src/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012, by the GROMACS development team, led by
   5  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   6  * others, as listed in the AUTHORS file in the top-level source
   7  * directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist * gmx_restrict                nlist,
  61                      rvec * gmx_restrict                    xx,
  62                      rvec * gmx_restrict                    ff,
  63                      t_forcerec * gmx_restrict              fr,
  64                      t_mdatoms * gmx_restrict               mdatoms,
  65                      nb_kernel_data_t * gmx_restrict        kernel_data,
  66                      t_nrnb * gmx_restrict                  nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 104     real             *vftab;
 105     _fjsp_v2r8       itab_tmp;
 106     _fjsp_v2r8       dummy_mask,cutoff_mask;
 107     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 108     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 109     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 110
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 125     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 126     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 127     nvdwtype         = fr->ntype;
 128     vdwparam         = fr->nbfp;
 129     vdwtype          = mdatoms->typeA;
 130
 131     vftab            = kernel_data->table_vdw->data;
 132     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 133
 134     /* Setup water-specific parameters */
 135     inr              = nlist->iinr[0];
 136     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 137     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 138     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 139     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 140
 141     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 142     rcutoff_scalar   = fr->rcoulomb;
 143     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 144     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 145
 146     /* Avoid stupid compiler warnings */
 147     jnrA = jnrB = 0;
 148     j_coord_offsetA = 0;
 149     j_coord_offsetB = 0;
 150
 151     outeriter        = 0;
 152     inneriter        = 0;
 153
 154     /* Start outer loop over neighborlists */
 155     for(iidx=0; iidx<nri; iidx++)
 156     {
 157         /* Load shift vector for this list */
 158         i_shift_offset   = DIM*shiftidx[iidx];
 159
 160         /* Load limits for loop over neighbors */
 161         j_index_start    = jindex[iidx];
 162         j_index_end      = jindex[iidx+1];
 163
 164         /* Get outer coordinate index */
 165         inr              = iinr[iidx];
 166         i_coord_offset   = DIM*inr;
 167
 168         /* Load i particle coords and add shift vector */
 169         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 170                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 171
 172         fix0             = _fjsp_setzero_v2r8();
 173         fiy0             = _fjsp_setzero_v2r8();
 174         fiz0             = _fjsp_setzero_v2r8();
 175         fix1             = _fjsp_setzero_v2r8();
 176         fiy1             = _fjsp_setzero_v2r8();
 177         fiz1             = _fjsp_setzero_v2r8();
 178         fix2             = _fjsp_setzero_v2r8();
 179         fiy2             = _fjsp_setzero_v2r8();
 180         fiz2             = _fjsp_setzero_v2r8();
 181         fix3             = _fjsp_setzero_v2r8();
 182         fiy3             = _fjsp_setzero_v2r8();
 183         fiz3             = _fjsp_setzero_v2r8();
 184
 185         /* Reset potential sums */
 186         velecsum         = _fjsp_setzero_v2r8();
 187         vvdwsum          = _fjsp_setzero_v2r8();
 188
 189         /* Start inner kernel loop */
 190         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 191         {
 192
 193             /* Get j neighbor index, and coordinate index */
 194             jnrA             = jjnr[jidx];
 195             jnrB             = jjnr[jidx+1];
 196             j_coord_offsetA  = DIM*jnrA;
 197             j_coord_offsetB  = DIM*jnrB;
 198
 199             /* load j atom coordinates */
 200             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 201                                               &jx0,&jy0,&jz0);
 202
 203             /* Calculate displacement vector */
 204             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 205             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 206             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 207             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 208             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 209             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 210             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 211             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 212             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 213             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 214             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 215             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 216
 217             /* Calculate squared distance and things based on it */
 218             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 219             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 220             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 221             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 222
 223             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 224             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 225             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 226             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 227
 228             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 229             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 230             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 231
 232             /* Load parameters for j particles */
 233             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 234             vdwjidx0A        = 2*vdwtype[jnrA+0];
 235             vdwjidx0B        = 2*vdwtype[jnrB+0];
 236
 237             fjx0             = _fjsp_setzero_v2r8();
 238             fjy0             = _fjsp_setzero_v2r8();
 239             fjz0             = _fjsp_setzero_v2r8();
 240
 241             /**************************
 242              * CALCULATE INTERACTIONS *
 243              **************************/
 244
 245             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 246
 247             /* Compute parameters for interactions between i and j atoms */
 248             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 249                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 250
 251             /* Calculate table index by multiplying r with table scale and truncate to integer */
 252             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 253             itab_tmp         = _fjsp_dtox_v2r8(rt);
 254             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 255             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 256             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 257
 258             vfconv.i[0]     *= 8;
 259             vfconv.i[1]     *= 8;
 260
 261             /* CUBIC SPLINE TABLE DISPERSION */
 262             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 263             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 264             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 265             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 266             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 267             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 268             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 269             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 270             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 271             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 272             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 273
 274             /* CUBIC SPLINE TABLE REPULSION */
 275             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 276             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 277             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 278             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 279             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 280             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 281             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 282             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 283             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 284             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 285             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 286             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 287             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 291
