src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_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 #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_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_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              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_elec->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* Avoid stupid compiler warnings */
 136     jnrA = jnrB = 0;
 137     j_coord_offsetA = 0;
 138     j_coord_offsetB = 0;
 139
 140     outeriter        = 0;
 141     inneriter        = 0;
 142
 143     /* Start outer loop over neighborlists */
 144     for(iidx=0; iidx<nri; iidx++)
 145     {
 146         /* Load shift vector for this list */
 147         i_shift_offset   = DIM*shiftidx[iidx];
 148
 149         /* Load limits for loop over neighbors */
 150         j_index_start    = jindex[iidx];
 151         j_index_end      = jindex[iidx+1];
 152
 153         /* Get outer coordinate index */
 154         inr              = iinr[iidx];
 155         i_coord_offset   = DIM*inr;
 156
 157         /* Load i particle coords and add shift vector */
 158         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 159                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 160
 161         fix0             = _fjsp_setzero_v2r8();
 162         fiy0             = _fjsp_setzero_v2r8();
 163         fiz0             = _fjsp_setzero_v2r8();
 164         fix1             = _fjsp_setzero_v2r8();
 165         fiy1             = _fjsp_setzero_v2r8();
 166         fiz1             = _fjsp_setzero_v2r8();
 167         fix2             = _fjsp_setzero_v2r8();
 168         fiy2             = _fjsp_setzero_v2r8();
 169         fiz2             = _fjsp_setzero_v2r8();
 170
 171         /* Reset potential sums */
 172         velecsum         = _fjsp_setzero_v2r8();
 173         vvdwsum          = _fjsp_setzero_v2r8();
 174
 175         /* Start inner kernel loop */
 176         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 177         {
 178
 179             /* Get j neighbor index, and coordinate index */
 180             jnrA             = jjnr[jidx];
 181             jnrB             = jjnr[jidx+1];
 182             j_coord_offsetA  = DIM*jnrA;
 183             j_coord_offsetB  = DIM*jnrB;
 184
 185             /* load j atom coordinates */
 186             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 187                                               &jx0,&jy0,&jz0);
 188
 189             /* Calculate displacement vector */
 190             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 191             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 192             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 193             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 194             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 195             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 196             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 197             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 198             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 199
 200             /* Calculate squared distance and things based on it */
 201             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 202             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 203             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 204
 205             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 206             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 207             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 208
 209             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 210
 211             /* Load parameters for j particles */
 212             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 213             vdwjidx0A        = 2*vdwtype[jnrA+0];
 214             vdwjidx0B        = 2*vdwtype[jnrB+0];
 215
 216             fjx0             = _fjsp_setzero_v2r8();
 217             fjy0             = _fjsp_setzero_v2r8();
 218             fjz0             = _fjsp_setzero_v2r8();
 219
 220             /**************************
 221              * CALCULATE INTERACTIONS *
 222              **************************/
 223
 224             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 225
 226             /* Compute parameters for interactions between i and j atoms */
 227             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 228             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 229                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 230
 231             /* Calculate table index by multiplying r with table scale and truncate to integer */
 232             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 233             itab_tmp         = _fjsp_dtox_v2r8(rt);
 234             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 235             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 236             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 237
 238             vfconv.i[0]     *= 4;
 239             vfconv.i[1]     *= 4;
 240
 241             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 242             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 243             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 244             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 245             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 246             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 247             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 248             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 249             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 250             velec            = _fjsp_mul_v2r8(qq00,VV);
 251             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 252             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 253
 254             /* LENNARD-JONES DISPERSION/REPULSION */
 255
 256             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 257             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 258             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 259             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 260             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 261
 262             /* Update potential sum for this i atom from the interaction with this j atom. */
 263             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 264             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 265
 266             fscal            = _fjsp_add_v2r8(felec,fvdw);
 267
 268             /* Update vectorial force */
 269             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 270             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 271             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 272
 273             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 274             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 275             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 276
 277             /**************************
 278              * CALCULATE INTERACTIONS *
 279              **************************/
 280
 281             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 282
 283             /* Compute parameters for interactions between i and j atoms */
 284             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 285
 286             /* Calculate table index by multiplying r with table scale and truncate to integer */
 287             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 288             itab_tmp         = _fjsp_dtox_v2r8(rt);
 289             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 290             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 291             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 292
 293             vfconv.i[0]     *= 4;
 294             vfconv.i[1]     *= 4;
 295
 296             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 297             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 298             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 299             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 300             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 301             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 302             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 303             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 304             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 305             velec            = _fjsp_mul_v2r8(qq10,VV);
 306             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 307             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 308
 309             /* Update potential sum for this i atom from the interaction with this j atom. */
 310             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 311
 312             fscal            = felec;
 313
 314             /* Update vectorial force */
