src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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 "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/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                      struct 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->ic->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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 439                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 440
 441             /* Calculate table index by multiplying r with table scale and truncate to integer */
 442             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 443             itab_tmp         = _fjsp_dtox_v2r8(rt);
 444             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 445             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 446             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 447
 448             vfconv.i[0]     *= 12;
 449             vfconv.i[1]     *= 12;
 450
 451             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 452             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 453             F                = _fjsp_setzero_v2r8();
 454             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 455             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 456             H                = _fjsp_setzero_v2r8();
 457             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 458             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 459             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 460             velec            = _fjsp_mul_v2r8(qq00,VV);
 461             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 462             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 463
 464             /* CUBIC SPLINE TABLE DISPERSION */
 465             vfconv.i[0]       += 4;
 466             vfconv.i[1]       += 4;
 467             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 468             F                = _fjsp_setzero_v2r8();
 469             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 470             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 471             H                = _fjsp_setzero_v2r8();
 472             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 473             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 474             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 475             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 476             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 477             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 478
 479             /* CUBIC SPLINE TABLE REPULSION */
 480             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 481             F                = _fjsp_setzero_v2r8();
 482             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 483             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 484             H                = _fjsp_setzero_v2r8();
 485             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 486             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 487             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 488             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 489             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 490             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 491             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 492             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 496             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 497             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 498             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 499
 500             fscal            = _fjsp_add_v2r8(felec,fvdw);
 501
 502             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 503
 504             /* Update vectorial force */
 505             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 506             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 507             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 508
 509             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 510             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 511             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 518
 519             /* Compute parameters for interactions between i and j atoms */
 520             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 521
 522             /* Calculate table index by multiplying r with table scale and truncate to integer */
 523             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 524             itab_tmp         = _fjsp_dtox_v2r8(rt);
 525             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 526             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 527             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 528
 529             vfconv.i[0]     *= 12;
 530             vfconv.i[1]     *= 12;
 531
 532             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 533             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 534             F                = _fjsp_setzero_v2r8();
 535             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 536             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 537             H                = _fjsp_setzero_v2r8();
 538             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 539             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 540             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 541             velec            = _fjsp_mul_v2r8(qq10,VV);
 542             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 543             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 544
 545             /* Update potential sum for this i atom from the interaction with this j atom. */
 546             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 547             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 548
 549             fscal            = felec;
 550
 551             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 552
 553             /* Update vectorial force */
 554             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 555             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 556             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 557
 558             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 559             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 560             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 567
 568             /* Compute parameters for interactions between i and j atoms */
 569             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 570
 571             /* Calculate table index by multiplying r with table scale and truncate to integer */
 572             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 573             itab_tmp         = _fjsp_dtox_v2r8(rt);
 574             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 575             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 576             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 577
 578             vfconv.i[0]     *= 12;
 579             vfconv.i[1]     *= 12;
 580
 581             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 582             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 583             F                = _fjsp_setzero_v2r8();
 584             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 585             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 586             H                = _fjsp_setzero_v2r8();
 587             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 588             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 589             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 590             velec            = _fjsp_mul_v2r8(qq20,VV);
 591             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 592             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 593
 594             /* Update potential sum for this i atom from the interaction with this j atom. */
 595             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 596             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 597
 598             fscal            = felec;
 599
 600             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 601
 602             /* Update vectorial force */
 603             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 604             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 605             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 606
 607             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 608             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 609             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 610
 611             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 612
 613             /* Inner loop uses 171 flops */
 614         }
 615
 616         /* End of innermost loop */
 617
 618         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 619                                               f+i_coord_offset,fshift+i_shift_offset);
 620
 621         ggid                        = gid[iidx];
 622         /* Update potential energies */
 623         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 624         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 625
 626         /* Increment number of inner iterations */
 627         inneriter                  += j_index_end - j_index_start;
 628
 629         /* Outer loop uses 20 flops */
 630     }
 631
 632     /* Increment number of outer iterations */
 633     outeriter        += nri;
 634
 635     /* Update outer/inner flops */
 636
 637     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*171);
 638 }
 639 /*
 640  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 641  * Electrostatics interaction: CubicSplineTable
 642  * VdW interaction:            CubicSplineTable
 643  * Geometry:                   Water3-Particle
 644  * Calculate force/pot:        Force
 645  */
 646 void
 647 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 648                     (t_nblist                    * gmx_restrict       nlist,
 649                      rvec                        * gmx_restrict          xx,
 650                      rvec                        * gmx_restrict          ff,
 651                      struct t_forcerec           * gmx_restrict          fr,
 652                      t_mdatoms                   * gmx_restrict     mdatoms,
 653                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 654                      t_nrnb                      * gmx_restrict        nrnb)
 655 {
 656     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 657      * just 0 for non-waters.
