src/gromacs/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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_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_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 468                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 469
 470             /* Calculate table index by multiplying r with table scale and truncate to integer */
 471             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 472             itab_tmp         = _fjsp_dtox_v2r8(rt);
 473             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 474             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 475             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 476
 477             vfconv.i[0]     *= 8;
 478             vfconv.i[1]     *= 8;
 479
 480             /* CUBIC SPLINE TABLE DISPERSION */
 481             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 482             F                = _fjsp_setzero_v2r8();
 483             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 484             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 485             H                = _fjsp_setzero_v2r8();
 486             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 487             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 488             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 489             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 490             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 491             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 492
 493             /* CUBIC SPLINE TABLE REPULSION */
 494             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 495             F                = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 497             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 498             H                = _fjsp_setzero_v2r8();
 499             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 500             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 501             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 502             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 503             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 504             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 505             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 506             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 507
 508             /* Update potential sum for this i atom from the interaction with this j atom. */
 509             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 510             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 511
 512             fscal            = fvdw;
 513
 514             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 515
 516             /* Update vectorial force */
 517             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 518             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 519             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 520
 521             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 522             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 523             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 524
 525             /**************************
 526              * CALCULATE INTERACTIONS *
 527              **************************/
 528
 529             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 530             {
 531
 532             /* Compute parameters for interactions between i and j atoms */
 533             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 534
 535             /* REACTION-FIELD ELECTROSTATICS */
 536             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 537             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 538
 539             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 540
 541             /* Update potential sum for this i atom from the interaction with this j atom. */
 542             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 543             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 544             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 545
 546             fscal            = felec;
 547
 548             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 549
 550             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 551
 552             /* Update vectorial force */
 553             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 554             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 555             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 556
 557             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 558             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 559             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 560
 561             }
 562
 563             /**************************
 564              * CALCULATE INTERACTIONS *
 565              **************************/
 566
 567             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 568             {
 569
 570             /* Compute parameters for interactions between i and j atoms */
 571             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 572
 573             /* REACTION-FIELD ELECTROSTATICS */
 574             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 575             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 576
 577             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 578
 579             /* Update potential sum for this i atom from the interaction with this j atom. */
 580             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 581             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 582             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 583
 584             fscal            = felec;
 585
 586             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 587
 588             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 589
 590             /* Update vectorial force */
 591             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 592             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 593             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 594
 595             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 596             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 597             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 598
 599             }
 600
 601             /**************************
 602              * CALCULATE INTERACTIONS *
 603              **************************/
 604
 605             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 606             {
 607
 608             /* Compute parameters for interactions between i and j atoms */
 609             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 610
 611             /* REACTION-FIELD ELECTROSTATICS */
 612             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 613             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 614
 615             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 616
 617             /* Update potential sum for this i atom from the interaction with this j atom. */
 618             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 619             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 620             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 621
 622             fscal            = felec;
 623
 624             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 625
 626             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 627
 628             /* Update vectorial force */
 629             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 630             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 631             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 632
 633             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 634             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 635             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 636
 637             }
 638
 639             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 640
 641             /* Inner loop uses 179 flops */
 642         }
 643
 644         /* End of innermost loop */
 645
 646         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 647                                               f+i_coord_offset,fshift+i_shift_offset);
 648
 649         ggid                        = gid[iidx];
 650         /* Update potential energies */
 651         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 652         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 653
 654         /* Increment number of inner iterations */
 655         inneriter                  += j_index_end - j_index_start;
 656
 657         /* Outer loop uses 26 flops */
 658     }
 659
 660     /* Increment number of outer iterations */
 661     outeriter        += nri;
 662
 663     /* Update outer/inner flops */
 664
 665     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*179);
 666 }
 667 /*
 668  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 669  * Electrostatics interaction: ReactionField
 670  * VdW interaction:            CubicSplineTable
 671  * Geometry:                   Water4-Particle
 672  * Calculate force/pot:        Force
 673  */
 674 void
 675 nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 676                     (t_nblist                    * gmx_restrict       nlist,
 677                      rvec                        * gmx_restrict          xx,
 678                      rvec                        * gmx_restrict          ff,
 679                      t_forcerec                  * gmx_restrict          fr,
 680                      t_mdatoms                   * gmx_restrict     mdatoms,
 681                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 682                      t_nrnb                      * gmx_restrict        nrnb)
 683 {
 684     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 685      * just 0 for non-waters.