 315             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 316             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 317             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 318
 319             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 320             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 321             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 328
 329             /* Compute parameters for interactions between i and j atoms */
 330             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 331
 332             /* Calculate table index by multiplying r with table scale and truncate to integer */
 333             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 334             itab_tmp         = _fjsp_dtox_v2r8(rt);
 335             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 336             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 337             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 338
 339             vfconv.i[0]     *= 4;
 340             vfconv.i[1]     *= 4;
 341
 342             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 343             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 344             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 345             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 346             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 347             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 348             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 349             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 350             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 351             velec            = _fjsp_mul_v2r8(qq20,VV);
 352             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 353             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 357
 358             fscal            = felec;
 359
 360             /* Update vectorial force */
 361             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 362             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 363             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 364
 365             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 366             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 367             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 368
 369             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 370
 371             /* Inner loop uses 154 flops */
 372         }
 373
 374         if(jidx<j_index_end)
 375         {
 376
 377             jnrA             = jjnr[jidx];
 378             j_coord_offsetA  = DIM*jnrA;
 379
 380             /* load j atom coordinates */
 381             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 382                                               &jx0,&jy0,&jz0);
 383
 384             /* Calculate displacement vector */
 385             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 386             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 387             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 388             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 389             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 390             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 391             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 392             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 393             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 394
 395             /* Calculate squared distance and things based on it */
 396             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 397             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 398             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 399
 400             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 401             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 402             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 403
 404             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 405
 406             /* Load parameters for j particles */
 407             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 408             vdwjidx0A        = 2*vdwtype[jnrA+0];
 409
 410             fjx0             = _fjsp_setzero_v2r8();
 411             fjy0             = _fjsp_setzero_v2r8();
 412             fjz0             = _fjsp_setzero_v2r8();
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 419
 420             /* Compute parameters for interactions between i and j atoms */
 421             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 422             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 423                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 424
 425             /* Calculate table index by multiplying r with table scale and truncate to integer */
 426             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 427             itab_tmp         = _fjsp_dtox_v2r8(rt);
 428             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 429             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 430             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 431
 432             vfconv.i[0]     *= 4;
 433             vfconv.i[1]     *= 4;
 434
 435             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 436             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 437             F                = _fjsp_setzero_v2r8();
 438             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 439             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 440             H                = _fjsp_setzero_v2r8();
 441             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 442             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 443             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 444             velec            = _fjsp_mul_v2r8(qq00,VV);
 445             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 446             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 447
 448             /* LENNARD-JONES DISPERSION/REPULSION */
 449
 450             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 451             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 452             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 453             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 454             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 458             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 459             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 460             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 461
 462             fscal            = _fjsp_add_v2r8(felec,fvdw);
 463
 464             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 465
 466             /* Update vectorial force */
 467             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 468             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 469             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 470
 471             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 472             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 473             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 474
 475             /**************************
 476              * CALCULATE INTERACTIONS *
 477              **************************/
 478
 479             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 480
 481             /* Compute parameters for interactions between i and j atoms */
 482             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 483
 484             /* Calculate table index by multiplying r with table scale and truncate to integer */
 485             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 486             itab_tmp         = _fjsp_dtox_v2r8(rt);
 487             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 488             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 489             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 490
 491             vfconv.i[0]     *= 4;
 492             vfconv.i[1]     *= 4;
 493
 494             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 495             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 496             F                = _fjsp_setzero_v2r8();
 497             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 498             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 499             H                = _fjsp_setzero_v2r8();
 500             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 501             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 502             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 503             velec            = _fjsp_mul_v2r8(qq10,VV);
 504             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 505             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 506
 507             /* Update potential sum for this i atom from the interaction with this j atom. */
 508             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 509             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 510
