 574             fscal            = _fjsp_add_v2r8(felec,fvdw);
 575
 576             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 577
 578             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 579
 580             /* Update vectorial force */
 581             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 582             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 583             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 584
 585             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 586
 587             }
 588
 589             /* Inner loop uses 40 flops */
 590         }
 591
 592         /* End of innermost loop */
 593
 594         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 595                                               f+i_coord_offset,fshift+i_shift_offset);
 596
 597         /* Increment number of inner iterations */
 598         inneriter                  += j_index_end - j_index_start;
 599
 600         /* Outer loop uses 7 flops */
 601     }
 602
 603     /* Increment number of outer iterations */
 604     outeriter        += nri;
 605
 606     /* Update outer/inner flops */
 607
 608     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*40);
 609 }