src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_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 "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_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              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_elec_vdw->data;
 124     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     /* Start outer loop over neighborlists */
 142     for(iidx=0; iidx<nri; iidx++)
 143     {
 144         /* Load shift vector for this list */
 145         i_shift_offset   = DIM*shiftidx[iidx];
 146
 147         /* Load limits for loop over neighbors */
 148         j_index_start    = jindex[iidx];
 149         j_index_end      = jindex[iidx+1];
 150
 151         /* Get outer coordinate index */
 152         inr              = iinr[iidx];
 153         i_coord_offset   = DIM*inr;
 154
 155         /* Load i particle coords and add shift vector */
 156         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 157                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 158
 159         fix0             = _fjsp_setzero_v2r8();
 160         fiy0             = _fjsp_setzero_v2r8();
 161         fiz0             = _fjsp_setzero_v2r8();
 162         fix1             = _fjsp_setzero_v2r8();
 163         fiy1             = _fjsp_setzero_v2r8();
 164         fiz1             = _fjsp_setzero_v2r8();
 165         fix2             = _fjsp_setzero_v2r8();
 166         fiy2             = _fjsp_setzero_v2r8();
 167         fiz2             = _fjsp_setzero_v2r8();
 168
 169         /* Reset potential sums */
 170         velecsum         = _fjsp_setzero_v2r8();
 171         vvdwsum          = _fjsp_setzero_v2r8();
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 175         {
 176
 177             /* Get j neighbor index, and coordinate index */
 178             jnrA             = jjnr[jidx];
 179             jnrB             = jjnr[jidx+1];
 180             j_coord_offsetA  = DIM*jnrA;
 181             j_coord_offsetB  = DIM*jnrB;
 182
 183             /* load j atom coordinates */
 184             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                               &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 189             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 190             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 191             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 192             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 193             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 194             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 195             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 196             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 197
 198             /* Calculate squared distance and things based on it */
 199             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 200             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 201             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 202
 203             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 204             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 205             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 206
 207             /* Load parameters for j particles */
 208             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 209             vdwjidx0A        = 2*vdwtype[jnrA+0];
 210             vdwjidx0B        = 2*vdwtype[jnrB+0];
 211
 212             fjx0             = _fjsp_setzero_v2r8();
 213             fjy0             = _fjsp_setzero_v2r8();
 214             fjz0             = _fjsp_setzero_v2r8();
 215
 216             /**************************
 217              * CALCULATE INTERACTIONS *
 218              **************************/
 219
 220             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 221
 222             /* Compute parameters for interactions between i and j atoms */
 223             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 224             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 225                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 226
 227             /* Calculate table index by multiplying r with table scale and truncate to integer */
 228             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 229             itab_tmp         = _fjsp_dtox_v2r8(rt);
 230             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 231             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 232             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 233
 234             vfconv.i[0]     *= 12;
 235             vfconv.i[1]     *= 12;
 236
 237             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 238             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 239             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 240             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 241             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 242             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 243             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 244             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 245             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 246             velec            = _fjsp_mul_v2r8(qq00,VV);
 247             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 248             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 249
 250             /* CUBIC SPLINE TABLE DISPERSION */
 251             vfconv.i[0]       += 4;
 252             vfconv.i[1]       += 4;
 253             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 254             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 255             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 256             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 257             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 258             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 259             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 260             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 261             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 262             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 263             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 264
 265             /* CUBIC SPLINE TABLE REPULSION */
 266             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 267             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 268             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 269             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 270             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 271             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 272             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 273             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 274             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 275             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 276             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 277             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 278             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 279
 280             /* Update potential sum for this i atom from the interaction with this j atom. */
 281             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 282             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 283
 284             fscal            = _fjsp_add_v2r8(felec,fvdw);
 285
 286             /* Update vectorial force */
 287             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 288             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 289             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 300
 301             /* Compute parameters for interactions between i and j atoms */
 302             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 303
 304             /* Calculate table index by multiplying r with table scale and truncate to integer */
 305             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 306             itab_tmp         = _fjsp_dtox_v2r8(rt);
 307             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 308             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 309             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 310
 311             vfconv.i[0]     *= 12;
 312             vfconv.i[1]     *= 12;
 313
