src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecGB_VdwCSTab_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_ElecGB_VdwCSTab_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: GeneralizedBorn
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecGB_VdwCSTab_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     _fjsp_v2r8       vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
  89     _fjsp_v2r8       minushalf = gmx_fjsp_set1_v2r8(-0.5);
  90     real             *invsqrta,*dvda,*gbtab;
  91     int              nvdwtype;
  92     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  93     int              *vdwtype;
  94     real             *vdwparam;
  95     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  96     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  97     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  98     real             *vftab;
  99     _fjsp_v2r8       itab_tmp;
 100     _fjsp_v2r8       dummy_mask,cutoff_mask;
 101     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 102     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 103     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 104
 105     x                = xx[0];
 106     f                = ff[0];
 107
 108     nri              = nlist->nri;
 109     iinr             = nlist->iinr;
 110     jindex           = nlist->jindex;
 111     jjnr             = nlist->jjnr;
 112     shiftidx         = nlist->shift;
 113     gid              = nlist->gid;
 114     shiftvec         = fr->shift_vec[0];
 115     fshift           = fr->fshift[0];
 116     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 117     charge           = mdatoms->chargeA;
 118     nvdwtype         = fr->ntype;
 119     vdwparam         = fr->nbfp;
 120     vdwtype          = mdatoms->typeA;
 121
 122     vftab            = kernel_data->table_vdw->data;
 123     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 124
 125     invsqrta         = fr->invsqrta;
 126     dvda             = fr->dvda;
 127     gbtabscale       = gmx_fjsp_set1_v2r8(fr->gbtab.scale);
 128     gbtab            = fr->gbtab.data;
 129     gbinvepsdiff     = gmx_fjsp_set1_v2r8((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 130
 131     /* Avoid stupid compiler warnings */
 132     jnrA = jnrB = 0;
 133     j_coord_offsetA = 0;
 134     j_coord_offsetB = 0;
 135
 136     outeriter        = 0;
 137     inneriter        = 0;
 138
 139     /* Start outer loop over neighborlists */
 140     for(iidx=0; iidx<nri; iidx++)
 141     {
 142         /* Load shift vector for this list */
 143         i_shift_offset   = DIM*shiftidx[iidx];
 144
 145         /* Load limits for loop over neighbors */
 146         j_index_start    = jindex[iidx];
 147         j_index_end      = jindex[iidx+1];
 148
 149         /* Get outer coordinate index */
 150         inr              = iinr[iidx];
 151         i_coord_offset   = DIM*inr;
 152
 153         /* Load i particle coords and add shift vector */
 154         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 155
 156         fix0             = _fjsp_setzero_v2r8();
 157         fiy0             = _fjsp_setzero_v2r8();
 158         fiz0             = _fjsp_setzero_v2r8();
 159
 160         /* Load parameters for i particles */
 161         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 162         isai0            = gmx_fjsp_load1_v2r8(invsqrta+inr+0);
 163         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 164
 165         /* Reset potential sums */
 166         velecsum         = _fjsp_setzero_v2r8();
 167         vgbsum           = _fjsp_setzero_v2r8();
 168         vvdwsum          = _fjsp_setzero_v2r8();
 169         dvdasum          = _fjsp_setzero_v2r8();
 170
 171         /* Start inner kernel loop */
 172         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 173         {
 174
 175             /* Get j neighbor index, and coordinate index */
 176             jnrA             = jjnr[jidx];
 177             jnrB             = jjnr[jidx+1];
 178             j_coord_offsetA  = DIM*jnrA;
 179             j_coord_offsetB  = DIM*jnrB;
 180
 181             /* load j atom coordinates */
 182             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 183                                               &jx0,&jy0,&jz0);
 184
 185             /* Calculate displacement vector */
 186             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 187             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 188             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 189
 190             /* Calculate squared distance and things based on it */
 191             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 192
 193             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 194
 195             /* Load parameters for j particles */
 196             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 197             isaj0            = gmx_fjsp_load_2real_swizzle_v2r8(invsqrta+jnrA+0,invsqrta+jnrB+0);
 198             vdwjidx0A        = 2*vdwtype[jnrA+0];
 199             vdwjidx0B        = 2*vdwtype[jnrB+0];
 200
 201             /**************************
 202              * CALCULATE INTERACTIONS *
 203              **************************/
 204
 205             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 206
 207             /* Compute parameters for interactions between i and j atoms */
 208             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 209             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 210                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 211
 212             /* Calculate table index by multiplying r with table scale and truncate to integer */
 213             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 214             itab_tmp         = _fjsp_dtox_v2r8(rt);
 215             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 216             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 217             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 218
 219             vfconv.i[0]     *= 8;
 220             vfconv.i[1]     *= 8;
 221
 222             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 223             isaprod          = _fjsp_mul_v2r8(isai0,isaj0);
 224             gbqqfactor       = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
 225             gbscale          = _fjsp_mul_v2r8(isaprod,gbtabscale);
 226
 227             /* Calculate generalized born table index - this is a separate table from the normal one,
 228              * but we use the same procedure by multiplying r with scale and truncating to integer.