 686      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 687      * jnr indices corresponding to data put in the four positions in the SIMD register.
 688      */
 689     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 690     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 691     int              jnrA,jnrB;
 692     int              j_coord_offsetA,j_coord_offsetB;
 693     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 694     real             rcutoff_scalar;
 695     real             *shiftvec,*fshift,*x,*f;
 696     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 697     int              vdwioffset0;
 698     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 699     int              vdwioffset1;
 700     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 701     int              vdwioffset2;
 702     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 703     int              vdwioffset3;
 704     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 705     int              vdwjidx0A,vdwjidx0B;
 706     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 707     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 708     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 709     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 710     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 711     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 712     real             *charge;
 713     int              nvdwtype;
 714     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 715     int              *vdwtype;
 716     real             *vdwparam;
 717     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 718     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 719     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 720     real             *vftab;
 721     _fjsp_v2r8       itab_tmp;
 722     _fjsp_v2r8       dummy_mask,cutoff_mask;
 723     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 724     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 725     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 726
 727     x                = xx[0];
 728     f                = ff[0];
 729
 730     nri              = nlist->nri;
 731     iinr             = nlist->iinr;
 732     jindex           = nlist->jindex;
 733     jjnr             = nlist->jjnr;
 734     shiftidx         = nlist->shift;
 735     gid              = nlist->gid;
 736     shiftvec         = fr->shift_vec[0];
 737     fshift           = fr->fshift[0];
 738     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 739     charge           = mdatoms->chargeA;
 740     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 741     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 742     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 743     nvdwtype         = fr->ntype;
 744     vdwparam         = fr->nbfp;
 745     vdwtype          = mdatoms->typeA;
 746
 747     vftab            = kernel_data->table_vdw->data;
 748     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 749
 750     /* Setup water-specific parameters */
 751     inr              = nlist->iinr[0];
 752     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 753     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 754     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 755     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 756
 757     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 758     rcutoff_scalar   = fr->rcoulomb;
 759     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 760     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 761
 762     /* Avoid stupid compiler warnings */
 763     jnrA = jnrB = 0;
 764     j_coord_offsetA = 0;
 765     j_coord_offsetB = 0;
 766
 767     outeriter        = 0;
 768     inneriter        = 0;
 769
 770     /* Start outer loop over neighborlists */
 771     for(iidx=0; iidx<nri; iidx++)
 772     {
 773         /* Load shift vector for this list */
 774         i_shift_offset   = DIM*shiftidx[iidx];
 775
 776         /* Load limits for loop over neighbors */
 777         j_index_start    = jindex[iidx];
 778         j_index_end      = jindex[iidx+1];
 779
 780         /* Get outer coordinate index */
 781         inr              = iinr[iidx];
 782         i_coord_offset   = DIM*inr;
 783
 784         /* Load i particle coords and add shift vector */
 785         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 786                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 787
 788         fix0             = _fjsp_setzero_v2r8();
 789         fiy0             = _fjsp_setzero_v2r8();
 790         fiz0             = _fjsp_setzero_v2r8();
 791         fix1             = _fjsp_setzero_v2r8();
 792         fiy1             = _fjsp_setzero_v2r8();
 793         fiz1             = _fjsp_setzero_v2r8();
 794         fix2             = _fjsp_setzero_v2r8();
 795         fiy2             = _fjsp_setzero_v2r8();
 796         fiz2             = _fjsp_setzero_v2r8();
 797         fix3             = _fjsp_setzero_v2r8();
 798         fiy3             = _fjsp_setzero_v2r8();
 799         fiz3             = _fjsp_setzero_v2r8();
 800
 801         /* Start inner kernel loop */
 802         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 803         {
 804
 805             /* Get j neighbor index, and coordinate index */
 806             jnrA             = jjnr[jidx];
 807             jnrB             = jjnr[jidx+1];
 808             j_coord_offsetA  = DIM*jnrA;
 809             j_coord_offsetB  = DIM*jnrB;
 810
 811             /* load j atom coordinates */
 812             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 813                                               &jx0,&jy0,&jz0);
 814
 815             /* Calculate displacement vector */
 816             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 817             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 818             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 819             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 820             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 821             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 822             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 823             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 824             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 825             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 826             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 827             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 828
 829             /* Calculate squared distance and things based on it */
 830             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 831             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 832             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 833             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 834
 835             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 836             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 837             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 838             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 839
 840             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 841             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 842             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 843
 844             /* Load parameters for j particles */
 845             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 846             vdwjidx0A        = 2*vdwtype[jnrA+0];
 847             vdwjidx0B        = 2*vdwtype[jnrB+0];
 848
 849             fjx0             = _fjsp_setzero_v2r8();
 850             fjy0             = _fjsp_setzero_v2r8();
 851             fjz0             = _fjsp_setzero_v2r8();
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 858
 859             /* Compute parameters for interactions between i and j atoms */
 860             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 861                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 862
