src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 118     charge           = mdatoms->chargeA;
 119     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 120     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 121     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 132     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 133     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 137     rcutoff_scalar   = fr->ic->rcoulomb;
 138     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 139     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 140
 141     /* Avoid stupid compiler warnings */
 142     jnrA = jnrB = 0;
 143     j_coord_offsetA = 0;
 144     j_coord_offsetB = 0;
 145
 146     outeriter        = 0;
 147     inneriter        = 0;
 148
 149     /* Start outer loop over neighborlists */
 150     for(iidx=0; iidx<nri; iidx++)
 151     {
 152         /* Load shift vector for this list */
 153         i_shift_offset   = DIM*shiftidx[iidx];
 154
 155         /* Load limits for loop over neighbors */
 156         j_index_start    = jindex[iidx];
 157         j_index_end      = jindex[iidx+1];
 158
 159         /* Get outer coordinate index */
 160         inr              = iinr[iidx];
 161         i_coord_offset   = DIM*inr;
 162
 163         /* Load i particle coords and add shift vector */
 164         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 165                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 166
 167         fix0             = _fjsp_setzero_v2r8();
 168         fiy0             = _fjsp_setzero_v2r8();
 169         fiz0             = _fjsp_setzero_v2r8();
 170         fix1             = _fjsp_setzero_v2r8();
 171         fiy1             = _fjsp_setzero_v2r8();
 172         fiz1             = _fjsp_setzero_v2r8();
 173         fix2             = _fjsp_setzero_v2r8();
 174         fiy2             = _fjsp_setzero_v2r8();
 175         fiz2             = _fjsp_setzero_v2r8();
 176
 177         /* Reset potential sums */
 178         velecsum         = _fjsp_setzero_v2r8();
 179         vvdwsum          = _fjsp_setzero_v2r8();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 193                                               &jx0,&jy0,&jz0);
 194
 195             /* Calculate displacement vector */
 196             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 197             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 198             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 199             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 200             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 201             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 202             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 203             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 204             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 205
 206             /* Calculate squared distance and things based on it */
 207             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 208             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 209             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 210
 211             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 212             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 213             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 214
 215             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 216             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 217             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 218
 219             /* Load parameters for j particles */
 220             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 221             vdwjidx0A        = 2*vdwtype[jnrA+0];
 222             vdwjidx0B        = 2*vdwtype[jnrB+0];
 223
 224             fjx0             = _fjsp_setzero_v2r8();
 225             fjy0             = _fjsp_setzero_v2r8();
 226             fjz0             = _fjsp_setzero_v2r8();
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 233             {
 234
 235             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 239             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 240                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 241
 242             /* Calculate table index by multiplying r with table scale and truncate to integer */
 243             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 244             itab_tmp         = _fjsp_dtox_v2r8(rt);
 245             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 246             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 247             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 248
 249             vfconv.i[0]     *= 8;
 250             vfconv.i[1]     *= 8;
 251
 252             /* REACTION-FIELD ELECTROSTATICS */
 253             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 254             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 255
 256             /* CUBIC SPLINE TABLE DISPERSION */
 257             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 258             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 259             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 260             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 261             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 262             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 263             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 264             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 265             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 266             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 267             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 268
 269             /* CUBIC SPLINE TABLE REPULSION */
 270             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 271             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 272             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 273             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 274             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 276             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 277             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 278             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 279             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 280             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 281             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 282             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 283
 284             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 285
 286             /* Update potential sum for this i atom from the interaction with this j atom. */
 287             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 288             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 289             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 290             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 291
 292             fscal            = _fjsp_add_v2r8(felec,fvdw);
 293
 294             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 295
 296             /* Update vectorial force */
 297             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 298             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 299             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 300
 301             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 302             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 303             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 304
 305             }
 306
 307             /**************************
 308              * CALCULATE INTERACTIONS *
 309              **************************/
 310
 311             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 312             {
 313
 314             /* Compute parameters for interactions between i and j atoms */
 315             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 316
 317             /* REACTION-FIELD ELECTROSTATICS */
 318             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 319             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 320
 321             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 322
 323             /* Update potential sum for this i atom from the interaction with this j atom. */
 324             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 325             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 326
 327             fscal            = felec;
 328
 329             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 330
 331             /* Update vectorial force */