 229              */
 230             rt               = _fjsp_mul_v2r8(r00,gbscale);
 231             itab_tmp         = _fjsp_dtox_v2r8(rt);
 232             gbeps            = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
 233             _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
 234
 235             Y                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
 236             F                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] );
 237             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 238             G                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
 239             H                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] +2);
 240             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 241             Fp               = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
 242             VV               = _fjsp_madd_v2r8(gbeps,Fp,Y);
 243             vgb              = _fjsp_mul_v2r8(gbqqfactor,VV);
 244
 245             twogbeps         = _fjsp_add_v2r8(gbeps,gbeps);
 246             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
 247             fgb              = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
 248             dvdatmp          = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
 249             dvdasum          = _fjsp_add_v2r8(dvdasum,dvdatmp);
 250             gmx_fjsp_increment_2real_swizzle_v2r8(dvda+jnrA,dvda+jnrB,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
 251             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 252             felec            = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
 253
 254             /* CUBIC SPLINE TABLE DISPERSION */
 255             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 256             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 257             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 258             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 259             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 260             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 261             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 262             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 263             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 264             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 265             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 266
 267             /* CUBIC SPLINE TABLE REPULSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 273             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 274             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 275             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 276             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 279             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 280             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 281
 282             /* Update potential sum for this i atom from the interaction with this j atom. */
 283             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 284             vgbsum           = _fjsp_add_v2r8(vgbsum,vgb);
 285             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 286
 287             fscal            = _fjsp_add_v2r8(felec,fvdw);
 288
 289             /* Update vectorial force */
 290             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 291             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 292             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 293
 294             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 295
 296             /* Inner loop uses 95 flops */
 297         }
 298
 299         if(jidx<j_index_end)
 300         {
 301
 302             jnrA             = jjnr[jidx];
 303             j_coord_offsetA  = DIM*jnrA;
 304
 305             /* load j atom coordinates */
 306             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 307                                               &jx0,&jy0,&jz0);
 308
 309             /* Calculate displacement vector */
 310             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 311             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 312             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 313
 314             /* Calculate squared distance and things based on it */
 315             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 316
 317             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 318
 319             /* Load parameters for j particles */
 320             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 321             isaj0            = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),invsqrta+jnrA+0);
 322             vdwjidx0A        = 2*vdwtype[jnrA+0];
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 329
 330             /* Compute parameters for interactions between i and j atoms */
 331             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 332             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 333                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 334
 335             /* Calculate table index by multiplying r with table scale and truncate to integer */
 336             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 337             itab_tmp         = _fjsp_dtox_v2r8(rt);
 338             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 339             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 340             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 341
 342             vfconv.i[0]     *= 8;
 343             vfconv.i[1]     *= 8;
 344
 345             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 346             isaprod          = _fjsp_mul_v2r8(isai0,isaj0);
 347             gbqqfactor       = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
 348             gbscale          = _fjsp_mul_v2r8(isaprod,gbtabscale);
 349
 350             /* Calculate generalized born table index - this is a separate table from the normal one,
 351              * but we use the same procedure by multiplying r with scale and truncating to integer.