 314             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 315             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 316             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 317             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 318             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 319             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 320             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 321             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 322             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 323             velec            = _fjsp_mul_v2r8(qq10,VV);
 324             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 325             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 326
 327             /* Update potential sum for this i atom from the interaction with this j atom. */
 328             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 329
 330             fscal            = felec;
 331
 332             /* Update vectorial force */
 333             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 334             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 335             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 336
 337             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 338             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 339             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 340
 341             /**************************
 342              * CALCULATE INTERACTIONS *
 343              **************************/
 344
 345             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 346
 347             /* Compute parameters for interactions between i and j atoms */
 348             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 349
 350             /* Calculate table index by multiplying r with table scale and truncate to integer */
 351             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 352             itab_tmp         = _fjsp_dtox_v2r8(rt);
 353             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 354             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 355             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 356
 357             vfconv.i[0]     *= 12;
 358             vfconv.i[1]     *= 12;
 359
 360             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 361             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 362             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 363             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 364             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 365             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 366             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 367             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 368             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 369             velec            = _fjsp_mul_v2r8(qq20,VV);
 370             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 371             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 372
 373             /* Update potential sum for this i atom from the interaction with this j atom. */
 374             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 375
 376             fscal            = felec;
 377
 378             /* Update vectorial force */
 379             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 380             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 381             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 382
 383             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 384             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 385             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 386
 387             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 388
 389             /* Inner loop uses 171 flops */
 390         }
 391
 392         if(jidx<j_index_end)
 393         {
 394
 395             jnrA             = jjnr[jidx];
 396             j_coord_offsetA  = DIM*jnrA;
 397
 398             /* load j atom coordinates */
 399             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 400                                               &jx0,&jy0,&jz0);
 401
 402             /* Calculate displacement vector */
 403             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 404             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 405             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 406             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 407             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 408             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 409             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 410             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 411             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 412
 413             /* Calculate squared distance and things based on it */
 414             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 415             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 416             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 417
 418             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 419             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 420             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 421
 422             /* Load parameters for j particles */
 423             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 424             vdwjidx0A        = 2*vdwtype[jnrA+0];
 425
 426             fjx0             = _fjsp_setzero_v2r8();
 427             fjy0             = _fjsp_setzero_v2r8();
 428             fjz0             = _fjsp_setzero_v2r8();
 429
 430             /**************************
 431              * CALCULATE INTERACTIONS *
 432              **************************/
 433
 434             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 435
 436             /* Compute parameters for interactions between i and j atoms */
 437             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 438             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 439
 440             /* Calculate table index by multiplying r with table scale and truncate to integer */
 441             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 442             itab_tmp         = _fjsp_dtox_v2r8(rt);
 443             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 444             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 445             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 446
 447             vfconv.i[0]     *= 12;
 448             vfconv.i[1]     *= 12;
 449
 450             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 451             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 452             F                = _fjsp_setzero_v2r8();
 453             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 454             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 455             H                = _fjsp_setzero_v2r8();
 456             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 457             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 458             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 459             velec            = _fjsp_mul_v2r8(qq00,VV);
 460             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 461             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 462
 463             /* CUBIC SPLINE TABLE DISPERSION */
 464             vfconv.i[0]       += 4;
 465             vfconv.i[1]       += 4;
 466             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 467             F                = _fjsp_setzero_v2r8();
 468             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 469             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 470             H                = _fjsp_setzero_v2r8();
 471             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 472             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 473             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 474             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 475             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 476             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 477
 478             /* CUBIC SPLINE TABLE REPULSION */