 292             fscal            = fvdw;
 293
 294             /* Update vectorial force */
 295             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 296             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 297             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 298
 299             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 300             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 301             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 302
 303             /**************************
 304              * CALCULATE INTERACTIONS *
 305              **************************/
 306
 307             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 308             {
 309
 310             /* Compute parameters for interactions between i and j atoms */
 311             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 312
 313             /* REACTION-FIELD ELECTROSTATICS */
 314             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 315             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 316
 317             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 318
 319             /* Update potential sum for this i atom from the interaction with this j atom. */
 320             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 321             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 322
 323             fscal            = felec;
 324
 325             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 326
 327             /* Update vectorial force */
 328             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 329             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 330             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 331
 332             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 333             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 334             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 335
 336             }
 337
 338             /**************************
 339              * CALCULATE INTERACTIONS *
 340              **************************/
 341
 342             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 343             {
 344
 345             /* Compute parameters for interactions between i and j atoms */
 346             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 347
 348             /* REACTION-FIELD ELECTROSTATICS */
 349             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 350             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 351
 352             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 356             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 357
 358             fscal            = felec;
 359
 360             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 361
 362             /* Update vectorial force */
 363             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 364             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 365             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 366
 367             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 368             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 369             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 370
 371             }
 372
 373             /**************************
 374              * CALCULATE INTERACTIONS *
 375              **************************/
 376
 377             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 378             {
 379
 380             /* Compute parameters for interactions between i and j atoms */
 381             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 382
 383             /* REACTION-FIELD ELECTROSTATICS */
 384             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 385             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 386
 387             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 388
 389             /* Update potential sum for this i atom from the interaction with this j atom. */
 390             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 391             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 396
 397             /* Update vectorial force */
 398             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 399             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 400             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 401
 402             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 403             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 404             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 405
 406             }
 407
 408             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 409
 410             /* Inner loop uses 179 flops */
 411         }
 412
 413         if(jidx<j_index_end)
 414         {
 415
 416             jnrA             = jjnr[jidx];
 417             j_coord_offsetA  = DIM*jnrA;
 418
 419             /* load j atom coordinates */
 420             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 421                                               &jx0,&jy0,&jz0);
 422
 423             /* Calculate displacement vector */
 424             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 425             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 426             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 427             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 428             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 429             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 430             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 431             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 432             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 433             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 434             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 435             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 436
 437             /* Calculate squared distance and things based on it */
 438             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 439             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 440             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 441             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 442
 443             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 444             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 445             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 446             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 447
 448             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 449             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 450             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 451
 452             /* Load parameters for j particles */
 453             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 454             vdwjidx0A        = 2*vdwtype[jnrA+0];
 455
 456             fjx0             = _fjsp_setzero_v2r8();
 457             fjy0             = _fjsp_setzero_v2r8();
 458             fjz0             = _fjsp_setzero_v2r8();
 459
 460             /**************************
 461              * CALCULATE INTERACTIONS *
 462              **************************/
 463
 464             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 465
 466             /* Compute parameters for interactions between i and j atoms */
 467             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 468
 469             /* Calculate table index by multiplying r with table scale and truncate to integer */
 470             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 471             itab_tmp         = _fjsp_dtox_v2r8(rt);
 472             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 473             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 474             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 475
 476             vfconv.i[0]     *= 8;
 477             vfconv.i[1]     *= 8;
 478
 479             /* CUBIC SPLINE TABLE DISPERSION */
 480             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 481             F                = _fjsp_setzero_v2r8();
