src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_VdwLJ_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_ElecEw_VdwLJ_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEw_VdwLJ_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       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  95     real             *ewtab;
  96     _fjsp_v2r8       itab_tmp;
  97     _fjsp_v2r8       dummy_mask,cutoff_mask;
  98     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  99     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 100     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 101
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 114     charge           = mdatoms->chargeA;
 115     nvdwtype         = fr->ntype;
 116     vdwparam         = fr->nbfp;
 117     vdwtype          = mdatoms->typeA;
 118
 119     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 120     ewtab            = fr->ic->tabq_coul_FDV0;
 121     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 122     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 123
 124     /* Avoid stupid compiler warnings */
 125     jnrA = jnrB = 0;
 126     j_coord_offsetA = 0;
 127     j_coord_offsetB = 0;
 128
 129     outeriter        = 0;
 130     inneriter        = 0;
 131
 132     /* Start outer loop over neighborlists */
 133     for(iidx=0; iidx<nri; iidx++)
 134     {
 135         /* Load shift vector for this list */
 136         i_shift_offset   = DIM*shiftidx[iidx];
 137
 138         /* Load limits for loop over neighbors */
 139         j_index_start    = jindex[iidx];
 140         j_index_end      = jindex[iidx+1];
 141
 142         /* Get outer coordinate index */
 143         inr              = iinr[iidx];
 144         i_coord_offset   = DIM*inr;
 145
 146         /* Load i particle coords and add shift vector */
 147         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 148
 149         fix0             = _fjsp_setzero_v2r8();
 150         fiy0             = _fjsp_setzero_v2r8();
 151         fiz0             = _fjsp_setzero_v2r8();
 152
 153         /* Load parameters for i particles */
 154         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 155         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 156
 157         /* Reset potential sums */
 158         velecsum         = _fjsp_setzero_v2r8();
 159         vvdwsum          = _fjsp_setzero_v2r8();
 160
 161         /* Start inner kernel loop */
 162         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 163         {
 164
 165             /* Get j neighbor index, and coordinate index */
 166             jnrA             = jjnr[jidx];
 167             jnrB             = jjnr[jidx+1];
 168             j_coord_offsetA  = DIM*jnrA;
 169             j_coord_offsetB  = DIM*jnrB;
 170
 171             /* load j atom coordinates */
 172             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 173                                               &jx0,&jy0,&jz0);
 174
 175             /* Calculate displacement vector */
 176             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 177             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 178             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 179
 180             /* Calculate squared distance and things based on it */
 181             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 182
 183             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 184
 185             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 186
 187             /* Load parameters for j particles */
 188             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 189             vdwjidx0A        = 2*vdwtype[jnrA+0];
 190             vdwjidx0B        = 2*vdwtype[jnrB+0];
 191
 192             /**************************
 193              * CALCULATE INTERACTIONS *
 194              **************************/
 195
 196             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 197
 198             /* Compute parameters for interactions between i and j atoms */
 199             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 200             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 201                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 202
 203             /* EWALD ELECTROSTATICS */
 204
 205             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 206             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 207             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 208             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 209             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 210
 211             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 212             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 213             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 214             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 215             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 216             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 217             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 218             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 219             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 220             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 221
 222             /* LENNARD-JONES DISPERSION/REPULSION */
 223
 224             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 225             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 226             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 227             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 228             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 229
 230             /* Update potential sum for this i atom from the interaction with this j atom. */
 231             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 232             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 233
 234             fscal            = _fjsp_add_v2r8(felec,fvdw);
 235
 236             /* Update vectorial force */
 237             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 238             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 239             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 240
 241             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 242
 243             /* Inner loop uses 56 flops */
 244         }
 245
 246         if(jidx<j_index_end)
 247         {
 248
 249             jnrA             = jjnr[jidx];
 250             j_coord_offsetA  = DIM*jnrA;
 251
 252             /* load j atom coordinates */
 253             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 254                                               &jx0,&jy0,&jz0);
 255
 256             /* Calculate displacement vector */
 257             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 258             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 259             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 260
 261             /* Calculate squared distance and things based on it */
 262             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 263
 264             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 265
 266             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 267
 268             /* Load parameters for j particles */