 479             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 480             F                = _fjsp_setzero_v2r8();
 481             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 482             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 483             H                = _fjsp_setzero_v2r8();
 484             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 485             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 486             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 487             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 488             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 489             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 490             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 491             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 492
 493             /* Update potential sum for this i atom from the interaction with this j atom. */
 494             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 495             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 496             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 497             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 498
 499             fscal            = _fjsp_add_v2r8(felec,fvdw);
 500
 501             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 502
 503             /* Update vectorial force */
 504             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 505             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 506             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 507
 508             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 509             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 510             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 517
 518             /* Compute parameters for interactions between i and j atoms */
 519             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 520
 521             /* Calculate table index by multiplying r with table scale and truncate to integer */
 522             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 523             itab_tmp         = _fjsp_dtox_v2r8(rt);
 524             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 525             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 526             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 527
 528             vfconv.i[0]     *= 12;
 529             vfconv.i[1]     *= 12;
 530
 531             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 532             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 533             F                = _fjsp_setzero_v2r8();
 534             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 535             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 536             H                = _fjsp_setzero_v2r8();
 537             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 538             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 539             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 540             velec            = _fjsp_mul_v2r8(qq10,VV);
 541             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 542             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 546             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 547
 548             fscal            = felec;
 549
 550             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 551
 552             /* Update vectorial force */
 553             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 554             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 555             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 556
 557             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 558             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 559             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 560
 561             /**************************
 562              * CALCULATE INTERACTIONS *
 563              **************************/
 564
 565             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 566
 567             /* Compute parameters for interactions between i and j atoms */
 568             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 569
 570             /* Calculate table index by multiplying r with table scale and truncate to integer */
 571             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 572             itab_tmp         = _fjsp_dtox_v2r8(rt);
 573             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 574             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 575             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 576
 577             vfconv.i[0]     *= 12;
 578             vfconv.i[1]     *= 12;
 579
 580             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 581             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 582             F                = _fjsp_setzero_v2r8();
 583             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 584             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 585             H                = _fjsp_setzero_v2r8();
 586             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 587             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 588             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 589             velec            = _fjsp_mul_v2r8(qq20,VV);
 590             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 591             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 592
 593             /* Update potential sum for this i atom from the interaction with this j atom. */
 594             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 595             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 596
 597             fscal            = felec;
 598
 599             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 600
 601             /* Update vectorial force */
 602             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 603             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 604             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 605
 606             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 607             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 608             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 609
 610             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 611
 612             /* Inner loop uses 171 flops */
 613         }
 614
 615         /* End of innermost loop */
 616
 617         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 618                                               f+i_coord_offset,fshift+i_shift_offset);
 619
 620         ggid                        = gid[iidx];
 621         /* Update potential energies */
 622         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 623         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 624
 625         /* Increment number of inner iterations */
 626         inneriter                  += j_index_end - j_index_start;
 627
 628         /* Outer loop uses 20 flops */
 629     }
 630
 631     /* Increment number of outer iterations */
 632     outeriter        += nri;
 633
 634     /* Update outer/inner flops */
 635
 636     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*171);
 637 }
 638 /*
 639  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 640  * Electrostatics interaction: CubicSplineTable
 641  * VdW interaction:            CubicSplineTable
 642  * Geometry:                   Water3-Particle
 643  * Calculate force/pot:        Force
 644  */
 645 void
 646 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 647                     (t_nblist                    * gmx_restrict       nlist,
 648                      rvec                        * gmx_restrict          xx,
 649                      rvec                        * gmx_restrict          ff,
 650                      t_forcerec                  * gmx_restrict          fr,
 651                      t_mdatoms                   * gmx_restrict     mdatoms,
 652                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 653                      t_nrnb                      * gmx_restrict        nrnb)
 654 {
 655     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 656      * just 0 for non-waters.
 657      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 658      * jnr indices corresponding to data put in the four positions in the SIMD register.