 482             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 483             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 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             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 489             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 490             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 491
 492             /* CUBIC SPLINE TABLE REPULSION */
 493             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 494             F                = _fjsp_setzero_v2r8();
 495             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 496             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 497             H                = _fjsp_setzero_v2r8();
 498             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 499             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 500             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 501             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 502             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 503             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 504             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 505             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 506
 507             /* Update potential sum for this i atom from the interaction with this j atom. */
 508             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 509             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 510
 511             fscal            = fvdw;
 512
 513             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 514
 515             /* Update vectorial force */
 516             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 517             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 518             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 519
 520             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 521             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 522             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 523
 524             /**************************
 525              * CALCULATE INTERACTIONS *
 526              **************************/
 527
 528             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 529             {
 530
 531             /* Compute parameters for interactions between i and j atoms */
 532             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 533
 534             /* REACTION-FIELD ELECTROSTATICS */
 535             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 536             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 537
 538             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 539
 540             /* Update potential sum for this i atom from the interaction with this j atom. */
 541             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 542             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 543             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 544
 545             fscal            = felec;
 546
 547             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 548
 549             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 550
 551             /* Update vectorial force */
 552             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 553             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 554             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 555
 556             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 557             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 558             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 559
 560             }
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 567             {
 568
 569             /* Compute parameters for interactions between i and j atoms */
 570             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 571
 572             /* REACTION-FIELD ELECTROSTATICS */
 573             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 574             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 575
 576             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 577
 578             /* Update potential sum for this i atom from the interaction with this j atom. */
 579             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 580             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 581             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 582
 583             fscal            = felec;
 584
 585             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 586
 587             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 588
 589             /* Update vectorial force */
 590             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 591             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 592             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 593
 594             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 595             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 596             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 597
 598             }
 599
 600             /**************************
 601              * CALCULATE INTERACTIONS *
 602              **************************/
 603
 604             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 605             {
 606
 607             /* Compute parameters for interactions between i and j atoms */
 608             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 609
 610             /* REACTION-FIELD ELECTROSTATICS */
 611             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 612             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 613
 614             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 615
 616             /* Update potential sum for this i atom from the interaction with this j atom. */
 617             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 618             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 619             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 620
 621             fscal            = felec;
 622
 623             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 624
 625             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 626
 627             /* Update vectorial force */
 628             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 629             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 630             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 631
 632             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 633             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 634             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 635
 636             }
 637
 638             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 639
 640             /* Inner loop uses 179 flops */
 641         }
 642
 643         /* End of innermost loop */
 644
 645         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 646                                               f+i_coord_offset,fshift+i_shift_offset);
 647
 648         ggid                        = gid[iidx];
 649         /* Update potential energies */
 650         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 651         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 652
 653         /* Increment number of inner iterations */
 654         inneriter                  += j_index_end - j_index_start;
 655
 656         /* Outer loop uses 26 flops */
 657     }
 658
 659     /* Increment number of outer iterations */
 660     outeriter        += nri;
 661
 662     /* Update outer/inner flops */
 663
 664     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*179);
 665 }
 666 /*
 667  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 668  * Electrostatics interaction: ReactionField
 669  * VdW interaction:            CubicSplineTable
 670  * Geometry:                   Water4-Particle
 671  * Calculate force/pot:        Force
 672  */
 673 void
 674 nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 675                     (t_nblist * gmx_restrict                nlist,
 676                      rvec * gmx_restrict                    xx,
 677                      rvec * gmx_restrict                    ff,
 678                      t_forcerec * gmx_restrict              fr,
 679                      t_mdatoms * gmx_restrict               mdatoms,
 680                      nb_kernel_data_t * gmx_restrict        kernel_data,
 681                      t_nrnb * gmx_restrict                  nrnb)
 682 {
 683     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 684      * just 0 for non-waters.