 863             /* Calculate table index by multiplying r with table scale and truncate to integer */
 864             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 865             itab_tmp         = _fjsp_dtox_v2r8(rt);
 866             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 867             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 868             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 869
 870             vfconv.i[0]     *= 8;
 871             vfconv.i[1]     *= 8;
 872
 873             /* CUBIC SPLINE TABLE DISPERSION */
 874             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 875             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 876             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 877             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 878             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 879             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 880             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 881             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 882             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 883
 884             /* CUBIC SPLINE TABLE REPULSION */
 885             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 886             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 887             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 888             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 889             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 890             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 891             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 892             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 893             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 894             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 895
 896             fscal            = fvdw;
 897
 898             /* Update vectorial force */
 899             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 900             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 901             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 902
 903             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 904             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 905             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 906
 907             /**************************
 908              * CALCULATE INTERACTIONS *
 909              **************************/
 910
 911             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 912             {
 913
 914             /* Compute parameters for interactions between i and j atoms */
 915             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 916
 917             /* REACTION-FIELD ELECTROSTATICS */
 918             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 919
 920             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 921
 922             fscal            = felec;
 923
 924             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 925
 926             /* Update vectorial force */
 927             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 928             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 929             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 930
 931             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 932             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 933             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 934
 935             }
 936
 937             /**************************
 938              * CALCULATE INTERACTIONS *
 939              **************************/
 940
 941             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 942             {
 943
 944             /* Compute parameters for interactions between i and j atoms */
 945             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 946
 947             /* REACTION-FIELD ELECTROSTATICS */
 948             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 949
 950             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 951
 952             fscal            = felec;
 953
 954             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 955
 956             /* Update vectorial force */
 957             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 958             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 959             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 960
 961             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 962             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 963             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 964
 965             }
 966
 967             /**************************
 968              * CALCULATE INTERACTIONS *
 969              **************************/
 970
 971             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 972             {
 973
 974             /* Compute parameters for interactions between i and j atoms */
 975             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 976
 977             /* REACTION-FIELD ELECTROSTATICS */
 978             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 979
 980             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 981
 982             fscal            = felec;
 983
 984             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 985
 986             /* Update vectorial force */
 987             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 988             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 989             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 990
 991             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 992             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 993             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 994
 995             }
 996
 997             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 998
 999             /* Inner loop uses 153 flops */
1000         }
1001
1002         if(jidx<j_index_end)
1003         {
1004
1005             jnrA             = jjnr[jidx];
1006             j_coord_offsetA  = DIM*jnrA;
1007
1008             /* load j atom coordinates */
1009             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1010                                               &jx0,&jy0,&jz0);
1011
1012             /* Calculate displacement vector */
1013             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1014             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1015             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1016             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1017             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1018             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1019             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1020             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1021             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1022             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1023             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1024             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1025
1026             /* Calculate squared distance and things based on it */
1027             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1028             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1029             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1030             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1031
1032             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1033             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1034             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1035             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1036