 352              */
 353             rt               = _fjsp_mul_v2r8(r00,gbscale);
 354             itab_tmp         = _fjsp_dtox_v2r8(rt);
 355             gbeps            = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
 356             _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
 357
 358             Y                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
 359             F                = _fjsp_setzero_v2r8();
 360             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 361             G                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
 362             H                = _fjsp_setzero_v2r8();
 363             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 364             Fp               = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
 365             VV               = _fjsp_madd_v2r8(gbeps,Fp,Y);
 366             vgb              = _fjsp_mul_v2r8(gbqqfactor,VV);
 367
 368             twogbeps         = _fjsp_add_v2r8(gbeps,gbeps);
 369             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
 370             fgb              = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
 371             dvdatmp          = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
 372             dvdasum          = _fjsp_add_v2r8(dvdasum,dvdatmp);
 373             gmx_fjsp_increment_1real_v2r8(dvda+jnrA,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
 374             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 375             felec            = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
 376
 377             /* CUBIC SPLINE TABLE DISPERSION */
 378             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 379             F                = _fjsp_setzero_v2r8();
 380             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 381             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 382             H                = _fjsp_setzero_v2r8();
 383             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 384             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 385             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 386             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 387             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 388             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 389
 390             /* CUBIC SPLINE TABLE REPULSION */
 391             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 392             F                = _fjsp_setzero_v2r8();
 393             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 394             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 395             H                = _fjsp_setzero_v2r8();
 396             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 397             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 398             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 399             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 400             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 401             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 402             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 403             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 404
 405             /* Update potential sum for this i atom from the interaction with this j atom. */
 406             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 407             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 408             vgb              = _fjsp_unpacklo_v2r8(vgb,_fjsp_setzero_v2r8());
 409             vgbsum           = _fjsp_add_v2r8(vgbsum,vgb);
 410             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 411             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 412
 413             fscal            = _fjsp_add_v2r8(felec,fvdw);
 414
 415             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 416
 417             /* Update vectorial force */
 418             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 419             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 420             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 421
 422             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 423
 424             /* Inner loop uses 95 flops */
 425         }
 426
 427         /* End of innermost loop */
 428
 429         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 430                                               f+i_coord_offset,fshift+i_shift_offset);
 431
 432         ggid                        = gid[iidx];
 433         /* Update potential energies */
 434         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 435         gmx_fjsp_update_1pot_v2r8(vgbsum,kernel_data->energygrp_polarization+ggid);
 436         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 437         dvdasum = _fjsp_mul_v2r8(dvdasum, _fjsp_mul_v2r8(isai0,isai0));
 438         gmx_fjsp_update_1pot_v2r8(dvdasum,dvda+inr);
 439
 440         /* Increment number of inner iterations */
 441         inneriter                  += j_index_end - j_index_start;
 442
 443         /* Outer loop uses 10 flops */
 444     }
 445
 446     /* Increment number of outer iterations */
 447     outeriter        += nri;
 448
 449     /* Update outer/inner flops */
 450
 451     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*10 + inneriter*95);
 452 }
 453 /*
 454  * Gromacs nonbonded kernel:   nb_kernel_ElecGB_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 455  * Electrostatics interaction: GeneralizedBorn
 456  * VdW interaction:            CubicSplineTable
 457  * Geometry:                   Particle-Particle
 458  * Calculate force/pot:        Force
 459  */
 460 void
 461 nb_kernel_ElecGB_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 462                     (t_nblist                    * gmx_restrict       nlist,
 463                      rvec                        * gmx_restrict          xx,
 464                      rvec                        * gmx_restrict          ff,
 465                      t_forcerec                  * gmx_restrict          fr,
 466                      t_mdatoms                   * gmx_restrict     mdatoms,
 467                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 468                      t_nrnb                      * gmx_restrict        nrnb)
 469 {
 470     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 471      * just 0 for non-waters.
