src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  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       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_elec->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[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     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* Avoid stupid compiler warnings */
 137     jnrA = jnrB = 0;
 138     j_coord_offsetA = 0;
 139     j_coord_offsetB = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     /* Start outer loop over neighborlists */
 145     for(iidx=0; iidx<nri; iidx++)
 146     {
 147         /* Load shift vector for this list */
 148         i_shift_offset   = DIM*shiftidx[iidx];
 149
 150         /* Load limits for loop over neighbors */
 151         j_index_start    = jindex[iidx];
 152         j_index_end      = jindex[iidx+1];
 153
 154         /* Get outer coordinate index */
 155         inr              = iinr[iidx];
 156         i_coord_offset   = DIM*inr;
 157
 158         /* Load i particle coords and add shift vector */
 159         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 160                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165         fix1             = _fjsp_setzero_v2r8();
 166         fiy1             = _fjsp_setzero_v2r8();
 167         fiz1             = _fjsp_setzero_v2r8();
 168         fix2             = _fjsp_setzero_v2r8();
 169         fiy2             = _fjsp_setzero_v2r8();
 170         fiz2             = _fjsp_setzero_v2r8();
 171         fix3             = _fjsp_setzero_v2r8();
 172         fiy3             = _fjsp_setzero_v2r8();
 173         fiz3             = _fjsp_setzero_v2r8();
 174
 175         /* Reset potential sums */
 176         velecsum         = _fjsp_setzero_v2r8();
 177         vvdwsum          = _fjsp_setzero_v2r8();
 178
 179         /* Start inner kernel loop */
 180         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 181         {
 182
 183             /* Get j neighbor index, and coordinate index */
 184             jnrA             = jjnr[jidx];
 185             jnrB             = jjnr[jidx+1];
 186             j_coord_offsetA  = DIM*jnrA;
 187             j_coord_offsetB  = DIM*jnrB;
 188
 189             /* load j atom coordinates */
 190             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 191                                               &jx0,&jy0,&jz0);
 192
 193             /* Calculate displacement vector */
 194             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 195             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 196             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 197             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 198             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 199             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 200             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 201             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 202             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 203             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 204             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 205             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 212
 213             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 214             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 215             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 216
 217             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 218
 219             /* Load parameters for j particles */
 220             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 221             vdwjidx0A        = 2*vdwtype[jnrA+0];
 222             vdwjidx0B        = 2*vdwtype[jnrB+0];
 223
 224             fjx0             = _fjsp_setzero_v2r8();
 225             fjy0             = _fjsp_setzero_v2r8();
 226             fjz0             = _fjsp_setzero_v2r8();
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             /* Compute parameters for interactions between i and j atoms */
 233             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 234                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 235
 236             /* LENNARD-JONES DISPERSION/REPULSION */
 237
 238             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 239             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 240             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 241             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 242             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 243
 244             /* Update potential sum for this i atom from the interaction with this j atom. */
 245             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 246
 247             fscal            = fvdw;
 248
 249             /* Update vectorial force */
 250             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 251             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 252             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 253
 254             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 255             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 256             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 257
 258             /**************************
 259              * CALCULATE INTERACTIONS *
 260              **************************/
 261
 262             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 263
 264             /* Compute parameters for interactions between i and j atoms */
 265             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 266
 267             /* Calculate table index by multiplying r with table scale and truncate to integer */
 268             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 269             itab_tmp         = _fjsp_dtox_v2r8(rt);
 270             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 271             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 272             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 273
 274             vfconv.i[0]     *= 4;
 275             vfconv.i[1]     *= 4;
 276
 277             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 278             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 279             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 280             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 281             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 282             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 283             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 284             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 285             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 286             velec            = _fjsp_mul_v2r8(qq10,VV);
 287             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 288             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 289
 290             /* Update potential sum for this i atom from the interaction with this j atom. */
 291             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 292
 293             fscal            = felec;
 294
 295             /* Update vectorial force */
 296             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 297             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 298             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 299
 300             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 301             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 302             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 303
 304             /**************************
 305              * CALCULATE INTERACTIONS *
 306              **************************/
 307
 308             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 309
 310             /* Compute parameters for interactions between i and j atoms */
 311             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 312
 313             /* Calculate table index by multiplying r with table scale and truncate to integer */
 314             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 315             itab_tmp         = _fjsp_dtox_v2r8(rt);
 316             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 317             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 318             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 319