 659      */
 660     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 661     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 662     int              jnrA,jnrB;
 663     int              j_coord_offsetA,j_coord_offsetB;
 664     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 665     real             rcutoff_scalar;
 666     real             *shiftvec,*fshift,*x,*f;
 667     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 668     int              vdwioffset0;
 669     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 670     int              vdwioffset1;
 671     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 672     int              vdwioffset2;
 673     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 674     int              vdwjidx0A,vdwjidx0B;
 675     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 676     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 677     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 678     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 679     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 680     real             *charge;
 681     int              nvdwtype;
 682     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 683     int              *vdwtype;
 684     real             *vdwparam;
 685     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 686     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 687     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 688     real             *vftab;
 689     _fjsp_v2r8       itab_tmp;
 690     _fjsp_v2r8       dummy_mask,cutoff_mask;
 691     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 692     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 693     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 694
 695     x                = xx[0];
 696     f                = ff[0];
 697
 698     nri              = nlist->nri;
 699     iinr             = nlist->iinr;
 700     jindex           = nlist->jindex;
 701     jjnr             = nlist->jjnr;
 702     shiftidx         = nlist->shift;
 703     gid              = nlist->gid;
 704     shiftvec         = fr->shift_vec[0];
 705     fshift           = fr->fshift[0];
 706     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 707     charge           = mdatoms->chargeA;
 708     nvdwtype         = fr->ntype;
 709     vdwparam         = fr->nbfp;
 710     vdwtype          = mdatoms->typeA;
 711
 712     vftab            = kernel_data->table_elec_vdw->data;
 713     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 714
 715     /* Setup water-specific parameters */
 716     inr              = nlist->iinr[0];
 717     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 718     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 719     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 720     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 721
 722     /* Avoid stupid compiler warnings */
 723     jnrA = jnrB = 0;
 724     j_coord_offsetA = 0;
 725     j_coord_offsetB = 0;
 726
 727     outeriter        = 0;
 728     inneriter        = 0;
 729
 730     /* Start outer loop over neighborlists */
 731     for(iidx=0; iidx<nri; iidx++)
 732     {
 733         /* Load shift vector for this list */
 734         i_shift_offset   = DIM*shiftidx[iidx];
 735
 736         /* Load limits for loop over neighbors */
 737         j_index_start    = jindex[iidx];
 738         j_index_end      = jindex[iidx+1];
 739
 740         /* Get outer coordinate index */
 741         inr              = iinr[iidx];
 742         i_coord_offset   = DIM*inr;
 743
 744         /* Load i particle coords and add shift vector */
 745         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 746                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 747
 748         fix0             = _fjsp_setzero_v2r8();
 749         fiy0             = _fjsp_setzero_v2r8();
 750         fiz0             = _fjsp_setzero_v2r8();
 751         fix1             = _fjsp_setzero_v2r8();
 752         fiy1             = _fjsp_setzero_v2r8();
 753         fiz1             = _fjsp_setzero_v2r8();
 754         fix2             = _fjsp_setzero_v2r8();
 755         fiy2             = _fjsp_setzero_v2r8();
 756         fiz2             = _fjsp_setzero_v2r8();
 757
 758         /* Start inner kernel loop */
 759         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 760         {
 761
 762             /* Get j neighbor index, and coordinate index */
 763             jnrA             = jjnr[jidx];
 764             jnrB             = jjnr[jidx+1];
 765             j_coord_offsetA  = DIM*jnrA;
 766             j_coord_offsetB  = DIM*jnrB;
 767
 768             /* load j atom coordinates */
 769             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 770                                               &jx0,&jy0,&jz0);
 771
 772             /* Calculate displacement vector */
 773             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 774             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 775             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 776             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 777             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 778             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 779             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 780             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 781             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 782
 783             /* Calculate squared distance and things based on it */
 784             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 785             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 786             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 787
 788             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 789             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 790             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 791
 792             /* Load parameters for j particles */
 793             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 794             vdwjidx0A        = 2*vdwtype[jnrA+0];
 795             vdwjidx0B        = 2*vdwtype[jnrB+0];
 796
 797             fjx0             = _fjsp_setzero_v2r8();
 798             fjy0             = _fjsp_setzero_v2r8();
 799             fjz0             = _fjsp_setzero_v2r8();
 800
 801             /**************************
 802              * CALCULATE INTERACTIONS *
 803              **************************/
 804
 805             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 806
 807             /* Compute parameters for interactions between i and j atoms */
 808             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 809             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 810                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 811
 812             /* Calculate table index by multiplying r with table scale and truncate to integer */
 813             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 814             itab_tmp         = _fjsp_dtox_v2r8(rt);
 815             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 816             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 817             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 818
 819             vfconv.i[0]     *= 12;
 820             vfconv.i[1]     *= 12;
 821
 822             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 823             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 824             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 825             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 826             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 827             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 828             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 829             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 830             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 831             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 832
 833             /* CUBIC SPLINE TABLE DISPERSION */
 834             vfconv.i[0]       += 4;
 835             vfconv.i[1]       += 4;