 685      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 686      * jnr indices corresponding to data put in the four positions in the SIMD register.
 687      */
 688     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 689     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 690     int              jnrA,jnrB;
 691     int              j_coord_offsetA,j_coord_offsetB;
 692     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 693     real             rcutoff_scalar;
 694     real             *shiftvec,*fshift,*x,*f;
 695     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 696     int              vdwioffset0;
 697     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 698     int              vdwioffset1;
 699     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 700     int              vdwioffset2;
 701     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 702     int              vdwioffset3;
 703     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 704     int              vdwjidx0A,vdwjidx0B;
 705     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 706     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 707     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 708     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 709     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 710     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 711     real             *charge;
 712     int              nvdwtype;
 713     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 714     int              *vdwtype;
 715     real             *vdwparam;
 716     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 717     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 718     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 719     real             *vftab;
 720     _fjsp_v2r8       itab_tmp;
 721     _fjsp_v2r8       dummy_mask,cutoff_mask;
 722     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 723     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 724     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 725
 726     x                = xx[0];
 727     f                = ff[0];
 728
 729     nri              = nlist->nri;
 730     iinr             = nlist->iinr;
 731     jindex           = nlist->jindex;
 732     jjnr             = nlist->jjnr;
 733     shiftidx         = nlist->shift;
 734     gid              = nlist->gid;
 735     shiftvec         = fr->shift_vec[0];
 736     fshift           = fr->fshift[0];
 737     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 738     charge           = mdatoms->chargeA;
 739     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 740     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 741     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 742     nvdwtype         = fr->ntype;
 743     vdwparam         = fr->nbfp;
 744     vdwtype          = mdatoms->typeA;
 745
 746     vftab            = kernel_data->table_vdw->data;
 747     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 748
 749     /* Setup water-specific parameters */
 750     inr              = nlist->iinr[0];
 751     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 752     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 753     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 754     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 755
 756     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 757     rcutoff_scalar   = fr->rcoulomb;
 758     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 759     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 760
 761     /* Avoid stupid compiler warnings */
 762     jnrA = jnrB = 0;
 763     j_coord_offsetA = 0;
 764     j_coord_offsetB = 0;
 765
 766     outeriter        = 0;
 767     inneriter        = 0;
 768
 769     /* Start outer loop over neighborlists */
 770     for(iidx=0; iidx<nri; iidx++)
 771     {
 772         /* Load shift vector for this list */
 773         i_shift_offset   = DIM*shiftidx[iidx];
 774
 775         /* Load limits for loop over neighbors */
 776         j_index_start    = jindex[iidx];
 777         j_index_end      = jindex[iidx+1];
 778
 779         /* Get outer coordinate index */
 780         inr              = iinr[iidx];
 781         i_coord_offset   = DIM*inr;
 782
 783         /* Load i particle coords and add shift vector */
 784         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 785                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 786
 787         fix0             = _fjsp_setzero_v2r8();
 788         fiy0             = _fjsp_setzero_v2r8();
 789         fiz0             = _fjsp_setzero_v2r8();
 790         fix1             = _fjsp_setzero_v2r8();
 791         fiy1             = _fjsp_setzero_v2r8();
 792         fiz1             = _fjsp_setzero_v2r8();
 793         fix2             = _fjsp_setzero_v2r8();
 794         fiy2             = _fjsp_setzero_v2r8();
 795         fiz2             = _fjsp_setzero_v2r8();
 796         fix3             = _fjsp_setzero_v2r8();
 797         fiy3             = _fjsp_setzero_v2r8();
 798         fiz3             = _fjsp_setzero_v2r8();
 799
 800         /* Start inner kernel loop */
 801         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 802         {
 803
 804             /* Get j neighbor index, and coordinate index */
 805             jnrA             = jjnr[jidx];
 806             jnrB             = jjnr[jidx+1];
 807             j_coord_offsetA  = DIM*jnrA;
 808             j_coord_offsetB  = DIM*jnrB;
 809
 810             /* load j atom coordinates */
 811             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 812                                               &jx0,&jy0,&jz0);
 813
 814             /* Calculate displacement vector */
 815             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 816             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 817             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 818             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 819             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 820             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 821             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 822             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 823             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 824             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 825             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 826             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 827
 828             /* Calculate squared distance and things based on it */
 829             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 830             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 831             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 832             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 833
 834             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 835             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 836             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 837             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 838
 839             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 840             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 841             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 842
 843             /* Load parameters for j particles */
 844             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 845             vdwjidx0A        = 2*vdwtype[jnrA+0];
 846             vdwjidx0B        = 2*vdwtype[jnrB+0];
 847
 848             fjx0             = _fjsp_setzero_v2r8();