 320             vfconv.i[0]     *= 4;
 321             vfconv.i[1]     *= 4;
 322
 323             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 324             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 325             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 326             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 327             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 328             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 329             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 330             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 331             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 332             velec            = _fjsp_mul_v2r8(qq20,VV);
 333             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 334             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 338
 339             fscal            = felec;
 340
 341             /* Update vectorial force */
 342             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 343             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 344             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 345
 346             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 347             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 348             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 349
 350             /**************************
 351              * CALCULATE INTERACTIONS *
 352              **************************/
 353
 354             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 355
 356             /* Compute parameters for interactions between i and j atoms */
 357             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 358
 359             /* Calculate table index by multiplying r with table scale and truncate to integer */
 360             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 361             itab_tmp         = _fjsp_dtox_v2r8(rt);
 362             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 363             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 364             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 365
 366             vfconv.i[0]     *= 4;
 367             vfconv.i[1]     *= 4;
 368
 369             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 370             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 371             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 372             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 373             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 374             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 375             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 376             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 377             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 378             velec            = _fjsp_mul_v2r8(qq30,VV);
 379             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 380             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Update vectorial force */
 388             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 389             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 390             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 391
 392             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 393             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 394             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 395
 396             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 397
 398             /* Inner loop uses 176 flops */
 399         }
 400
 401         if(jidx<j_index_end)
 402         {
 403
 404             jnrA             = jjnr[jidx];
 405             j_coord_offsetA  = DIM*jnrA;
 406
 407             /* load j atom coordinates */
 408             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 409                                               &jx0,&jy0,&jz0);
 410
 411             /* Calculate displacement vector */
 412             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 413             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 414             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 415             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 416             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 417             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 418             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 419             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 420             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 421             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 422             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 423             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 424
 425             /* Calculate squared distance and things based on it */
 426             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 427             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 428             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 429             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 430
 431             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 432             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 433             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 434
 435             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 436
 437             /* Load parameters for j particles */
 438             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 439             vdwjidx0A        = 2*vdwtype[jnrA+0];
 440
 441             fjx0             = _fjsp_setzero_v2r8();
 442             fjy0             = _fjsp_setzero_v2r8();
 443             fjz0             = _fjsp_setzero_v2r8();
 444
 445             /**************************
 446              * CALCULATE INTERACTIONS *
 447              **************************/
 448
 449             /* Compute parameters for interactions between i and j atoms */
 450             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 451
 452             /* LENNARD-JONES DISPERSION/REPULSION */
 453
 454             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 455             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 456             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 457             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 458             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 459
 460             /* Update potential sum for this i atom from the interaction with this j atom. */
 461             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 462             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 463
 464             fscal            = fvdw;
 465
 466             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 467
 468             /* Update vectorial force */
 469             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 470             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 471             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 472
 473             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 474             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 475             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 476
 477             /**************************
 478              * CALCULATE INTERACTIONS *
 479              **************************/
 480
 481             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 482
 483             /* Compute parameters for interactions between i and j atoms */
 484             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 485
 486             /* Calculate table index by multiplying r with table scale and truncate to integer */
 487             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 488             itab_tmp         = _fjsp_dtox_v2r8(rt);
 489             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 490             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 491             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 492
 493             vfconv.i[0]     *= 4;
 494             vfconv.i[1]     *= 4;
 495
 496             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 497             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 498             F                = _fjsp_setzero_v2r8();