 472      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 473      * jnr indices corresponding to data put in the four positions in the SIMD register.
 474      */
 475     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 476     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 477     int              jnrA,jnrB;
 478     int              j_coord_offsetA,j_coord_offsetB;
 479     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 480     real             rcutoff_scalar;
 481     real             *shiftvec,*fshift,*x,*f;
 482     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 483     int              vdwioffset0;
 484     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 485     int              vdwjidx0A,vdwjidx0B;
 486     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 487     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 488     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 489     real             *charge;
 490     _fjsp_v2r8       vgb,fgb,vgbsum,dvdasum,gbscale,gbtabscale,isaprod,gbqqfactor,gbinvepsdiff,dvdaj,gbeps,twogbeps,dvdatmp;
 491     _fjsp_v2r8       minushalf = gmx_fjsp_set1_v2r8(-0.5);
 492     real             *invsqrta,*dvda,*gbtab;
 493     int              nvdwtype;
 494     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 495     int              *vdwtype;
 496     real             *vdwparam;
 497     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 498     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 499     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 500     real             *vftab;
 501     _fjsp_v2r8       itab_tmp;
 502     _fjsp_v2r8       dummy_mask,cutoff_mask;
 503     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 504     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 505     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 506
 507     x                = xx[0];
 508     f                = ff[0];
 509
 510     nri              = nlist->nri;
 511     iinr             = nlist->iinr;
 512     jindex           = nlist->jindex;
 513     jjnr             = nlist->jjnr;
 514     shiftidx         = nlist->shift;
 515     gid              = nlist->gid;
 516     shiftvec         = fr->shift_vec[0];
 517     fshift           = fr->fshift[0];
 518     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 519     charge           = mdatoms->chargeA;
 520     nvdwtype         = fr->ntype;
 521     vdwparam         = fr->nbfp;
 522     vdwtype          = mdatoms->typeA;
 523
 524     vftab            = kernel_data->table_vdw->data;
 525     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 526
 527     invsqrta         = fr->invsqrta;
 528     dvda             = fr->dvda;
 529     gbtabscale       = gmx_fjsp_set1_v2r8(fr->gbtab.scale);
 530     gbtab            = fr->gbtab.data;
 531     gbinvepsdiff     = gmx_fjsp_set1_v2r8((1.0/fr->epsilon_r) - (1.0/fr->gb_epsilon_solvent));
 532
 533     /* Avoid stupid compiler warnings */
 534     jnrA = jnrB = 0;
 535     j_coord_offsetA = 0;
 536     j_coord_offsetB = 0;
 537
 538     outeriter        = 0;
 539     inneriter        = 0;
 540
 541     /* Start outer loop over neighborlists */
 542     for(iidx=0; iidx<nri; iidx++)
 543     {
 544         /* Load shift vector for this list */
 545         i_shift_offset   = DIM*shiftidx[iidx];
 546
 547         /* Load limits for loop over neighbors */
 548         j_index_start    = jindex[iidx];
 549         j_index_end      = jindex[iidx+1];
 550
 551         /* Get outer coordinate index */
 552         inr              = iinr[iidx];
 553         i_coord_offset   = DIM*inr;
 554
 555         /* Load i particle coords and add shift vector */
 556         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 557
 558         fix0             = _fjsp_setzero_v2r8();
 559         fiy0             = _fjsp_setzero_v2r8();
 560         fiz0             = _fjsp_setzero_v2r8();
 561
 562         /* Load parameters for i particles */
 563         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 564         isai0            = gmx_fjsp_load1_v2r8(invsqrta+inr+0);
 565         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 566
 567         dvdasum          = _fjsp_setzero_v2r8();
 568
 569         /* Start inner kernel loop */
 570         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 571         {
 572
 573             /* Get j neighbor index, and coordinate index */
 574             jnrA             = jjnr[jidx];
 575             jnrB             = jjnr[jidx+1];
 576             j_coord_offsetA  = DIM*jnrA;
 577             j_coord_offsetB  = DIM*jnrB;
 578
 579             /* load j atom coordinates */
 580             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 581                                               &jx0,&jy0,&jz0);
 582
 583             /* Calculate displacement vector */
 584             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 585             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 586             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 587
 588             /* Calculate squared distance and things based on it */
 589             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 590
 591             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 592
 593             /* Load parameters for j particles */
 594             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 595             isaj0            = gmx_fjsp_load_2real_swizzle_v2r8(invsqrta+jnrA+0,invsqrta+jnrB+0);
 596             vdwjidx0A        = 2*vdwtype[jnrA+0];
 597             vdwjidx0B        = 2*vdwtype[jnrB+0];
 598
 599             /**************************
 600              * CALCULATE INTERACTIONS *
 601              **************************/
 602
 603             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 604
 605             /* Compute parameters for interactions between i and j atoms */
 606             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 607             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 608                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 609
 610             /* Calculate table index by multiplying r with table scale and truncate to integer */
 611             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 612             itab_tmp         = _fjsp_dtox_v2r8(rt);
 613             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 614             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 615             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 616
 617             vfconv.i[0]     *= 8;
 618             vfconv.i[1]     *= 8;
 619