 332             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 333             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 334             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 335
 336             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 337             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 338             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 339
 340             }
 341
 342             /**************************
 343              * CALCULATE INTERACTIONS *
 344              **************************/
 345
 346             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 347             {
 348
 349             /* Compute parameters for interactions between i and j atoms */
 350             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 351
 352             /* REACTION-FIELD ELECTROSTATICS */
 353             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 354             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 355
 356             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 357
 358             /* Update potential sum for this i atom from the interaction with this j atom. */
 359             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 360             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 361
 362             fscal            = felec;
 363
 364             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 365
 366             /* Update vectorial force */
 367             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 368             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 369             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 370
 371             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 372             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 373             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 374
 375             }
 376
 377             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 378
 379             /* Inner loop uses 156 flops */
 380         }
 381
 382         if(jidx<j_index_end)
 383         {
 384
 385             jnrA             = jjnr[jidx];
 386             j_coord_offsetA  = DIM*jnrA;
 387
 388             /* load j atom coordinates */
 389             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 390                                               &jx0,&jy0,&jz0);
 391
 392             /* Calculate displacement vector */
 393             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 394             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 395             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 396             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 397             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 398             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 399             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 400             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 401             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 402
 403             /* Calculate squared distance and things based on it */
 404             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 405             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 406             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 407
 408             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 409             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 410             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 411
 412             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 413             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 414             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 415
 416             /* Load parameters for j particles */
 417             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 418             vdwjidx0A        = 2*vdwtype[jnrA+0];
 419
 420             fjx0             = _fjsp_setzero_v2r8();
 421             fjy0             = _fjsp_setzero_v2r8();
 422             fjz0             = _fjsp_setzero_v2r8();
 423
 424             /**************************
 425              * CALCULATE INTERACTIONS *
 426              **************************/
 427
 428             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 429             {
 430
 431             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 432
 433             /* Compute parameters for interactions between i and j atoms */
 434             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 435             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 436                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 437
 438             /* Calculate table index by multiplying r with table scale and truncate to integer */
 439             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 440             itab_tmp         = _fjsp_dtox_v2r8(rt);
 441             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 442             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 443             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 444
 445             vfconv.i[0]     *= 8;
 446             vfconv.i[1]     *= 8;
 447
 448             /* REACTION-FIELD ELECTROSTATICS */
 449             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 450             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 451
 452             /* CUBIC SPLINE TABLE DISPERSION */
 453             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 454             F                = _fjsp_setzero_v2r8();
 455             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 456             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 457             H                = _fjsp_setzero_v2r8();
 458             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 459             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 460             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 461             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 462             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 463             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 464
 465             /* CUBIC SPLINE TABLE REPULSION */
 466             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 467             F                = _fjsp_setzero_v2r8();
 468             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 469             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 470             H                = _fjsp_setzero_v2r8();
 471             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 472             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 473             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 474             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 475             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 476             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 477             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 478             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 479
 480             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 481
 482             /* Update potential sum for this i atom from the interaction with this j atom. */
 483             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 484             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 485             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 486             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 487             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 488             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 489
 490             fscal            = _fjsp_add_v2r8(felec,fvdw);
 491
 492             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 493
 494             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 495
 496             /* Update vectorial force */
 497             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 498             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 499             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 500
 501             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 502             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 503             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 504
 505             }
 506
 507             /**************************
 508              * CALCULATE INTERACTIONS *
 509              **************************/
 510
 511             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 512             {
 513
 514             /* Compute parameters for interactions between i and j atoms */
 515             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 516