 511             fscal            = felec;
 512
 513             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 514
 515             /* Update vectorial force */
 516             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 517             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 518             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 519
 520             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 521             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 522             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 523
 524             /**************************
 525              * CALCULATE INTERACTIONS *
 526              **************************/
 527
 528             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 529
 530             /* Compute parameters for interactions between i and j atoms */
 531             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 532
 533             /* Calculate table index by multiplying r with table scale and truncate to integer */
 534             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 535             itab_tmp         = _fjsp_dtox_v2r8(rt);
 536             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 537             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 538             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 539
 540             vfconv.i[0]     *= 4;
 541             vfconv.i[1]     *= 4;
 542
 543             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 544             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 545             F                = _fjsp_setzero_v2r8();
 546             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 547             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 548             H                = _fjsp_setzero_v2r8();
 549             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 550             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 551             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 552             velec            = _fjsp_mul_v2r8(qq20,VV);
 553             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 554             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 555
 556             /* Update potential sum for this i atom from the interaction with this j atom. */
 557             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 558             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 559
 560             fscal            = felec;
 561
 562             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 563
 564             /* Update vectorial force */
 565             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 566             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 567             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 568
 569             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 570             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 571             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 572
 573             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 574
 575             /* Inner loop uses 154 flops */
 576         }
 577
 578         /* End of innermost loop */
 579
 580         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 581                                               f+i_coord_offset,fshift+i_shift_offset);
 582
 583         ggid                        = gid[iidx];
 584         /* Update potential energies */
 585         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 586         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 587
 588         /* Increment number of inner iterations */
 589         inneriter                  += j_index_end - j_index_start;
 590
 591         /* Outer loop uses 20 flops */
 592     }
 593
 594     /* Increment number of outer iterations */
 595     outeriter        += nri;
 596
 597     /* Update outer/inner flops */
 598
 599     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 600 }
 601 /*
 602  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 603  * Electrostatics interaction: CubicSplineTable
 604  * VdW interaction:            LennardJones
 605  * Geometry:                   Water3-Particle
 606  * Calculate force/pot:        Force
 607  */
 608 void
 609 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 610                     (t_nblist                    * gmx_restrict       nlist,
 611                      rvec                        * gmx_restrict          xx,
 612                      rvec                        * gmx_restrict          ff,
 613                      t_forcerec                  * gmx_restrict          fr,
 614                      t_mdatoms                   * gmx_restrict     mdatoms,
 615                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 616                      t_nrnb                      * gmx_restrict        nrnb)
 617 {
 618     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 619      * just 0 for non-waters.
 620      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 621      * jnr indices corresponding to data put in the four positions in the SIMD register.
 622      */
 623     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 624     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 625     int              jnrA,jnrB;
 626     int              j_coord_offsetA,j_coord_offsetB;
 627     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 628     real             rcutoff_scalar;
 629     real             *shiftvec,*fshift,*x,*f;
 630     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 631     int              vdwioffset0;
 632     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 633     int              vdwioffset1;
 634     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 635     int              vdwioffset2;
 636     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 637     int              vdwjidx0A,vdwjidx0B;
 638     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 639     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 640     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 641     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 642     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 643     real             *charge;
 644     int              nvdwtype;
 645     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 646     int              *vdwtype;
 647     real             *vdwparam;
 648     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 649     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 650     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 651     real             *vftab;
 652     _fjsp_v2r8       itab_tmp;
 653     _fjsp_v2r8       dummy_mask,cutoff_mask;
 654     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 655     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 656     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 657
 658     x                = xx[0];
 659     f                = ff[0];
 660
 661     nri              = nlist->nri;
 662     iinr             = nlist->iinr;
 663     jindex           = nlist->jindex;
 664     jjnr             = nlist->jjnr;
 665     shiftidx         = nlist->shift;
 666     gid              = nlist->gid;
 667     shiftvec         = fr->shift_vec[0];
 668     fshift           = fr->fshift[0];
 669     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 670     charge           = mdatoms->chargeA;
 671     nvdwtype         = fr->ntype;
 672     vdwparam         = fr->nbfp;
 673     vdwtype          = mdatoms->typeA;
 674
 675     vftab            = kernel_data->table_elec->data;
 676     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 677
 678     /* Setup water-specific parameters */
 679     inr              = nlist->iinr[0];
 680     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 681     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 682     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 683     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 684
 685     /* Avoid stupid compiler warnings */
 686     jnrA = jnrB = 0;
 687     j_coord_offsetA = 0;
 688     j_coord_offsetB = 0;
 689
 690     outeriter        = 0;
 691     inneriter        = 0;
 692
 693     /* Start outer loop over neighborlists */
 694     for(iidx=0; iidx<nri; iidx++)
 695     {
 696         /* Load shift vector for this list */
 697         i_shift_offset   = DIM*shiftidx[iidx];
 698
 699         /* Load limits for loop over neighbors */
 700         j_index_start    = jindex[iidx];
 701         j_index_end      = jindex[iidx+1];
 702
 703         /* Get outer coordinate index */
 704         inr              = iinr[iidx];
 705         i_coord_offset   = DIM*inr;
 706
 707         /* Load i particle coords and add shift vector */
 708         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 709                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 710
 711         fix0             = _fjsp_setzero_v2r8();
 712         fiy0             = _fjsp_setzero_v2r8();
 713         fiz0             = _fjsp_setzero_v2r8();
 714         fix1             = _fjsp_setzero_v2r8();
 715         fiy1             = _fjsp_setzero_v2r8();
