src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecNone_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 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwCSTab_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: None
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecNone_VdwCSTab_GeomP1P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwjidx0A,vdwjidx0B;
  82     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  83     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  84     int              nvdwtype;
  85     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  86     int              *vdwtype;
  87     real             *vdwparam;
  88     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  89     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  90     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  91     real             *vftab;
  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     nvdwtype         = fr->ntype;
 110     vdwparam         = fr->nbfp;
 111     vdwtype          = mdatoms->typeA;
 112
 113     vftab            = kernel_data->table_vdw->data;
 114     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 115
 116     /* Avoid stupid compiler warnings */
 117     jnrA = jnrB = 0;
 118     j_coord_offsetA = 0;
 119     j_coord_offsetB = 0;
 120
 121     outeriter        = 0;
 122     inneriter        = 0;
 123
 124     /* Start outer loop over neighborlists */
 125     for(iidx=0; iidx<nri; iidx++)
 126     {
 127         /* Load shift vector for this list */
 128         i_shift_offset   = DIM*shiftidx[iidx];
 129
 130         /* Load limits for loop over neighbors */
 131         j_index_start    = jindex[iidx];
 132         j_index_end      = jindex[iidx+1];
 133
 134         /* Get outer coordinate index */
 135         inr              = iinr[iidx];
 136         i_coord_offset   = DIM*inr;
 137
 138         /* Load i particle coords and add shift vector */
 139         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 140
 141         fix0             = _fjsp_setzero_v2r8();
 142         fiy0             = _fjsp_setzero_v2r8();
 143         fiz0             = _fjsp_setzero_v2r8();
 144
 145         /* Load parameters for i particles */
 146         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 147
 148         /* Reset potential sums */
 149         vvdwsum          = _fjsp_setzero_v2r8();
 150
 151         /* Start inner kernel loop */
 152         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 153         {
 154
 155             /* Get j neighbor index, and coordinate index */
 156             jnrA             = jjnr[jidx];
 157             jnrB             = jjnr[jidx+1];
 158             j_coord_offsetA  = DIM*jnrA;
 159             j_coord_offsetB  = DIM*jnrB;
 160
 161             /* load j atom coordinates */
 162             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 163                                               &jx0,&jy0,&jz0);
 164
 165             /* Calculate displacement vector */
 166             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 167             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 168             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 169
 170             /* Calculate squared distance and things based on it */
 171             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 172
 173             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 174
 175             /* Load parameters for j particles */
 176             vdwjidx0A        = 2*vdwtype[jnrA+0];
 177             vdwjidx0B        = 2*vdwtype[jnrB+0];
 178
 179             /**************************
 180              * CALCULATE INTERACTIONS *
 181              **************************/
 182
 183             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 184
 185             /* Compute parameters for interactions between i and j atoms */
 186             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 187                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 188
 189             /* Calculate table index by multiplying r with table scale and truncate to integer */
 190             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 191             itab_tmp         = _fjsp_dtox_v2r8(rt);
 192             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 193             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 194             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 195
 196             vfconv.i[0]     *= 8;
 197             vfconv.i[1]     *= 8;
 198
 199             /* CUBIC SPLINE TABLE DISPERSION */
 200             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 201             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 202             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 203             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 204             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 205             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 206             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 207             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 208             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 209             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 210             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 211
 212             /* CUBIC SPLINE TABLE REPULSION */
 213             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 214             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 215             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 216             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 217             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 218             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 219             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 220             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 221             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 222             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 223             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 224             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 225             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 226
 227             /* Update potential sum for this i atom from the interaction with this j atom. */
 228             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 229
 230             fscal            = fvdw;
 231
 232             /* Update vectorial force */
 233             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 234             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 235             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 236
 237             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 238
 239             /* Inner loop uses 59 flops */
 240         }
 241
 242         if(jidx<j_index_end)
 243         {
 244
 245             jnrA             = jjnr[jidx];
 246             j_coord_offsetA  = DIM*jnrA;
 247
 248             /* load j atom coordinates */
 249             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 250                                               &jx0,&jy0,&jz0);
 251
 252             /* Calculate displacement vector */
 253             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 254             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 255             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 256
 257             /* Calculate squared distance and things based on it */
 258             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 259
 260             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 261
 262             /* Load parameters for j particles */