 836             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 837             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 838             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 839             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 840             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 841             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 842             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 843             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 844             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 845
 846             /* CUBIC SPLINE TABLE REPULSION */
 847             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 848             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 849             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 850             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 851             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 852             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 853             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 854             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 855             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 856             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 857
 858             fscal            = _fjsp_add_v2r8(felec,fvdw);
 859
 860             /* Update vectorial force */
 861             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 862             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 863             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 864
 865             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 866             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 867             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 868
 869             /**************************
 870              * CALCULATE INTERACTIONS *
 871              **************************/
 872
 873             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 874
 875             /* Compute parameters for interactions between i and j atoms */
 876             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 877
 878             /* Calculate table index by multiplying r with table scale and truncate to integer */
 879             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 880             itab_tmp         = _fjsp_dtox_v2r8(rt);
 881             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 882             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 883             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 884
 885             vfconv.i[0]     *= 12;
 886             vfconv.i[1]     *= 12;
 887
 888             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 889             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 890             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 891             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 892             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 893             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 894             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 895             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 896             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 897             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 898
 899             fscal            = felec;
 900
 901             /* Update vectorial force */
 902             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 903             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 904             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 905
 906             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 907             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 908             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 909
 910             /**************************
 911              * CALCULATE INTERACTIONS *
 912              **************************/
 913
 914             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 915
 916             /* Compute parameters for interactions between i and j atoms */
 917             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 918
 919             /* Calculate table index by multiplying r with table scale and truncate to integer */
 920             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 921             itab_tmp         = _fjsp_dtox_v2r8(rt);
 922             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 923             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 924             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 925
 926             vfconv.i[0]     *= 12;
 927             vfconv.i[1]     *= 12;
 928
 929             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 930             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 931             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 932             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 933             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 934             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 935             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 936             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 937             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 938             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 939
 940             fscal            = felec;
 941
 942             /* Update vectorial force */
 943             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 944             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 945             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 946
 947             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 948             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 949             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 950
 951             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 952
 953             /* Inner loop uses 151 flops */
 954         }
 955
 956         if(jidx<j_index_end)
 957         {
 958
 959             jnrA             = jjnr[jidx];
 960             j_coord_offsetA  = DIM*jnrA;
 961
 962             /* load j atom coordinates */
 963             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 964                                               &jx0,&jy0,&jz0);
 965
 966             /* Calculate displacement vector */
 967             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 968             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 969             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 970             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 971             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 972             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 973             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 974             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 975             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 976
 977             /* Calculate squared distance and things based on it */
 978             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 979             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 980             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 981
 982             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 983             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 984             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 985
 986             /* Load parameters for j particles */
 987             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 988             vdwjidx0A        = 2*vdwtype[jnrA+0];
 989
 990             fjx0             = _fjsp_setzero_v2r8();
 991             fjy0             = _fjsp_setzero_v2r8();
 992             fjz0             = _fjsp_setzero_v2r8();
 993
 994             /**************************
 995              * CALCULATE INTERACTIONS *
 996              **************************/
 997
 998             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 999
1000             /* Compute parameters for interactions between i and j atoms */