 499             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 500             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 501             H                = _fjsp_setzero_v2r8();
 502             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 503             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 504             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 505             velec            = _fjsp_mul_v2r8(qq10,VV);
 506             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 507             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 511             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 512
 513             fscal            = felec;
 514
 515             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 516
 517             /* Update vectorial force */
 518             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 519             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 520             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 521
 522             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 523             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 524             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 525
 526             /**************************
 527              * CALCULATE INTERACTIONS *
 528              **************************/
 529
 530             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 531
 532             /* Compute parameters for interactions between i and j atoms */
 533             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 534
 535             /* Calculate table index by multiplying r with table scale and truncate to integer */
 536             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 537             itab_tmp         = _fjsp_dtox_v2r8(rt);
 538             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 539             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 540             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 541
 542             vfconv.i[0]     *= 4;
 543             vfconv.i[1]     *= 4;
 544
 545             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 546             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 547             F                = _fjsp_setzero_v2r8();
 548             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 549             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 550             H                = _fjsp_setzero_v2r8();
 551             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 552             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 553             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 554             velec            = _fjsp_mul_v2r8(qq20,VV);
 555             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 556             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 557
 558             /* Update potential sum for this i atom from the interaction with this j atom. */
 559             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 560             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 561
 562             fscal            = felec;
 563
 564             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 565
 566             /* Update vectorial force */
 567             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 568             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 569             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 570
 571             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 572             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 573             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 574
 575             /**************************
 576              * CALCULATE INTERACTIONS *
 577              **************************/
 578
 579             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 580
 581             /* Compute parameters for interactions between i and j atoms */
 582             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 583
 584             /* Calculate table index by multiplying r with table scale and truncate to integer */
 585             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 586             itab_tmp         = _fjsp_dtox_v2r8(rt);
 587             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 588             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 589             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 590
 591             vfconv.i[0]     *= 4;
 592             vfconv.i[1]     *= 4;
 593
 594             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 595             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 596             F                = _fjsp_setzero_v2r8();
 597             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 598             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 599             H                = _fjsp_setzero_v2r8();
 600             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 601             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 602             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 603             velec            = _fjsp_mul_v2r8(qq30,VV);
 604             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 605             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 606
 607             /* Update potential sum for this i atom from the interaction with this j atom. */
 608             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 609             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 610
 611             fscal            = felec;
 612
 613             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 614
 615             /* Update vectorial force */
 616             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 617             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 618             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 619
 620             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 621             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 622             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 623
 624             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 625
 626             /* Inner loop uses 176 flops */
 627         }
 628
 629         /* End of innermost loop */
 630
 631         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 632                                               f+i_coord_offset,fshift+i_shift_offset);
 633
 634         ggid                        = gid[iidx];
 635         /* Update potential energies */
 636         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 637         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 638
 639         /* Increment number of inner iterations */
 640         inneriter                  += j_index_end - j_index_start;
 641
 642         /* Outer loop uses 26 flops */
 643     }
 644
 645     /* Increment number of outer iterations */
 646     outeriter        += nri;
 647
 648     /* Update outer/inner flops */
 649
 650     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*176);
 651 }
 652 /*
 653  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 654  * Electrostatics interaction: CubicSplineTable
 655  * VdW interaction:            LennardJones
 656  * Geometry:                   Water4-Particle
 657  * Calculate force/pot:        Force
 658  */
 659 void
 660 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 661                     (t_nblist                    * gmx_restrict       nlist,
 662                      rvec                        * gmx_restrict          xx,
 663                      rvec                        * gmx_restrict          ff,
 664                      t_forcerec                  * gmx_restrict          fr,
 665                      t_mdatoms                   * gmx_restrict     mdatoms,
 666                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 667                      t_nrnb                      * gmx_restrict        nrnb)
 668 {
 669     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 670      * just 0 for non-waters.
 671      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 672      * jnr indices corresponding to data put in the four positions in the SIMD register.