 716         fiz1             = _fjsp_setzero_v2r8();
 717         fix2             = _fjsp_setzero_v2r8();
 718         fiy2             = _fjsp_setzero_v2r8();
 719         fiz2             = _fjsp_setzero_v2r8();
 720
 721         /* Start inner kernel loop */
 722         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 723         {
 724
 725             /* Get j neighbor index, and coordinate index */
 726             jnrA             = jjnr[jidx];
 727             jnrB             = jjnr[jidx+1];
 728             j_coord_offsetA  = DIM*jnrA;
 729             j_coord_offsetB  = DIM*jnrB;
 730
 731             /* load j atom coordinates */
 732             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 733                                               &jx0,&jy0,&jz0);
 734
 735             /* Calculate displacement vector */
 736             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 737             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 738             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 739             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 740             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 741             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 742             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 743             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 744             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 745
 746             /* Calculate squared distance and things based on it */
 747             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 748             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 749             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 750
 751             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 752             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 753             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 754
 755             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 756
 757             /* Load parameters for j particles */
 758             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 759             vdwjidx0A        = 2*vdwtype[jnrA+0];
 760             vdwjidx0B        = 2*vdwtype[jnrB+0];
 761
 762             fjx0             = _fjsp_setzero_v2r8();
 763             fjy0             = _fjsp_setzero_v2r8();
 764             fjz0             = _fjsp_setzero_v2r8();
 765
 766             /**************************
 767              * CALCULATE INTERACTIONS *
 768              **************************/
 769
 770             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 771
 772             /* Compute parameters for interactions between i and j atoms */
 773             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 774             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 775                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 776
 777             /* Calculate table index by multiplying r with table scale and truncate to integer */
 778             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 779             itab_tmp         = _fjsp_dtox_v2r8(rt);
 780             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 781             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 782             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 783
 784             vfconv.i[0]     *= 4;
 785             vfconv.i[1]     *= 4;
 786
 787             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 788             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 789             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 790             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 791             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 792             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 793             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 794             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 795             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 796             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 797
 798             /* LENNARD-JONES DISPERSION/REPULSION */
 799
 800             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 801             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 802
 803             fscal            = _fjsp_add_v2r8(felec,fvdw);
 804
 805             /* Update vectorial force */
 806             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 807             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 808             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 809
 810             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 811             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 812             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 813
 814             /**************************
 815              * CALCULATE INTERACTIONS *
 816              **************************/
 817
 818             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 819
 820             /* Compute parameters for interactions between i and j atoms */
 821             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 822
 823             /* Calculate table index by multiplying r with table scale and truncate to integer */
 824             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 825             itab_tmp         = _fjsp_dtox_v2r8(rt);
 826             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 827             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 828             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 829
 830             vfconv.i[0]     *= 4;
 831             vfconv.i[1]     *= 4;
 832
 833             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 834             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 835             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 836             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 837             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 838             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 839             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 840             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 841             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 842             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 843
 844             fscal            = felec;
 845
 846             /* Update vectorial force */
 847             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 848             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 849             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 850
 851             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 852             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 853             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 854
 855             /**************************
 856              * CALCULATE INTERACTIONS *
 857              **************************/
 858
 859             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 860
 861             /* Compute parameters for interactions between i and j atoms */
 862             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 863
 864             /* Calculate table index by multiplying r with table scale and truncate to integer */
 865             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 866             itab_tmp         = _fjsp_dtox_v2r8(rt);
 867             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 868             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 869             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 870
 871             vfconv.i[0]     *= 4;
 872             vfconv.i[1]     *= 4;
 873
 874             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 875             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 876             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 877             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 878             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 879             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 880             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 881             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 882             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 883             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 884
 885             fscal            = felec;
 886
 887             /* Update vectorial force */
 888             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 889             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 890             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 891
 892             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 893             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 894             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 895
 896             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 897
 898             /* Inner loop uses 137 flops */
 899         }
 900
 901         if(jidx<j_index_end)