1001             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1002             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1003
1004             /* Calculate table index by multiplying r with table scale and truncate to integer */
1005             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1006             itab_tmp         = _fjsp_dtox_v2r8(rt);
1007             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1008             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1009             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1010
1011             vfconv.i[0]     *= 12;
1012             vfconv.i[1]     *= 12;
1013
1014             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1015             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1016             F                = _fjsp_setzero_v2r8();
1017             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1018             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1019             H                = _fjsp_setzero_v2r8();
1020             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1021             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1022             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1023             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
1024
1025             /* CUBIC SPLINE TABLE DISPERSION */
1026             vfconv.i[0]       += 4;
1027             vfconv.i[1]       += 4;
1028             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1029             F                = _fjsp_setzero_v2r8();
1030             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1031             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1032             H                = _fjsp_setzero_v2r8();
1033             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1034             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1035             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1036             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1037
1038             /* CUBIC SPLINE TABLE REPULSION */
1039             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1040             F                = _fjsp_setzero_v2r8();
1041             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1042             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1043             H                = _fjsp_setzero_v2r8();
1044             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1045             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1046             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1047             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1048             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1049
1050             fscal            = _fjsp_add_v2r8(felec,fvdw);
1051
1052             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1053
1054             /* Update vectorial force */
1055             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1056             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1057             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1058
1059             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1060             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1061             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1062
1063             /**************************
1064              * CALCULATE INTERACTIONS *
1065              **************************/
1066
1067             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1068
1069             /* Compute parameters for interactions between i and j atoms */
1070             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1071
1072             /* Calculate table index by multiplying r with table scale and truncate to integer */
1073             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1074             itab_tmp         = _fjsp_dtox_v2r8(rt);
1075             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1076             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1077             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1078
1079             vfconv.i[0]     *= 12;
1080             vfconv.i[1]     *= 12;
1081
1082             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1083             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1084             F                = _fjsp_setzero_v2r8();
1085             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1086             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1087             H                = _fjsp_setzero_v2r8();
1088             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1089             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1090             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1091             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1092
1093             fscal            = felec;
1094
1095             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1096
1097             /* Update vectorial force */
1098             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1099             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1100             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1101
1102             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1103             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1104             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1105
1106             /**************************
1107              * CALCULATE INTERACTIONS *
1108              **************************/
1109
1110             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1111
1112             /* Compute parameters for interactions between i and j atoms */
1113             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1114
1115             /* Calculate table index by multiplying r with table scale and truncate to integer */
1116             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1117             itab_tmp         = _fjsp_dtox_v2r8(rt);
1118             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1119             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1120             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1121
1122             vfconv.i[0]     *= 12;
1123             vfconv.i[1]     *= 12;
1124
1125             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1126             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1127             F                = _fjsp_setzero_v2r8();
1128             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1129             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1130             H                = _fjsp_setzero_v2r8();
1131             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1132             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1133             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1134             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1135
1136             fscal            = felec;
1137
1138             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1139
1140             /* Update vectorial force */
1141             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1142             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1143             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1144
1145             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1146             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1147             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1148
1149             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1150
1151             /* Inner loop uses 151 flops */
1152         }
1153
1154         /* End of innermost loop */
1155
1156         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1157                                               f+i_coord_offset,fshift+i_shift_offset);
1158
1159         /* Increment number of inner iterations */
1160         inneriter                  += j_index_end - j_index_start;
1161
1162         /* Outer loop uses 18 flops */
1163     }
1164
1165     /* Increment number of outer iterations */
1166     outeriter        += nri;
1167
1168     /* Update outer/inner flops */
1169
1170     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*151);
1171 }