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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_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                      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              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->ic->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->ic->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->ic->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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 279                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 280
 281             /* REACTION-FIELD ELECTROSTATICS */
 282             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 283             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 284
 285             /* LENNARD-JONES DISPERSION/REPULSION */
 286
 287             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 288             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 289             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 290             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 291                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 292             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 293
 294             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 295
 296             /* Update potential sum for this i atom from the interaction with this j atom. */
 297             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 298             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 299             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 300             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 301             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 302             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 303
 304             fscal            = _fjsp_add_v2r8(felec,fvdw);
 305
 306             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 307
 308             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 309
 310             /* Update vectorial force */
 311             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 312             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 313             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 314
 315             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 316
 317             }
 318
 319             /* Inner loop uses 57 flops */
 320         }
 321
 322         /* End of innermost loop */
 323
 324         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 325                                               f+i_coord_offset,fshift+i_shift_offset);
 326
 327         ggid                        = gid[iidx];
 328         /* Update potential energies */
 329         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 330         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 331
 332         /* Increment number of inner iterations */
 333         inneriter                  += j_index_end - j_index_start;
 334
 335         /* Outer loop uses 9 flops */
 336     }
 337
 338     /* Increment number of outer iterations */
 339     outeriter        += nri;
 340
 341     /* Update outer/inner flops */
 342
 343     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*57);
 344 }
 345 /*
 346  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 347  * Electrostatics interaction: ReactionField
 348  * VdW interaction:            LennardJones
 349  * Geometry:                   Particle-Particle
 350  * Calculate force/pot:        Force
 351  */
 352 void
 353 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sparc64_hpc_ace_double
 354                     (t_nblist                    * gmx_restrict       nlist,
 355                      rvec                        * gmx_restrict          xx,
 356                      rvec                        * gmx_restrict          ff,
 357                      struct t_forcerec           * gmx_restrict          fr,
 358                      t_mdatoms                   * gmx_restrict     mdatoms,
 359                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 360                      t_nrnb                      * gmx_restrict        nrnb)
 361 {
 362     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 363      * just 0 for non-waters.
 364      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 365      * jnr indices corresponding to data put in the four positions in the SIMD register.
 366      */
 367     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 368     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 369     int              jnrA,jnrB;
 370     int              j_coord_offsetA,j_coord_offsetB;
 371     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 372     real             rcutoff_scalar;
 373     real             *shiftvec,*fshift,*x,*f;
 374     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 375     int              vdwioffset0;
 376     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 377     int              vdwjidx0A,vdwjidx0B;
 378     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 379     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 380     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 381     real             *charge;
 382     int              nvdwtype;
 383     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 384     int              *vdwtype;
 385     real             *vdwparam;
 386     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 387     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 388     _fjsp_v2r8       itab_tmp;
 389     _fjsp_v2r8       dummy_mask,cutoff_mask;
 390     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 391     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 392     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 393
 394     x                = xx[0];
 395     f                = ff[0];
 396
 397     nri              = nlist->nri;
 398     iinr             = nlist->iinr;
 399     jindex           = nlist->jindex;
 400     jjnr             = nlist->jjnr;
 401     shiftidx         = nlist->shift;
 402     gid              = nlist->gid;
 403     shiftvec         = fr->shift_vec[0];
 404     fshift           = fr->fshift[0];
 405     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 406     charge           = mdatoms->chargeA;
 407     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 408     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 409     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 410     nvdwtype         = fr->ntype;
 411     vdwparam         = fr->nbfp;
 412     vdwtype          = mdatoms->typeA;
 413
 414     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 415     rcutoff_scalar   = fr->ic->rcoulomb;
 416     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 417     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 418
 419     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 420     rvdw             = gmx_fjsp_set1_v2r8(fr->ic->rvdw);
 421
 422     /* Avoid stupid compiler warnings */
 423     jnrA = jnrB = 0;
 424     j_coord_offsetA = 0;
 425     j_coord_offsetB = 0;
 426
 427     outeriter        = 0;
 428     inneriter        = 0;
 429
 430     /* Start outer loop over neighborlists */
 431     for(iidx=0; iidx<nri; iidx++)
 432     {
 433         /* Load shift vector for this list */
 434         i_shift_offset   = DIM*shiftidx[iidx];
 435
 436         /* Load limits for loop over neighbors */
 437         j_index_start    = jindex[iidx];
 438         j_index_end      = jindex[iidx+1];
 439
 440         /* Get outer coordinate index */
 441         inr              = iinr[iidx];
 442         i_coord_offset   = DIM*inr;
 443
 444         /* Load i particle coords and add shift vector */
 445         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 446
 447         fix0             = _fjsp_setzero_v2r8();
 448         fiy0             = _fjsp_setzero_v2r8();
 449         fiz0             = _fjsp_setzero_v2r8();
 450
 451         /* Load parameters for i particles */
 452         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 453         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 454
 455         /* Start inner kernel loop */
 456         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 457         {
 458
 459             /* Get j neighbor index, and coordinate index */
 460             jnrA             = jjnr[jidx];
 461             jnrB             = jjnr[jidx+1];
 462             j_coord_offsetA  = DIM*jnrA;
 463             j_coord_offsetB  = DIM*jnrB;
 464
 465             /* load j atom coordinates */
 466             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 467                                               &jx0,&jy0,&jz0);
 468
 469             /* Calculate displacement vector */
 470             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 471             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 472             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 473
 474             /* Calculate squared distance and things based on it */
 475             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 476
 477             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 478
 479             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 480
 481             /* Load parameters for j particles */
 482             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 483             vdwjidx0A        = 2*vdwtype[jnrA+0];
 484             vdwjidx0B        = 2*vdwtype[jnrB+0];
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 491             {
 492
 493             /* Compute parameters for interactions between i and j atoms */
 494             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 495             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 496                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 497
 498             /* REACTION-FIELD ELECTROSTATICS */
 499             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 500
 501             /* LENNARD-JONES DISPERSION/REPULSION */
 502
 503             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 504             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 505
 506             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 507
 508             fscal            = _fjsp_add_v2r8(felec,fvdw);
 509
 510             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 511
 512             /* Update vectorial force */
 513             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 514             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 515             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 516
 517             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 518
 519             }
 520
 521             /* Inner loop uses 40 flops */
 522         }
 523
 524         if(jidx<j_index_end)
 525         {
 526
 527             jnrA             = jjnr[jidx];
 528             j_coord_offsetA  = DIM*jnrA;
 529
 530             /* load j atom coordinates */
 531             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 532                                               &jx0,&jy0,&jz0);
 533
 534             /* Calculate displacement vector */
 535             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 536             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 537             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 538
 539             /* Calculate squared distance and things based on it */
 540             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 541
 542             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 543
 544             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 545
 546             /* Load parameters for j particles */
 547             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 548             vdwjidx0A        = 2*vdwtype[jnrA+0];
 549
 550             /**************************
 551              * CALCULATE INTERACTIONS *
 552              **************************/
 553
 554             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 555             {
 556
 557             /* Compute parameters for interactions between i and j atoms */
 558             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 559             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 560                                          vdwparam+vdwioffset0+vdwjidx0B,&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 }