 658      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 659      * jnr indices corresponding to data put in the four positions in the SIMD register.
 660      */
 661     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 662     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 663     int              jnrA,jnrB;
 664     int              j_coord_offsetA,j_coord_offsetB;
 665     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 666     real             rcutoff_scalar;
 667     real             *shiftvec,*fshift,*x,*f;
 668     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 669     int              vdwioffset0;
 670     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 671     int              vdwioffset1;
 672     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 673     int              vdwioffset2;
 674     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 675     int              vdwjidx0A,vdwjidx0B;
 676     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 677     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 678     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 679     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 680     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 681     real             *charge;
 682     int              nvdwtype;
 683     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 684     int              *vdwtype;
 685     real             *vdwparam;
 686     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 687     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 688     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 689     real             *vftab;
 690     _fjsp_v2r8       itab_tmp;
 691     _fjsp_v2r8       dummy_mask,cutoff_mask;
 692     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 693     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 694     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 695
 696     x                = xx[0];
 697     f                = ff[0];
 698
 699     nri              = nlist->nri;
 700     iinr             = nlist->iinr;
 701     jindex           = nlist->jindex;
 702     jjnr             = nlist->jjnr;
 703     shiftidx         = nlist->shift;
 704     gid              = nlist->gid;
 705     shiftvec         = fr->shift_vec[0];
 706     fshift           = fr->fshift[0];
 707     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 708     charge           = mdatoms->chargeA;
 709     nvdwtype         = fr->ntype;
 710     vdwparam         = fr->nbfp;
 711     vdwtype          = mdatoms->typeA;
 712
 713     vftab            = kernel_data->table_elec_vdw->data;
 714     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 715
 716     /* Setup water-specific parameters */
 717     inr              = nlist->iinr[0];
 718     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 719     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 720     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 721     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 722
 723     /* Avoid stupid compiler warnings */
 724     jnrA = jnrB = 0;
 725     j_coord_offsetA = 0;
 726     j_coord_offsetB = 0;
 727
 728     outeriter        = 0;
 729     inneriter        = 0;
 730
 731     /* Start outer loop over neighborlists */
 732     for(iidx=0; iidx<nri; iidx++)
 733     {
 734         /* Load shift vector for this list */
 735         i_shift_offset   = DIM*shiftidx[iidx];
 736
 737         /* Load limits for loop over neighbors */
 738         j_index_start    = jindex[iidx];
 739         j_index_end      = jindex[iidx+1];
 740
 741         /* Get outer coordinate index */
 742         inr              = iinr[iidx];
 743         i_coord_offset   = DIM*inr;
 744
 745         /* Load i particle coords and add shift vector */
 746         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 747                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 748
 749         fix0             = _fjsp_setzero_v2r8();
 750         fiy0             = _fjsp_setzero_v2r8();
 751         fiz0             = _fjsp_setzero_v2r8();
 752         fix1             = _fjsp_setzero_v2r8();
 753         fiy1             = _fjsp_setzero_v2r8();
 754         fiz1             = _fjsp_setzero_v2r8();
 755         fix2             = _fjsp_setzero_v2r8();
 756         fiy2             = _fjsp_setzero_v2r8();
 757         fiz2             = _fjsp_setzero_v2r8();
 758
 759         /* Start inner kernel loop */
 760         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 761         {
 762
 763             /* Get j neighbor index, and coordinate index */
 764             jnrA             = jjnr[jidx];
 765             jnrB             = jjnr[jidx+1];
 766             j_coord_offsetA  = DIM*jnrA;
 767             j_coord_offsetB  = DIM*jnrB;
 768
 769             /* load j atom coordinates */
 770             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 771                                               &jx0,&jy0,&jz0);
 772
 773             /* Calculate displacement vector */
 774             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 775             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 776             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 777             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 778             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 779             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 780             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 781             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 782             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 783
 784             /* Calculate squared distance and things based on it */
 785             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 786             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 787             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 788
 789             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 790             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 791             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 792
 793             /* Load parameters for j particles */
 794             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 795             vdwjidx0A        = 2*vdwtype[jnrA+0];
 796             vdwjidx0B        = 2*vdwtype[jnrB+0];
 797
 798             fjx0             = _fjsp_setzero_v2r8();
 799             fjy0             = _fjsp_setzero_v2r8();
 800             fjz0             = _fjsp_setzero_v2r8();
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 807
 808             /* Compute parameters for interactions between i and j atoms */
 809             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 810             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 811                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 812
 813             /* Calculate table index by multiplying r with table scale and truncate to integer */
 814             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 815             itab_tmp         = _fjsp_dtox_v2r8(rt);
 816             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 817             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 818             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 819
 820             vfconv.i[0]     *= 12;
 821             vfconv.i[1]     *= 12;
 822
 823             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 824             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 825             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 826             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 827             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 828             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 829             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 830             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 831             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 832             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 833
 834             /* CUBIC SPLINE TABLE DISPERSION */
 835             vfconv.i[0]       += 4;
 836             vfconv.i[1]       += 4;