 673      */
 674     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 675     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 676     int              jnrA,jnrB;
 677     int              j_coord_offsetA,j_coord_offsetB;
 678     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 679     real             rcutoff_scalar;
 680     real             *shiftvec,*fshift,*x,*f;
 681     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 682     int              vdwioffset0;
 683     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 684     int              vdwioffset1;
 685     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 686     int              vdwioffset2;
 687     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 688     int              vdwioffset3;
 689     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 690     int              vdwjidx0A,vdwjidx0B;
 691     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 692     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 693     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 694     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 695     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 696     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 697     real             *charge;
 698     int              nvdwtype;
 699     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 700     int              *vdwtype;
 701     real             *vdwparam;
 702     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 703     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 704     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 705     real             *vftab;
 706     _fjsp_v2r8       itab_tmp;
 707     _fjsp_v2r8       dummy_mask,cutoff_mask;
 708     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 709     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 710     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 711
 712     x                = xx[0];
 713     f                = ff[0];
 714
 715     nri              = nlist->nri;
 716     iinr             = nlist->iinr;
 717     jindex           = nlist->jindex;
 718     jjnr             = nlist->jjnr;
 719     shiftidx         = nlist->shift;
 720     gid              = nlist->gid;
 721     shiftvec         = fr->shift_vec[0];
 722     fshift           = fr->fshift[0];
 723     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 724     charge           = mdatoms->chargeA;
 725     nvdwtype         = fr->ntype;
 726     vdwparam         = fr->nbfp;
 727     vdwtype          = mdatoms->typeA;
 728
 729     vftab            = kernel_data->table_elec->data;
 730     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 731
 732     /* Setup water-specific parameters */
 733     inr              = nlist->iinr[0];
 734     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 735     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 736     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 737     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 738
 739     /* Avoid stupid compiler warnings */
 740     jnrA = jnrB = 0;
 741     j_coord_offsetA = 0;
 742     j_coord_offsetB = 0;
 743
 744     outeriter        = 0;
 745     inneriter        = 0;
 746
 747     /* Start outer loop over neighborlists */
 748     for(iidx=0; iidx<nri; iidx++)
 749     {
 750         /* Load shift vector for this list */
 751         i_shift_offset   = DIM*shiftidx[iidx];
 752
 753         /* Load limits for loop over neighbors */
 754         j_index_start    = jindex[iidx];
 755         j_index_end      = jindex[iidx+1];
 756
 757         /* Get outer coordinate index */
 758         inr              = iinr[iidx];
 759         i_coord_offset   = DIM*inr;
 760
 761         /* Load i particle coords and add shift vector */
 762         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 763                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 764
 765         fix0             = _fjsp_setzero_v2r8();
 766         fiy0             = _fjsp_setzero_v2r8();
 767         fiz0             = _fjsp_setzero_v2r8();
 768         fix1             = _fjsp_setzero_v2r8();
 769         fiy1             = _fjsp_setzero_v2r8();
 770         fiz1             = _fjsp_setzero_v2r8();
 771         fix2             = _fjsp_setzero_v2r8();
 772         fiy2             = _fjsp_setzero_v2r8();
 773         fiz2             = _fjsp_setzero_v2r8();
 774         fix3             = _fjsp_setzero_v2r8();
 775         fiy3             = _fjsp_setzero_v2r8();
 776         fiz3             = _fjsp_setzero_v2r8();
 777
 778         /* Start inner kernel loop */
 779         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 780         {
 781
 782             /* Get j neighbor index, and coordinate index */
 783             jnrA             = jjnr[jidx];
 784             jnrB             = jjnr[jidx+1];
 785             j_coord_offsetA  = DIM*jnrA;
 786             j_coord_offsetB  = DIM*jnrB;
 787
 788             /* load j atom coordinates */
 789             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 790                                               &jx0,&jy0,&jz0);
 791
 792             /* Calculate displacement vector */
 793             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 794             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 795             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 796             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 797             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 798             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 799             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 800             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 801             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 802             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 803             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 804             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 805
 806             /* Calculate squared distance and things based on it */
 807             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 808             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 809             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 810             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 811
 812             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 813             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 814             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 815
 816             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 817
 818             /* Load parameters for j particles */
 819             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 820             vdwjidx0A        = 2*vdwtype[jnrA+0];
 821             vdwjidx0B        = 2*vdwtype[jnrB+0];
 822
 823             fjx0             = _fjsp_setzero_v2r8();
 824             fjy0             = _fjsp_setzero_v2r8();
 825             fjz0             = _fjsp_setzero_v2r8();
 826
 827             /**************************
 828              * CALCULATE INTERACTIONS *
 829              **************************/
 830
 831             /* Compute parameters for interactions between i and j atoms */
 832             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 833                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 834
 835             /* LENNARD-JONES DISPERSION/REPULSION */
 836
 837             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 838             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 839
 840             fscal            = fvdw;
 841
 842             /* Update vectorial force */
 843             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 844             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 845             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 846
 847             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 848             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 849             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 856
 857             /* Compute parameters for interactions between i and j atoms */
 858             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 859