 269             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 270             vdwjidx0A        = 2*vdwtype[jnrA+0];
 271
 272             /**************************
 273              * CALCULATE INTERACTIONS *
 274              **************************/
 275
 276             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 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             /* EWALD ELECTROSTATICS */
 284
 285             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 286             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 287             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 288             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 289             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 290
 291             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 292             ewtabD           = _fjsp_setzero_v2r8();
 293             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 294             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 295             ewtabFn          = _fjsp_setzero_v2r8();
 296             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 297             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 298             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 299             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 300             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 301
 302             /* LENNARD-JONES DISPERSION/REPULSION */
 303
 304             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 305             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 306             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 307             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 308             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 309
 310             /* Update potential sum for this i atom from the interaction with this j atom. */
 311             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 312             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 313             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 314             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 315
 316             fscal            = _fjsp_add_v2r8(felec,fvdw);
 317
 318             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 319
 320             /* Update vectorial force */
 321             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 322             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 323             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 324
 325             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 326
 327             /* Inner loop uses 56 flops */
 328         }
 329
 330         /* End of innermost loop */
 331
 332         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 333                                               f+i_coord_offset,fshift+i_shift_offset);
 334
 335         ggid                        = gid[iidx];
 336         /* Update potential energies */
 337         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 338         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 339
 340         /* Increment number of inner iterations */
 341         inneriter                  += j_index_end - j_index_start;
 342
 343         /* Outer loop uses 9 flops */
 344     }
 345
 346     /* Increment number of outer iterations */
 347     outeriter        += nri;
 348
 349     /* Update outer/inner flops */
 350
 351     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*56);
 352 }
 353 /*
 354  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 355  * Electrostatics interaction: Ewald
 356  * VdW interaction:            LennardJones
 357  * Geometry:                   Particle-Particle
 358  * Calculate force/pot:        Force
 359  */
 360 void
 361 nb_kernel_ElecEw_VdwLJ_GeomP1P1_F_sparc64_hpc_ace_double
 362                     (t_nblist                    * gmx_restrict       nlist,
 363                      rvec                        * gmx_restrict          xx,
 364                      rvec                        * gmx_restrict          ff,
 365                      t_forcerec                  * gmx_restrict          fr,
 366                      t_mdatoms                   * gmx_restrict     mdatoms,
 367                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 368                      t_nrnb                      * gmx_restrict        nrnb)
 369 {
 370     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 371      * just 0 for non-waters.
 372      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 373      * jnr indices corresponding to data put in the four positions in the SIMD register.
 374      */
 375     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 376     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 377     int              jnrA,jnrB;
 378     int              j_coord_offsetA,j_coord_offsetB;
 379     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 380     real             rcutoff_scalar;
 381     real             *shiftvec,*fshift,*x,*f;
 382     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 383     int              vdwioffset0;
 384     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 385     int              vdwjidx0A,vdwjidx0B;
 386     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 387     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 388     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 389     real             *charge;
 390     int              nvdwtype;
 391     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 392     int              *vdwtype;
 393     real             *vdwparam;
 394     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 395     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 396     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 397     real             *ewtab;
 398     _fjsp_v2r8       itab_tmp;
 399     _fjsp_v2r8       dummy_mask,cutoff_mask;
 400     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 401     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 402     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 403
 404     x                = xx[0];
 405     f                = ff[0];
 406
 407     nri              = nlist->nri;
 408     iinr             = nlist->iinr;
 409     jindex           = nlist->jindex;
 410     jjnr             = nlist->jjnr;
 411     shiftidx         = nlist->shift;
 412     gid              = nlist->gid;
 413     shiftvec         = fr->shift_vec[0];
 414     fshift           = fr->fshift[0];
 415     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 416     charge           = mdatoms->chargeA;
 417     nvdwtype         = fr->ntype;
 418     vdwparam         = fr->nbfp;
 419     vdwtype          = mdatoms->typeA;
 420
 421     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 422     ewtab            = fr->ic->tabq_coul_F;
 423     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 424     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 425
 426     /* Avoid stupid compiler warnings */
 427     jnrA = jnrB = 0;
 428     j_coord_offsetA = 0;
 429     j_coord_offsetB = 0;
 430
 431     outeriter        = 0;
 432     inneriter        = 0;
 433
 434     /* Start outer loop over neighborlists */
 435     for(iidx=0; iidx<nri; iidx++)
 436     {
 437         /* Load shift vector for this list */