 902         {
 903
 904             jnrA             = jjnr[jidx];
 905             j_coord_offsetA  = DIM*jnrA;
 906
 907             /* load j atom coordinates */
 908             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 909                                               &jx0,&jy0,&jz0);
 910
 911             /* Calculate displacement vector */
 912             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 913             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 914             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 915             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 916             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 917             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 918             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 919             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 920             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 921
 922             /* Calculate squared distance and things based on it */
 923             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 924             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 925             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 926
 927             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 928             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 929             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 930
 931             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 932
 933             /* Load parameters for j particles */
 934             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 935             vdwjidx0A        = 2*vdwtype[jnrA+0];
 936
 937             fjx0             = _fjsp_setzero_v2r8();
 938             fjy0             = _fjsp_setzero_v2r8();
 939             fjz0             = _fjsp_setzero_v2r8();
 940
 941             /**************************
 942              * CALCULATE INTERACTIONS *
 943              **************************/
 944
 945             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 946
 947             /* Compute parameters for interactions between i and j atoms */
 948             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 949             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 950                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 951
 952             /* Calculate table index by multiplying r with table scale and truncate to integer */
 953             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 954             itab_tmp         = _fjsp_dtox_v2r8(rt);
 955             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 956             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 957             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 958
 959             vfconv.i[0]     *= 4;
 960             vfconv.i[1]     *= 4;
 961
 962             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 963             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 964             F                = _fjsp_setzero_v2r8();
 965             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 966             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 967             H                = _fjsp_setzero_v2r8();
 968             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 969             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 970             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 971             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 972
 973             /* LENNARD-JONES DISPERSION/REPULSION */
 974
 975             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 976             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 977
 978             fscal            = _fjsp_add_v2r8(felec,fvdw);
 979
 980             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 981
 982             /* Update vectorial force */
 983             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 984             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 985             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 986
 987             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 988             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 989             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 996
 997             /* Compute parameters for interactions between i and j atoms */
 998             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 999
1000             /* Calculate table index by multiplying r with table scale and truncate to integer */
1001             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1002             itab_tmp         = _fjsp_dtox_v2r8(rt);
1003             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1004             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1005             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1006
1007             vfconv.i[0]     *= 4;
1008             vfconv.i[1]     *= 4;
1009
1010             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1011             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1012             F                = _fjsp_setzero_v2r8();
1013             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1014             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1015             H                = _fjsp_setzero_v2r8();
1016             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1017             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1018             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1019             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1020
1021             fscal            = felec;
1022
1023             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1024
1025             /* Update vectorial force */
1026             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1027             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1028             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1029
1030             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1031             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1032             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1033
1034             /**************************
1035              * CALCULATE INTERACTIONS *
1036              **************************/
1037
1038             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1039
1040             /* Compute parameters for interactions between i and j atoms */
1041             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1042
1043             /* Calculate table index by multiplying r with table scale and truncate to integer */
1044             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1045             itab_tmp         = _fjsp_dtox_v2r8(rt);
1046             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1047             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1048             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1049
1050             vfconv.i[0]     *= 4;
1051             vfconv.i[1]     *= 4;
1052
1053             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1054             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1055             F                = _fjsp_setzero_v2r8();
1056             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1057             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1058             H                = _fjsp_setzero_v2r8();
1059             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1060             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1061             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1062             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1063
1064             fscal            = felec;
1065
1066             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1067
1068             /* Update vectorial force */
1069             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1070             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1071             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1072
1073             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1074             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1075             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1076
1077             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1078
1079             /* Inner loop uses 137 flops */
1080         }
1081
1082         /* End of innermost loop */
1083
1084         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1085                                               f+i_coord_offset,fshift+i_shift_offset);
1086
1087         /* Increment number of inner iterations */
1088         inneriter                  += j_index_end - j_index_start;
1089
1090         /* Outer loop uses 18 flops */
1091     }
1092
1093     /* Increment number of outer iterations */
1094     outeriter        += nri;
1095
1096     /* Update outer/inner flops */
1097
1098     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*137);
1099 }