 860             /* Calculate table index by multiplying r with table scale and truncate to integer */
 861             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 862             itab_tmp         = _fjsp_dtox_v2r8(rt);
 863             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 864             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 865             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 866
 867             vfconv.i[0]     *= 4;
 868             vfconv.i[1]     *= 4;
 869
 870             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 871             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 872             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 873             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 874             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 875             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 876             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 877             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 878             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 879             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 880
 881             fscal            = felec;
 882
 883             /* Update vectorial force */
 884             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 885             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 886             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 887
 888             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 889             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 890             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 891
 892             /**************************
 893              * CALCULATE INTERACTIONS *
 894              **************************/
 895
 896             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 897
 898             /* Compute parameters for interactions between i and j atoms */
 899             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 900
 901             /* Calculate table index by multiplying r with table scale and truncate to integer */
 902             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 903             itab_tmp         = _fjsp_dtox_v2r8(rt);
 904             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 905             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 906             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 907
 908             vfconv.i[0]     *= 4;
 909             vfconv.i[1]     *= 4;
 910
 911             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 912             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 913             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 914             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 915             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 916             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 917             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 918             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 919             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 920             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 921
 922             fscal            = felec;
 923
 924             /* Update vectorial force */
 925             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 926             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 927             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 928
 929             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 930             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 931             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 932
 933             /**************************
 934              * CALCULATE INTERACTIONS *
 935              **************************/
 936
 937             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 938
 939             /* Compute parameters for interactions between i and j atoms */
 940             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 941
 942             /* Calculate table index by multiplying r with table scale and truncate to integer */
 943             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 944             itab_tmp         = _fjsp_dtox_v2r8(rt);
 945             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 946             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 947             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 948
 949             vfconv.i[0]     *= 4;
 950             vfconv.i[1]     *= 4;
 951
 952             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 953             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 954             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 955             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 956             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 957             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 958             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 959             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 960             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 961             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 962
 963             fscal            = felec;
 964
 965             /* Update vectorial force */
 966             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 967             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 968             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 969
 970             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 971             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 972             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 973
 974             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 975
 976             /* Inner loop uses 159 flops */
 977         }
 978
 979         if(jidx<j_index_end)
 980         {
 981
 982             jnrA             = jjnr[jidx];
 983             j_coord_offsetA  = DIM*jnrA;
 984
 985             /* load j atom coordinates */
 986             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 987                                               &jx0,&jy0,&jz0);
 988
 989             /* Calculate displacement vector */
 990             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 991             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 992             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 993             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 994             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 995             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 996             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 997             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 998             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 999             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1000             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1001             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1002
1003             /* Calculate squared distance and things based on it */
1004             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1005             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1006             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1007             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1008
1009             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1010             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1011             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1012
1013             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
1014
1015             /* Load parameters for j particles */
1016             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1017             vdwjidx0A        = 2*vdwtype[jnrA+0];
1018
1019             fjx0             = _fjsp_setzero_v2r8();
1020             fjy0             = _fjsp_setzero_v2r8();
1021             fjz0             = _fjsp_setzero_v2r8();