 849             fjy0             = _fjsp_setzero_v2r8();
 850             fjz0             = _fjsp_setzero_v2r8();
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 857
 858             /* Compute parameters for interactions between i and j atoms */
 859             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 860                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 861
 862             /* Calculate table index by multiplying r with table scale and truncate to integer */
 863             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 864             itab_tmp         = _fjsp_dtox_v2r8(rt);
 865             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 866             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 867             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 868
 869             vfconv.i[0]     *= 8;
 870             vfconv.i[1]     *= 8;
 871
 872             /* CUBIC SPLINE TABLE DISPERSION */
 873             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 874             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 875             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 876             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 877             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 878             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 879             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 880             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 881             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 882
 883             /* CUBIC SPLINE TABLE REPULSION */
 884             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 885             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 886             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 887             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 888             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 889             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 890             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 891             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 892             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 893             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 894
 895             fscal            = fvdw;
 896
 897             /* Update vectorial force */
 898             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 899             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 900             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 901
 902             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 903             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 904             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 905
 906             /**************************
 907              * CALCULATE INTERACTIONS *
 908              **************************/
 909
 910             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 911             {
 912
 913             /* Compute parameters for interactions between i and j atoms */
 914             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 915
 916             /* REACTION-FIELD ELECTROSTATICS */
 917             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 918
 919             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 920
 921             fscal            = felec;
 922
 923             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 924
 925             /* Update vectorial force */
 926             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 927             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 928             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 929
 930             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 931             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 932             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 933
 934             }
 935
 936             /**************************
 937              * CALCULATE INTERACTIONS *
 938              **************************/
 939
 940             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 941             {
 942
 943             /* Compute parameters for interactions between i and j atoms */
 944             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 945
 946             /* REACTION-FIELD ELECTROSTATICS */
 947             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 948
 949             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 950
 951             fscal            = felec;
 952
 953             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 954
 955             /* Update vectorial force */
 956             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 957             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 958             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 959
 960             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 961             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 962             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 963
 964             }
 965
 966             /**************************
 967              * CALCULATE INTERACTIONS *
 968              **************************/
 969
 970             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 971             {
 972
 973             /* Compute parameters for interactions between i and j atoms */
 974             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 975
 976             /* REACTION-FIELD ELECTROSTATICS */
 977             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 978
 979             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 980
 981             fscal            = felec;
 982
 983             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 984
 985             /* Update vectorial force */
 986             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 987             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 988             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 989
 990             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 991             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 992             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 993
 994             }
 995
 996             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 997
 998             /* Inner loop uses 153 flops */
 999         }
1000
1001         if(jidx<j_index_end)
1002         {
1003
1004             jnrA             = jjnr[jidx];
1005             j_coord_offsetA  = DIM*jnrA;
1006
1007             /* load j atom coordinates */
1008             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1009                                               &jx0,&jy0,&jz0);
1010
1011             /* Calculate displacement vector */
1012             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1013             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1014             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1015             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1016             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1017             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1018             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1019             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1020             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1021             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1022             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1023             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1024