 837             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 838             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 839             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 840             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 841             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 842             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 843             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 844             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 845             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 846
 847             /* CUBIC SPLINE TABLE REPULSION */
 848             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 849             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 850             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 851             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 852             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 853             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 854             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 855             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 856             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 857             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 858
 859             fscal            = _fjsp_add_v2r8(felec,fvdw);
 860
 861             /* Update vectorial force */
 862             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 863             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 864             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 865
 866             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 867             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 868             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 869
 870             /**************************
 871              * CALCULATE INTERACTIONS *
 872              **************************/
 873
 874             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 875
 876             /* Compute parameters for interactions between i and j atoms */
 877             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 878
 879             /* Calculate table index by multiplying r with table scale and truncate to integer */
 880             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 881             itab_tmp         = _fjsp_dtox_v2r8(rt);
 882             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 883             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 884             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 885
 886             vfconv.i[0]     *= 12;
 887             vfconv.i[1]     *= 12;
 888
 889             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 890             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 891             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 892             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 893             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 894             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 895             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 896             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 897             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 898             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 899
 900             fscal            = felec;
 901
 902             /* Update vectorial force */
 903             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 904             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 905             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 906
 907             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 908             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 909             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 916
 917             /* Compute parameters for interactions between i and j atoms */
 918             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 919
 920             /* Calculate table index by multiplying r with table scale and truncate to integer */
 921             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 922             itab_tmp         = _fjsp_dtox_v2r8(rt);
 923             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 924             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 925             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 926
 927             vfconv.i[0]     *= 12;
 928             vfconv.i[1]     *= 12;
 929
 930             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 931             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 932             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 933             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 934             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 935             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 936             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 937             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 938             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 939             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 940
 941             fscal            = felec;
 942
 943             /* Update vectorial force */
 944             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 945             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 946             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 947
 948             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 949             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 950             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 951
 952             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 953
 954             /* Inner loop uses 151 flops */
 955         }
 956
 957         if(jidx<j_index_end)
 958         {
 959
 960             jnrA             = jjnr[jidx];
 961             j_coord_offsetA  = DIM*jnrA;
 962
 963             /* load j atom coordinates */
 964             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 965                                               &jx0,&jy0,&jz0);
 966
 967             /* Calculate displacement vector */
 968             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 969             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 970             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 971             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 972             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 973             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 974             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 975             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 976             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 977
 978             /* Calculate squared distance and things based on it */
 979             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 980             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 981             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 982
 983             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 984             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 985             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 986
 987             /* Load parameters for j particles */
 988             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 989             vdwjidx0A        = 2*vdwtype[jnrA+0];
 990
 991             fjx0             = _fjsp_setzero_v2r8();
 992             fjy0             = _fjsp_setzero_v2r8();
 993             fjz0             = _fjsp_setzero_v2r8();
 994
 995             /**************************
 996              * CALCULATE INTERACTIONS *
 997              **************************/
 998
 999             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1000
1001             /* Compute parameters for interactions between i and j atoms */
1002             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1003             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1004                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1005