 517             /* REACTION-FIELD ELECTROSTATICS */
 518             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 519             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 520
 521             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 522
 523             /* Update potential sum for this i atom from the interaction with this j atom. */
 524             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 525             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 526             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 531
 532             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 533
 534             /* Update vectorial force */
 535             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 536             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 537             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 538
 539             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 540             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 541             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 542
 543             }
 544
 545             /**************************
 546              * CALCULATE INTERACTIONS *
 547              **************************/
 548
 549             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 550             {
 551
 552             /* Compute parameters for interactions between i and j atoms */
 553             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 554
 555             /* REACTION-FIELD ELECTROSTATICS */
 556             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 557             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 558
 559             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 560
 561             /* Update potential sum for this i atom from the interaction with this j atom. */
 562             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 563             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 564             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 565
 566             fscal            = felec;
 567
 568             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 569
 570             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 571
 572             /* Update vectorial force */
 573             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 574             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 575             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 576
 577             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 578             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 579             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 580
 581             }
 582
 583             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 584
 585             /* Inner loop uses 156 flops */
 586         }
 587
 588         /* End of innermost loop */
 589
 590         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 591                                               f+i_coord_offset,fshift+i_shift_offset);
 592
 593         ggid                        = gid[iidx];
 594         /* Update potential energies */
 595         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 596         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 597
 598         /* Increment number of inner iterations */
 599         inneriter                  += j_index_end - j_index_start;
 600
 601         /* Outer loop uses 20 flops */
 602     }
 603
 604     /* Increment number of outer iterations */
 605     outeriter        += nri;
 606
 607     /* Update outer/inner flops */
 608
 609     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*156);
 610 }
 611 /*
 612  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 613  * Electrostatics interaction: ReactionField
 614  * VdW interaction:            CubicSplineTable
 615  * Geometry:                   Water3-Particle
 616  * Calculate force/pot:        Force
 617  */
 618 void
 619 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 620                     (t_nblist                    * gmx_restrict       nlist,
 621                      rvec                        * gmx_restrict          xx,
 622                      rvec                        * gmx_restrict          ff,
 623                      struct t_forcerec           * gmx_restrict          fr,
 624                      t_mdatoms                   * gmx_restrict     mdatoms,
 625                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 626                      t_nrnb                      * gmx_restrict        nrnb)
 627 {
 628     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 629      * just 0 for non-waters.
 630      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 631      * jnr indices corresponding to data put in the four positions in the SIMD register.
 632      */
 633     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 634     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 635     int              jnrA,jnrB;
 636     int              j_coord_offsetA,j_coord_offsetB;
 637     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 638     real             rcutoff_scalar;
 639     real             *shiftvec,*fshift,*x,*f;
 640     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 641     int              vdwioffset0;
 642     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 643     int              vdwioffset1;
 644     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 645     int              vdwioffset2;
 646     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 647     int              vdwjidx0A,vdwjidx0B;
 648     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 649     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 650     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 651     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 652     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 653     real             *charge;
 654     int              nvdwtype;
 655     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 656     int              *vdwtype;
 657     real             *vdwparam;
 658     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 659     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 660     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 661     real             *vftab;
 662     _fjsp_v2r8       itab_tmp;
 663     _fjsp_v2r8       dummy_mask,cutoff_mask;
 664     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 665     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 666     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 667
 668     x                = xx[0];
 669     f                = ff[0];
 670
 671     nri              = nlist->nri;
 672     iinr             = nlist->iinr;
 673     jindex           = nlist->jindex;
 674     jjnr             = nlist->jjnr;
 675     shiftidx         = nlist->shift;
 676     gid              = nlist->gid;
 677     shiftvec         = fr->shift_vec[0];
 678     fshift           = fr->fshift[0];
 679     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 680     charge           = mdatoms->chargeA;
 681     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 682     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 683     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 684     nvdwtype         = fr->ntype;
 685     vdwparam         = fr->nbfp;
 686     vdwtype          = mdatoms->typeA;
 687
 688     vftab            = kernel_data->table_vdw->data;
 689     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 690
 691     /* Setup water-specific parameters */
 692     inr              = nlist->iinr[0];
 693     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 694     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 695     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 696     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 697
 698     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 699     rcutoff_scalar   = fr->ic->rcoulomb;
 700     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 701     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 702
 703     /* Avoid stupid compiler warnings */
 704     jnrA = jnrB = 0;
 705     j_coord_offsetA = 0;
 706     j_coord_offsetB = 0;
 707
 708     outeriter        = 0;
 709     inneriter        = 0;
 710
 711     /* Start outer loop over neighborlists */
 712     for(iidx=0; iidx<nri; iidx++)
 713     {
 714         /* Load shift vector for this list */
 715         i_shift_offset   = DIM*shiftidx[iidx];
 716
 717         /* Load limits for loop over neighbors */
 718         j_index_start    = jindex[iidx];
 719         j_index_end      = jindex[iidx+1];
 720
 721         /* Get outer coordinate index */
 722         inr              = iinr[iidx];
 723         i_coord_offset   = DIM*inr;