1022
1023             /**************************
1024              * CALCULATE INTERACTIONS *
1025              **************************/
1026
1027             /* Compute parameters for interactions between i and j atoms */
1028             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1029
1030             /* LENNARD-JONES DISPERSION/REPULSION */
1031
1032             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1033             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1034
1035             fscal            = fvdw;
1036
1037             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1038
1039             /* Update vectorial force */
1040             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1041             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1042             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1043
1044             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1045             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1046             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1047
1048             /**************************
1049              * CALCULATE INTERACTIONS *
1050              **************************/
1051
1052             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1053
1054             /* Compute parameters for interactions between i and j atoms */
1055             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1056
1057             /* Calculate table index by multiplying r with table scale and truncate to integer */
1058             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1059             itab_tmp         = _fjsp_dtox_v2r8(rt);
1060             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1061             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1062             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1063
1064             vfconv.i[0]     *= 4;
1065             vfconv.i[1]     *= 4;
1066
1067             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1068             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1069             F                = _fjsp_setzero_v2r8();
1070             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1071             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1072             H                = _fjsp_setzero_v2r8();
1073             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1074             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1075             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1076             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1077
1078             fscal            = felec;
1079
1080             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1081
1082             /* Update vectorial force */
1083             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1084             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1085             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1086
1087             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1088             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1089             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1090
1091             /**************************
1092              * CALCULATE INTERACTIONS *
1093              **************************/
1094
1095             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1096
1097             /* Compute parameters for interactions between i and j atoms */
1098             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1099
1100             /* Calculate table index by multiplying r with table scale and truncate to integer */
1101             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1102             itab_tmp         = _fjsp_dtox_v2r8(rt);
1103             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1104             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1105             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1106
1107             vfconv.i[0]     *= 4;
1108             vfconv.i[1]     *= 4;
1109
1110             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1111             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1112             F                = _fjsp_setzero_v2r8();
1113             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1114             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1115             H                = _fjsp_setzero_v2r8();
1116             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1117             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1118             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1119             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1120
1121             fscal            = felec;
1122
1123             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1124
1125             /* Update vectorial force */
1126             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1127             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1128             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1129
1130             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1131             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1132             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1133
1134             /**************************
1135              * CALCULATE INTERACTIONS *
1136              **************************/
1137
1138             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1139
1140             /* Compute parameters for interactions between i and j atoms */
1141             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1142
1143             /* Calculate table index by multiplying r with table scale and truncate to integer */
1144             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1145             itab_tmp         = _fjsp_dtox_v2r8(rt);
1146             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1147             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1148             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1149
1150             vfconv.i[0]     *= 4;
1151             vfconv.i[1]     *= 4;
1152
1153             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1154             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1155             F                = _fjsp_setzero_v2r8();
1156             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1157             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1158             H                = _fjsp_setzero_v2r8();
1159             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1160             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1161             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1162             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1163
1164             fscal            = felec;
1165
1166             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1167
1168             /* Update vectorial force */
1169             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1170             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1171             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1172
1173             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1174             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1175             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1176
1177             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1178
1179             /* Inner loop uses 159 flops */
1180         }
1181
1182         /* End of innermost loop */
1183
1184         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1185                                               f+i_coord_offset,fshift+i_shift_offset);
1186
1187         /* Increment number of inner iterations */
1188         inneriter                  += j_index_end - j_index_start;
1189
1190         /* Outer loop uses 24 flops */
1191     }
1192
1193     /* Increment number of outer iterations */
1194     outeriter        += nri;
1195
1196     /* Update outer/inner flops */
1197
1198     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);
1199 }