1037             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1038             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1039             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1040
1041             /* Load parameters for j particles */
1042             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1043             vdwjidx0A        = 2*vdwtype[jnrA+0];
1044
1045             fjx0             = _fjsp_setzero_v2r8();
1046             fjy0             = _fjsp_setzero_v2r8();
1047             fjz0             = _fjsp_setzero_v2r8();
1048
1049             /**************************
1050              * CALCULATE INTERACTIONS *
1051              **************************/
1052
1053             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1054
1055             /* Compute parameters for interactions between i and j atoms */
1056             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1057                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1058
1059             /* Calculate table index by multiplying r with table scale and truncate to integer */
1060             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1061             itab_tmp         = _fjsp_dtox_v2r8(rt);
1062             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1063             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1064             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1065
1066             vfconv.i[0]     *= 8;
1067             vfconv.i[1]     *= 8;
1068
1069             /* CUBIC SPLINE TABLE DISPERSION */
1070             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1071             F                = _fjsp_setzero_v2r8();
1072             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1073             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1074             H                = _fjsp_setzero_v2r8();
1075             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1076             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1077             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1078             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1079
1080             /* CUBIC SPLINE TABLE REPULSION */
1081             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1082             F                = _fjsp_setzero_v2r8();
1083             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1084             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1085             H                = _fjsp_setzero_v2r8();
1086             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1087             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1088             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1089             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1090             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1091
1092             fscal            = fvdw;
1093
1094             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1095
1096             /* Update vectorial force */
1097             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1098             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1099             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1100
1101             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1102             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1103             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1104
1105             /**************************
1106              * CALCULATE INTERACTIONS *
1107              **************************/
1108
1109             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1110             {
1111
1112             /* Compute parameters for interactions between i and j atoms */
1113             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1114
1115             /* REACTION-FIELD ELECTROSTATICS */
1116             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1117
1118             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1119
1120             fscal            = felec;
1121
1122             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1123
1124             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1125
1126             /* Update vectorial force */
1127             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1128             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1129             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1130
1131             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1132             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1133             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1134
1135             }
1136
1137             /**************************
1138              * CALCULATE INTERACTIONS *
1139              **************************/
1140
1141             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1142             {
1143
1144             /* Compute parameters for interactions between i and j atoms */
1145             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1146
1147             /* REACTION-FIELD ELECTROSTATICS */
1148             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1149
1150             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1151
1152             fscal            = felec;
1153
1154             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1155
1156             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1157
1158             /* Update vectorial force */
1159             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1160             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1161             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1162
1163             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1164             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1165             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1166
1167             }
1168
1169             /**************************
1170              * CALCULATE INTERACTIONS *
1171              **************************/
1172
1173             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1174             {
1175
1176             /* Compute parameters for interactions between i and j atoms */
1177             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1178
1179             /* REACTION-FIELD ELECTROSTATICS */
1180             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1181
1182             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1183
1184             fscal            = felec;
1185
1186             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1187
1188             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1189
1190             /* Update vectorial force */
1191             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1192             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1193             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1194
1195             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1196             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1197             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1198
1199             }
1200
1201             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1202
1203             /* Inner loop uses 153 flops */
1204         }
1205
1206         /* End of innermost loop */
1207
1208         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1209                                               f+i_coord_offset,fshift+i_shift_offset);
1210
1211         /* Increment number of inner iterations */
1212         inneriter                  += j_index_end - j_index_start;
1213
1214         /* Outer loop uses 24 flops */
1215     }
1216
1217     /* Increment number of outer iterations */
1218     outeriter        += nri;
1219
1220     /* Update outer/inner flops */
1221
1222     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*153);
1223 }