 263             vdwjidx0A        = 2*vdwtype[jnrA+0];
 264
 265             /**************************
 266              * CALCULATE INTERACTIONS *
 267              **************************/
 268
 269             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 270
 271             /* Compute parameters for interactions between i and j atoms */
 272             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 273
 274             /* Calculate table index by multiplying r with table scale and truncate to integer */
 275             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 276             itab_tmp         = _fjsp_dtox_v2r8(rt);
 277             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 278             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 279             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 280
 281             vfconv.i[0]     *= 8;
 282             vfconv.i[1]     *= 8;
 283
 284             /* CUBIC SPLINE TABLE DISPERSION */
 285             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 286             F                = _fjsp_setzero_v2r8();
 287             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 288             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 289             H                = _fjsp_setzero_v2r8();
 290             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 291             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 292             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 293             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 294             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 295             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 296
 297             /* CUBIC SPLINE TABLE REPULSION */
 298             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 299             F                = _fjsp_setzero_v2r8();
 300             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 301             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 302             H                = _fjsp_setzero_v2r8();
 303             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 304             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 305             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 306             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 307             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 308             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 309             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 310             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 311
 312             /* Update potential sum for this i atom from the interaction with this j atom. */
 313             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 314             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 315
 316             fscal            = 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 59 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(vvdwsum,kernel_data->energygrp_vdw+ggid);
 338
 339         /* Increment number of inner iterations */
 340         inneriter                  += j_index_end - j_index_start;
 341
 342         /* Outer loop uses 7 flops */
 343     }
 344
 345     /* Increment number of outer iterations */
 346     outeriter        += nri;
 347
 348     /* Update outer/inner flops */
 349
 350     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*59);
 351 }
 352 /*
 353  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 354  * Electrostatics interaction: None
 355  * VdW interaction:            CubicSplineTable
 356  * Geometry:                   Particle-Particle
 357  * Calculate force/pot:        Force
 358  */
 359 void
 360 nb_kernel_ElecNone_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 361                     (t_nblist                    * gmx_restrict       nlist,
 362                      rvec                        * gmx_restrict          xx,
 363                      rvec                        * gmx_restrict          ff,
 364                      t_forcerec                  * gmx_restrict          fr,
 365                      t_mdatoms                   * gmx_restrict     mdatoms,
 366                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 367                      t_nrnb                      * gmx_restrict        nrnb)
 368 {
 369     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 370      * just 0 for non-waters.
 371      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 372      * jnr indices corresponding to data put in the four positions in the SIMD register.
 373      */
 374     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 375     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 376     int              jnrA,jnrB;
 377     int              j_coord_offsetA,j_coord_offsetB;
 378     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 379     real             rcutoff_scalar;
 380     real             *shiftvec,*fshift,*x,*f;
 381     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 382     int              vdwioffset0;
 383     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 384     int              vdwjidx0A,vdwjidx0B;
 385     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 386     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 387     int              nvdwtype;
 388     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 389     int              *vdwtype;
 390     real             *vdwparam;
 391     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 392     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 393     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 394     real             *vftab;
 395     _fjsp_v2r8       itab_tmp;
 396     _fjsp_v2r8       dummy_mask,cutoff_mask;
 397     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 398     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 399     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 400
 401     x                = xx[0];
 402     f                = ff[0];
 403
 404     nri              = nlist->nri;
 405     iinr             = nlist->iinr;
 406     jindex           = nlist->jindex;
 407     jjnr             = nlist->jjnr;
 408     shiftidx         = nlist->shift;
 409     gid              = nlist->gid;
 410     shiftvec         = fr->shift_vec[0];
 411     fshift           = fr->fshift[0];
 412     nvdwtype         = fr->ntype;
 413     vdwparam         = fr->nbfp;
 414     vdwtype          = mdatoms->typeA;
 415
 416     vftab            = kernel_data->table_vdw->data;
 417     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 418
 419     /* Avoid stupid compiler warnings */
 420     jnrA = jnrB = 0;
 421     j_coord_offsetA = 0;
 422     j_coord_offsetB = 0;
 423
 424     outeriter        = 0;
 425     inneriter        = 0;
 426
 427     /* Start outer loop over neighborlists */
 428     for(iidx=0; iidx<nri; iidx++)
 429     {
 430         /* Load shift vector for this list */
 431         i_shift_offset   = DIM*shiftidx[iidx];
 432
 433         /* Load limits for loop over neighbors */
 434         j_index_start    = jindex[iidx];
 435         j_index_end      = jindex[iidx+1];
 436
 437         /* Get outer coordinate index */
 438         inr              = iinr[iidx];
 439         i_coord_offset   = DIM*inr;
 440
 441         /* Load i particle coords and add shift vector */
 442         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 443
 444         fix0             = _fjsp_setzero_v2r8();
 445         fiy0             = _fjsp_setzero_v2r8();
 446         fiz0             = _fjsp_setzero_v2r8();
 447
 448         /* Load parameters for i particles */
 449         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 450
 451         /* Start inner kernel loop */