1025             /* Calculate squared distance and things based on it */
1026             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1027             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1028             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1029             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1030
1031             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1032             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1033             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1034             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1035
1036             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1037             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1038             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1039
1040             /* Load parameters for j particles */
1041             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1042             vdwjidx0A        = 2*vdwtype[jnrA+0];
1043
1044             fjx0             = _fjsp_setzero_v2r8();
1045             fjy0             = _fjsp_setzero_v2r8();
1046             fjz0             = _fjsp_setzero_v2r8();
1047
1048             /**************************
1049              * CALCULATE INTERACTIONS *
1050              **************************/
1051
1052             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1053
1054             /* Compute parameters for interactions between i and j atoms */
1055             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1056
1057             /* Calculate table index by multiplying r with table scale and truncate to integer */
1058             rt               = _fjsp_mul_v2r8(r00,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]     *= 8;
1065             vfconv.i[1]     *= 8;
1066
1067             /* CUBIC SPLINE TABLE DISPERSION */
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(H,vfeps,G),F);
1075             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1076             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1077
1078             /* CUBIC SPLINE TABLE REPULSION */
1079             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1080             F                = _fjsp_setzero_v2r8();
1081             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1082             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1083             H                = _fjsp_setzero_v2r8();
1084             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1085             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1086             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1087             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1088             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1089
1090             fscal            = fvdw;
1091
1092             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1093
1094             /* Update vectorial force */
1095             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1096             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1097             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1098
1099             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1100             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1101             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1108             {
1109
1110             /* Compute parameters for interactions between i and j atoms */
1111             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1112
1113             /* REACTION-FIELD ELECTROSTATICS */
1114             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1115
1116             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1117
1118             fscal            = felec;
1119
1120             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1121
1122             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1123
1124             /* Update vectorial force */
1125             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1126             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1127             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1128
1129             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1130             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1131             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1132
1133             }
1134
1135             /**************************
1136              * CALCULATE INTERACTIONS *
1137              **************************/
1138
1139             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1140             {
1141
1142             /* Compute parameters for interactions between i and j atoms */
1143             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1144
1145             /* REACTION-FIELD ELECTROSTATICS */
1146             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1147
1148             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1149
1150             fscal            = felec;
1151
1152             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1153
1154             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1155
1156             /* Update vectorial force */
1157             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1158             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1159             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1160
1161             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1162             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1163             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1164
1165             }
1166
1167             /**************************
1168              * CALCULATE INTERACTIONS *
1169              **************************/
1170
1171             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1172             {
1173
1174             /* Compute parameters for interactions between i and j atoms */
1175             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1176
1177             /* REACTION-FIELD ELECTROSTATICS */
1178             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1179
1180             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1181
1182             fscal            = felec;
1183
1184             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1185
1186             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1187
1188             /* Update vectorial force */
1189             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1190             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1191             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1192
1193             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1194             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1195             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1196
1197             }
1198
1199             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1200
1201             /* Inner loop uses 153 flops */
1202         }
1203
1204         /* End of innermost loop */
1205
1206         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1207                                               f+i_coord_offset,fshift+i_shift_offset);
1208
1209         /* Increment number of inner iterations */
1210         inneriter                  += j_index_end - j_index_start;
1211
1212         /* Outer loop uses 24 flops */
1213     }
1214
1215     /* Increment number of outer iterations */
1216     outeriter        += nri;
1217
1218     /* Update outer/inner flops */
1219
1220     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*153);
1221 }