1006             /* Calculate table index by multiplying r with table scale and truncate to integer */
1007             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1008             itab_tmp         = _fjsp_dtox_v2r8(rt);
1009             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1010             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1011             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1012
1013             vfconv.i[0]     *= 12;
1014             vfconv.i[1]     *= 12;
1015
1016             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1017             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1018             F                = _fjsp_setzero_v2r8();
1019             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1020             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1021             H                = _fjsp_setzero_v2r8();
1022             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1023             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1024             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1025             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
1026
1027             /* CUBIC SPLINE TABLE DISPERSION */
1028             vfconv.i[0]       += 4;
1029             vfconv.i[1]       += 4;
1030             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1031             F                = _fjsp_setzero_v2r8();
1032             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1033             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1034             H                = _fjsp_setzero_v2r8();
1035             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1036             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1037             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1038             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1039
1040             /* CUBIC SPLINE TABLE REPULSION */
1041             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1042             F                = _fjsp_setzero_v2r8();
1043             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1044             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1045             H                = _fjsp_setzero_v2r8();
1046             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1047             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1048             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1049             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1050             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1051
1052             fscal            = _fjsp_add_v2r8(felec,fvdw);
1053
1054             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1055
1056             /* Update vectorial force */
1057             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1058             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1059             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1060
1061             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1062             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1063             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1064
1065             /**************************
1066              * CALCULATE INTERACTIONS *
1067              **************************/
1068
1069             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1070
1071             /* Compute parameters for interactions between i and j atoms */
1072             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1073
1074             /* Calculate table index by multiplying r with table scale and truncate to integer */
1075             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1076             itab_tmp         = _fjsp_dtox_v2r8(rt);
1077             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1078             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1079             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1080
1081             vfconv.i[0]     *= 12;
1082             vfconv.i[1]     *= 12;
1083
1084             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1085             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1086             F                = _fjsp_setzero_v2r8();
1087             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1088             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1089             H                = _fjsp_setzero_v2r8();
1090             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1091             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1092             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1093             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1094
1095             fscal            = felec;
1096
1097             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1098
1099             /* Update vectorial force */
1100             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1101             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1102             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1103
1104             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1105             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1106             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1107
1108             /**************************
1109              * CALCULATE INTERACTIONS *
1110              **************************/
1111
1112             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1113
1114             /* Compute parameters for interactions between i and j atoms */
1115             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1116
1117             /* Calculate table index by multiplying r with table scale and truncate to integer */
1118             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1119             itab_tmp         = _fjsp_dtox_v2r8(rt);
1120             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1121             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1122             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1123
1124             vfconv.i[0]     *= 12;
1125             vfconv.i[1]     *= 12;
1126
1127             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1128             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1129             F                = _fjsp_setzero_v2r8();
1130             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1131             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1132             H                = _fjsp_setzero_v2r8();
1133             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1134             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1135             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1136             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1137
1138             fscal            = felec;
1139
1140             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1141
1142             /* Update vectorial force */
1143             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1144             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1145             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1146
1147             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1148             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1149             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1150
1151             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1152
1153             /* Inner loop uses 151 flops */
1154         }
1155
1156         /* End of innermost loop */
1157
1158         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1159                                               f+i_coord_offset,fshift+i_shift_offset);
1160
1161         /* Increment number of inner iterations */
1162         inneriter                  += j_index_end - j_index_start;
1163
1164         /* Outer loop uses 18 flops */
1165     }
1166
1167     /* Increment number of outer iterations */
1168     outeriter        += nri;
1169
1170     /* Update outer/inner flops */
1171
1172     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*151);
1173 }