 452         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 453         {
 454
 455             /* Get j neighbor index, and coordinate index */
 456             jnrA             = jjnr[jidx];
 457             jnrB             = jjnr[jidx+1];
 458             j_coord_offsetA  = DIM*jnrA;
 459             j_coord_offsetB  = DIM*jnrB;
 460
 461             /* load j atom coordinates */
 462             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 463                                               &jx0,&jy0,&jz0);
 464
 465             /* Calculate displacement vector */
 466             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 467             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 468             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 469
 470             /* Calculate squared distance and things based on it */
 471             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 472
 473             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 474
 475             /* Load parameters for j particles */
 476             vdwjidx0A        = 2*vdwtype[jnrA+0];
 477             vdwjidx0B        = 2*vdwtype[jnrB+0];
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 484
 485             /* Compute parameters for interactions between i and j atoms */
 486             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 487                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 488
 489             /* Calculate table index by multiplying r with table scale and truncate to integer */
 490             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 491             itab_tmp         = _fjsp_dtox_v2r8(rt);
 492             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 493             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 494             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 495
 496             vfconv.i[0]     *= 8;
 497             vfconv.i[1]     *= 8;
 498
 499             /* CUBIC SPLINE TABLE DISPERSION */
 500             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 501             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 502             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 503             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 504             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 505             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 506             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 507             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 508             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 509
 510             /* CUBIC SPLINE TABLE REPULSION */
 511             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 512             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 513             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 514             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 515             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 516             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 517             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 518             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 519             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 520             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 521
 522             fscal            = fvdw;
 523
 524             /* Update vectorial force */
 525             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 526             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 527             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 528
 529             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 530
 531             /* Inner loop uses 51 flops */
 532         }
 533
 534         if(jidx<j_index_end)
 535         {
 536
 537             jnrA             = jjnr[jidx];
 538             j_coord_offsetA  = DIM*jnrA;
 539
 540             /* load j atom coordinates */
 541             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 542                                               &jx0,&jy0,&jz0);
 543
 544             /* Calculate displacement vector */
 545             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 546             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 547             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 548
 549             /* Calculate squared distance and things based on it */
 550             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 551
 552             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 553
 554             /* Load parameters for j particles */
 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             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 565
 566             /* Calculate table index by multiplying r with table scale and truncate to integer */
 567             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 568             itab_tmp         = _fjsp_dtox_v2r8(rt);
 569             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 570             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 571             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 572
 573             vfconv.i[0]     *= 8;
 574             vfconv.i[1]     *= 8;
 575
 576             /* CUBIC SPLINE TABLE DISPERSION */
 577             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 578             F                = _fjsp_setzero_v2r8();
 579             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 580             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 581             H                = _fjsp_setzero_v2r8();
 582             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 583             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 584             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 585             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 586
 587             /* CUBIC SPLINE TABLE REPULSION */
 588             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 589             F                = _fjsp_setzero_v2r8();
 590             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 591             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 592             H                = _fjsp_setzero_v2r8();
 593             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 594             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 595             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 596             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 597             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 598
 599             fscal            = fvdw;
 600
 601             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 602
 603             /* Update vectorial force */
 604             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 605             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 606             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 607
 608             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 609
 610             /* Inner loop uses 51 flops */
 611         }
 612
 613         /* End of innermost loop */
 614
 615         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 616                                               f+i_coord_offset,fshift+i_shift_offset);
 617
 618         /* Increment number of inner iterations */
 619         inneriter                  += j_index_end - j_index_start;
 620
 621         /* Outer loop uses 6 flops */
 622     }
 623
 624     /* Increment number of outer iterations */
 625     outeriter        += nri;
 626
 627     /* Update outer/inner flops */
 628
 629     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*51);
 630 }