 438         i_shift_offset   = DIM*shiftidx[iidx];
 439
 440         /* Load limits for loop over neighbors */
 441         j_index_start    = jindex[iidx];
 442         j_index_end      = jindex[iidx+1];
 443
 444         /* Get outer coordinate index */
 445         inr              = iinr[iidx];
 446         i_coord_offset   = DIM*inr;
 447
 448         /* Load i particle coords and add shift vector */
 449         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 450
 451         fix0             = _fjsp_setzero_v2r8();
 452         fiy0             = _fjsp_setzero_v2r8();
 453         fiz0             = _fjsp_setzero_v2r8();
 454
 455         /* Load parameters for i particles */
 456         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 457         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 458
 459         /* Start inner kernel loop */
 460         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 461         {
 462
 463             /* Get j neighbor index, and coordinate index */
 464             jnrA             = jjnr[jidx];
 465             jnrB             = jjnr[jidx+1];
 466             j_coord_offsetA  = DIM*jnrA;
 467             j_coord_offsetB  = DIM*jnrB;
 468
 469             /* load j atom coordinates */
 470             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 471                                               &jx0,&jy0,&jz0);
 472
 473             /* Calculate displacement vector */
 474             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 475             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 476             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 477
 478             /* Calculate squared distance and things based on it */
 479             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 480
 481             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 482
 483             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 484
 485             /* Load parameters for j particles */
 486             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 487             vdwjidx0A        = 2*vdwtype[jnrA+0];
 488             vdwjidx0B        = 2*vdwtype[jnrB+0];
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 495
 496             /* Compute parameters for interactions between i and j atoms */
 497             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 498             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 499                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 500
 501             /* EWALD ELECTROSTATICS */
 502
 503             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 504             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 505             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 506             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 507             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 508
 509             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 510                                          &ewtabF,&ewtabFn);
 511             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 512             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 513
 514             /* LENNARD-JONES DISPERSION/REPULSION */
 515
 516             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 517             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 518
 519             fscal            = _fjsp_add_v2r8(felec,fvdw);
 520
 521             /* Update vectorial force */
 522             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 523             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 524             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 525
 526             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 527
 528             /* Inner loop uses 46 flops */
 529         }
 530
 531         if(jidx<j_index_end)
 532         {
 533
 534             jnrA             = jjnr[jidx];
 535             j_coord_offsetA  = DIM*jnrA;
 536
 537             /* load j atom coordinates */
 538             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 539                                               &jx0,&jy0,&jz0);
 540
 541             /* Calculate displacement vector */
 542             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 543             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 544             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 545
 546             /* Calculate squared distance and things based on it */
 547             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 548
 549             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 550
 551             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 552
 553             /* Load parameters for j particles */
 554             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 555             vdwjidx0A        = 2*vdwtype[jnrA+0];
 556
 557             /**************************
 558              * CALCULATE INTERACTIONS *
 559              **************************/
 560
 561             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 562
 563             /* Compute parameters for interactions between i and j atoms */
 564             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 565             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 566                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 567
 568             /* EWALD ELECTROSTATICS */
 569
 570             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 571             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 572             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 573             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 574             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 575
 576             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 577             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 578             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 579
 580             /* LENNARD-JONES DISPERSION/REPULSION */
 581
 582             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 583             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 584
 585             fscal            = _fjsp_add_v2r8(felec,fvdw);
 586
 587             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 588
 589             /* Update vectorial force */
 590             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 591             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 592             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 593
 594             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 595
 596             /* Inner loop uses 46 flops */
 597         }
 598
 599         /* End of innermost loop */
 600
 601         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 602                                               f+i_coord_offset,fshift+i_shift_offset);
 603
 604         /* Increment number of inner iterations */
 605         inneriter                  += j_index_end - j_index_start;
 606
 607         /* Outer loop uses 7 flops */
 608     }
 609
 610     /* Increment number of outer iterations */
 611     outeriter        += nri;
 612
 613     /* Update outer/inner flops */
 614
 615     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*46);
 616 }