 724
 725         /* Load i particle coords and add shift vector */
 726         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 727                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 728
 729         fix0             = _fjsp_setzero_v2r8();
 730         fiy0             = _fjsp_setzero_v2r8();
 731         fiz0             = _fjsp_setzero_v2r8();
 732         fix1             = _fjsp_setzero_v2r8();
 733         fiy1             = _fjsp_setzero_v2r8();
 734         fiz1             = _fjsp_setzero_v2r8();
 735         fix2             = _fjsp_setzero_v2r8();
 736         fiy2             = _fjsp_setzero_v2r8();
 737         fiz2             = _fjsp_setzero_v2r8();
 738
 739         /* Start inner kernel loop */
 740         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 741         {
 742
 743             /* Get j neighbor index, and coordinate index */
 744             jnrA             = jjnr[jidx];
 745             jnrB             = jjnr[jidx+1];
 746             j_coord_offsetA  = DIM*jnrA;
 747             j_coord_offsetB  = DIM*jnrB;
 748
 749             /* load j atom coordinates */
 750             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 751                                               &jx0,&jy0,&jz0);
 752
 753             /* Calculate displacement vector */
 754             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 755             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 756             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 757             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 758             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 759             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 760             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 761             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 762             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 763
 764             /* Calculate squared distance and things based on it */
 765             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 766             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 767             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 768
 769             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 770             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 771             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 772
 773             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 774             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 775             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 776
 777             /* Load parameters for j particles */
 778             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 779             vdwjidx0A        = 2*vdwtype[jnrA+0];
 780             vdwjidx0B        = 2*vdwtype[jnrB+0];
 781
 782             fjx0             = _fjsp_setzero_v2r8();
 783             fjy0             = _fjsp_setzero_v2r8();
 784             fjz0             = _fjsp_setzero_v2r8();
 785
 786             /**************************
 787              * CALCULATE INTERACTIONS *
 788              **************************/
 789
 790             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 791             {
 792
 793             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 794
 795             /* Compute parameters for interactions between i and j atoms */
 796             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 797             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 798                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 799
 800             /* Calculate table index by multiplying r with table scale and truncate to integer */
 801             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 802             itab_tmp         = _fjsp_dtox_v2r8(rt);
 803             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 804             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 805             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 806
 807             vfconv.i[0]     *= 8;
 808             vfconv.i[1]     *= 8;
 809
 810             /* REACTION-FIELD ELECTROSTATICS */
 811             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 812
 813             /* CUBIC SPLINE TABLE DISPERSION */
 814             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 815             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 816             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 817             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 818             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 819             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 820             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 821             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 822             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 823
 824             /* CUBIC SPLINE TABLE REPULSION */
 825             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 826             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 827             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 828             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 829             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 830             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 831             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 832             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 833             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 834             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 835
 836             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 837
 838             fscal            = _fjsp_add_v2r8(felec,fvdw);
 839
 840             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 841
 842             /* Update vectorial force */
 843             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 844             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 845             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 846
 847             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 848             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 849             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 850
 851             }
 852
 853             /**************************
 854              * CALCULATE INTERACTIONS *
 855              **************************/
 856
 857             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 858             {
 859
 860             /* Compute parameters for interactions between i and j atoms */
 861             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 862
 863             /* REACTION-FIELD ELECTROSTATICS */
 864             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 865
 866             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 867
 868             fscal            = felec;
 869
 870             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 871
 872             /* Update vectorial force */
 873             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 874             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 875             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 876
 877             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 878             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 879             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 880
 881             }
 882
 883             /**************************
 884              * CALCULATE INTERACTIONS *
 885              **************************/
 886
 887             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 888             {
 889
 890             /* Compute parameters for interactions between i and j atoms */
 891             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 892
 893             /* REACTION-FIELD ELECTROSTATICS */
 894             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 895
 896             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 897
 898             fscal            = felec;
 899
 900             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 901
 902             /* Update vectorial force */
 903             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 904             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 905             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 906
 907             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 908             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 909             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 910
 911             }
 912