 620             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 621             isaprod          = _fjsp_mul_v2r8(isai0,isaj0);
 622             gbqqfactor       = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
 623             gbscale          = _fjsp_mul_v2r8(isaprod,gbtabscale);
 624
 625             /* Calculate generalized born table index - this is a separate table from the normal one,
 626              * but we use the same procedure by multiplying r with scale and truncating to integer.
 627              */
 628             rt               = _fjsp_mul_v2r8(r00,gbscale);
 629             itab_tmp         = _fjsp_dtox_v2r8(rt);
 630             gbeps            = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
 631             _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
 632
 633             Y                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
 634             F                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] );
 635             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 636             G                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
 637             H                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[1] +2);
 638             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 639             Fp               = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
 640             VV               = _fjsp_madd_v2r8(gbeps,Fp,Y);
 641             vgb              = _fjsp_mul_v2r8(gbqqfactor,VV);
 642
 643             twogbeps         = _fjsp_add_v2r8(gbeps,gbeps);
 644             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
 645             fgb              = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
 646             dvdatmp          = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
 647             dvdasum          = _fjsp_add_v2r8(dvdasum,dvdatmp);
 648             gmx_fjsp_increment_2real_swizzle_v2r8(dvda+jnrA,dvda+jnrB,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
 649             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 650             felec            = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
 651
 652             /* CUBIC SPLINE TABLE DISPERSION */
 653             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 654             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 655             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 656             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 657             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 658             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 659             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 660             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 661             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 662
 663             /* CUBIC SPLINE TABLE REPULSION */
 664             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 665             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 666             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 667             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 668             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 669             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 670             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 671             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 672             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 673             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 674
 675             fscal            = _fjsp_add_v2r8(felec,fvdw);
 676
 677             /* Update vectorial force */
 678             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 679             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 680             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 681
 682             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 683
 684             /* Inner loop uses 85 flops */
 685         }
 686
 687         if(jidx<j_index_end)
 688         {
 689
 690             jnrA             = jjnr[jidx];
 691             j_coord_offsetA  = DIM*jnrA;
 692
 693             /* load j atom coordinates */
 694             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 695                                               &jx0,&jy0,&jz0);
 696
 697             /* Calculate displacement vector */
 698             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 699             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 700             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 701
 702             /* Calculate squared distance and things based on it */
 703             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 704
 705             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 706
 707             /* Load parameters for j particles */
 708             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 709             isaj0            = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),invsqrta+jnrA+0);
 710             vdwjidx0A        = 2*vdwtype[jnrA+0];
 711
 712             /**************************
 713              * CALCULATE INTERACTIONS *
 714              **************************/
 715
 716             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 717
 718             /* Compute parameters for interactions between i and j atoms */
 719             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 720             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 721                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 722
 723             /* Calculate table index by multiplying r with table scale and truncate to integer */
 724             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 725             itab_tmp         = _fjsp_dtox_v2r8(rt);
 726             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 727             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 728             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 729
 730             vfconv.i[0]     *= 8;
 731             vfconv.i[1]     *= 8;
 732
 733             /* GENERALIZED BORN AND COULOMB ELECTROSTATICS */
 734             isaprod          = _fjsp_mul_v2r8(isai0,isaj0);
 735             gbqqfactor       = _fjsp_neg_v2r8(_fjsp_mul_v2r8(qq00,_fjsp_mul_v2r8(isaprod,gbinvepsdiff)));
 736             gbscale          = _fjsp_mul_v2r8(isaprod,gbtabscale);
 737
 738             /* Calculate generalized born table index - this is a separate table from the normal one,
 739              * but we use the same procedure by multiplying r with scale and truncating to integer.