 913             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 914
 915             /* Inner loop uses 129 flops */
 916         }
 917
 918         if(jidx<j_index_end)
 919         {
 920
 921             jnrA             = jjnr[jidx];
 922             j_coord_offsetA  = DIM*jnrA;
 923
 924             /* load j atom coordinates */
 925             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 926                                               &jx0,&jy0,&jz0);
 927
 928             /* Calculate displacement vector */
 929             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 930             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 931             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 932             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 933             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 934             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 935             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 936             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 937             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 938
 939             /* Calculate squared distance and things based on it */
 940             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 941             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 942             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 943
 944             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 945             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 946             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 947
 948             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 949             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 950             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 951
 952             /* Load parameters for j particles */
 953             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 954             vdwjidx0A        = 2*vdwtype[jnrA+0];
 955
 956             fjx0             = _fjsp_setzero_v2r8();
 957             fjy0             = _fjsp_setzero_v2r8();
 958             fjz0             = _fjsp_setzero_v2r8();
 959
 960             /**************************
 961              * CALCULATE INTERACTIONS *
 962              **************************/
 963
 964             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 965             {
 966
 967             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 968
 969             /* Compute parameters for interactions between i and j atoms */
 970             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 971             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 972                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 973
 974             /* Calculate table index by multiplying r with table scale and truncate to integer */
 975             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 976             itab_tmp         = _fjsp_dtox_v2r8(rt);
 977             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 978             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 979             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 980
 981             vfconv.i[0]     *= 8;
 982             vfconv.i[1]     *= 8;
 983
 984             /* REACTION-FIELD ELECTROSTATICS */
 985             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 986
 987             /* CUBIC SPLINE TABLE DISPERSION */
 988             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 989             F                = _fjsp_setzero_v2r8();
 990             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 991             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 992             H                = _fjsp_setzero_v2r8();
 993             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 994             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 995             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 996             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 997
 998             /* CUBIC SPLINE TABLE REPULSION */
 999             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1000             F                = _fjsp_setzero_v2r8();
1001             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1002             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1003             H                = _fjsp_setzero_v2r8();
1004             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1005             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1006             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1007             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1008             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1009
1010             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1011
1012             fscal            = _fjsp_add_v2r8(felec,fvdw);
1013
1014             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1015
1016             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1017
1018             /* Update vectorial force */
1019             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1020             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1021             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1022
1023             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1024             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1025             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1026
1027             }
1028
1029             /**************************
1030              * CALCULATE INTERACTIONS *
1031              **************************/
1032
1033             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1034             {
1035
1036             /* Compute parameters for interactions between i and j atoms */
1037             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1038
1039             /* REACTION-FIELD ELECTROSTATICS */
1040             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1041
1042             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1043
1044             fscal            = felec;
1045
1046             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1047
1048             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1049
1050             /* Update vectorial force */
1051             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1052             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1053             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1054
1055             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1056             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1057             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1058
1059             }
1060
1061             /**************************
1062              * CALCULATE INTERACTIONS *
1063              **************************/
1064
1065             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1066             {
1067
1068             /* Compute parameters for interactions between i and j atoms */
1069             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1070
1071             /* REACTION-FIELD ELECTROSTATICS */
1072             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1073
1074             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1075
1076             fscal            = felec;
1077
1078             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1079
1080             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1081
1082             /* Update vectorial force */
1083             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1084             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1085             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1086
1087             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1088             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1089             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1090
1091             }
1092
1093             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1094
1095             /* Inner loop uses 129 flops */
1096         }
1097
1098         /* End of innermost loop */
1099
1100         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1101                                               f+i_coord_offset,fshift+i_shift_offset);
1102
1103         /* Increment number of inner iterations */
1104         inneriter                  += j_index_end - j_index_start;
1105
1106         /* Outer loop uses 18 flops */
1107     }
1108
1109     /* Increment number of outer iterations */
1110     outeriter        += nri;
1111
1112     /* Update outer/inner flops */
1113
1114     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*129);
1115 }