 740              */
 741             rt               = _fjsp_mul_v2r8(r00,gbscale);
 742             itab_tmp         = _fjsp_dtox_v2r8(rt);
 743             gbeps            = _fjsp_sub_v2r8(rt,_fjsp_xtod_v2r8(itab_tmp));
 744             _fjsp_store_v2r8(&gbconv.simd,itab_tmp);
 745
 746             Y                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] );
 747             F                = _fjsp_setzero_v2r8();
 748             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 749             G                = _fjsp_load_v2r8( gbtab + 4*gbconv.i[0] +2);
 750             H                = _fjsp_setzero_v2r8();
 751             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 752             Fp               = _fjsp_madd_v2r8(gbeps,_fjsp_madd_v2r8(gbeps,H,G),F);
 753             VV               = _fjsp_madd_v2r8(gbeps,Fp,Y);
 754             vgb              = _fjsp_mul_v2r8(gbqqfactor,VV);
 755
 756             twogbeps         = _fjsp_add_v2r8(gbeps,gbeps);
 757             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twogbeps,H,G),gbeps,Fp);
 758             fgb              = _fjsp_mul_v2r8(gbqqfactor,_fjsp_mul_v2r8(FF,gbscale));
 759             dvdatmp          = _fjsp_mul_v2r8(minushalf,_fjsp_madd_v2r8(fgb,r00,vgb));
 760             dvdasum          = _fjsp_add_v2r8(dvdasum,dvdatmp);
 761             gmx_fjsp_increment_1real_v2r8(dvda+jnrA,_fjsp_mul_v2r8(dvdatmp,_fjsp_mul_v2r8(isaj0,isaj0)));
 762             velec            = _fjsp_mul_v2r8(qq00,rinv00);
 763             felec            = _fjsp_mul_v2r8(_fjsp_msub_v2r8(velec,rinv00,fgb),rinv00);
 764
 765             /* CUBIC SPLINE TABLE DISPERSION */
 766             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 767             F                = _fjsp_setzero_v2r8();
 768             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 769             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 770             H                = _fjsp_setzero_v2r8();
 771             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 772             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 773             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 774             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 775
 776             /* CUBIC SPLINE TABLE REPULSION */
 777             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 778             F                = _fjsp_setzero_v2r8();
 779             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 780             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 781             H                = _fjsp_setzero_v2r8();
 782             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 783             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 784             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 785             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 786             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 787
 788             fscal            = _fjsp_add_v2r8(felec,fvdw);
 789
 790             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 791
 792             /* Update vectorial force */
 793             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 794             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 795             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 796
 797             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 798
 799             /* Inner loop uses 85 flops */
 800         }
 801
 802         /* End of innermost loop */
 803
 804         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 805                                               f+i_coord_offset,fshift+i_shift_offset);
 806
 807         dvdasum = _fjsp_mul_v2r8(dvdasum, _fjsp_mul_v2r8(isai0,isai0));
 808         gmx_fjsp_update_1pot_v2r8(dvdasum,dvda+inr);
 809
 810         /* Increment number of inner iterations */
 811         inneriter                  += j_index_end - j_index_start;
 812
 813         /* Outer loop uses 7 flops */
 814     }
 815
 816     /* Increment number of outer iterations */
 817     outeriter        += nri;
 818
 819     /* Update outer/inner flops */
 820
 821     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*85);
 822 }