src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRF_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, 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_ElecRF_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRF_VdwCSTab_GeomW4P1_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              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 121     charge           = mdatoms->chargeA;
 122     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 123     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 124     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 125     nvdwtype         = fr->ntype;
 126     vdwparam         = fr->nbfp;
 127     vdwtype          = mdatoms->typeA;
 128
 129     vftab            = kernel_data->table_vdw->data;
 130     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 131
 132     /* Setup water-specific parameters */
 133     inr              = nlist->iinr[0];
 134     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 135     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 136     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 137     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 138
 139     /* Avoid stupid compiler warnings */
 140     jnrA = jnrB = 0;
 141     j_coord_offsetA = 0;
 142     j_coord_offsetB = 0;
 143
 144     outeriter        = 0;
 145     inneriter        = 0;
 146
 147     /* Start outer loop over neighborlists */
 148     for(iidx=0; iidx<nri; iidx++)
 149     {
 150         /* Load shift vector for this list */
 151         i_shift_offset   = DIM*shiftidx[iidx];
 152
 153         /* Load limits for loop over neighbors */
 154         j_index_start    = jindex[iidx];
 155         j_index_end      = jindex[iidx+1];
 156
 157         /* Get outer coordinate index */
 158         inr              = iinr[iidx];
 159         i_coord_offset   = DIM*inr;
 160
 161         /* Load i particle coords and add shift vector */
 162         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 163                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 164
 165         fix0             = _fjsp_setzero_v2r8();
 166         fiy0             = _fjsp_setzero_v2r8();
 167         fiz0             = _fjsp_setzero_v2r8();
 168         fix1             = _fjsp_setzero_v2r8();
 169         fiy1             = _fjsp_setzero_v2r8();
 170         fiz1             = _fjsp_setzero_v2r8();
 171         fix2             = _fjsp_setzero_v2r8();
 172         fiy2             = _fjsp_setzero_v2r8();
 173         fiz2             = _fjsp_setzero_v2r8();
 174         fix3             = _fjsp_setzero_v2r8();
 175         fiy3             = _fjsp_setzero_v2r8();
 176         fiz3             = _fjsp_setzero_v2r8();
 177
 178         /* Reset potential sums */
 179         velecsum         = _fjsp_setzero_v2r8();
 180         vvdwsum          = _fjsp_setzero_v2r8();
 181
 182         /* Start inner kernel loop */
 183         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 184         {
 185
 186             /* Get j neighbor index, and coordinate index */
 187             jnrA             = jjnr[jidx];
 188             jnrB             = jjnr[jidx+1];
 189             j_coord_offsetA  = DIM*jnrA;
 190             j_coord_offsetB  = DIM*jnrB;
 191
 192             /* load j atom coordinates */
 193             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 194                                               &jx0,&jy0,&jz0);
 195
 196             /* Calculate displacement vector */
 197             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 198             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 199             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 200             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 201             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 202             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 203             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 204             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 205             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 206             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 207             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 208             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 209
 210             /* Calculate squared distance and things based on it */
 211             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 212             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 213             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 214             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 215
 216             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 217             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 218             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 219             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 220
 221             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 222             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 223             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 224
 225             /* Load parameters for j particles */
 226             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 227             vdwjidx0A        = 2*vdwtype[jnrA+0];
 228             vdwjidx0B        = 2*vdwtype[jnrB+0];
 229
 230             fjx0             = _fjsp_setzero_v2r8();
 231             fjy0             = _fjsp_setzero_v2r8();
 232             fjz0             = _fjsp_setzero_v2r8();
 233
 234             /**************************
 235              * CALCULATE INTERACTIONS *
 236              **************************/
 237
 238             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 239
 240             /* Compute parameters for interactions between i and j atoms */
 241             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 242                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 243
 244             /* Calculate table index by multiplying r with table scale and truncate to integer */
 245             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 246             itab_tmp         = _fjsp_dtox_v2r8(rt);
 247             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 248             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 249             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 250
 251             vfconv.i[0]     *= 8;
 252             vfconv.i[1]     *= 8;
 253
 254             /* CUBIC SPLINE TABLE DISPERSION */
 255             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 256             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 257             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 258             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 259             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 260             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 261             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 262             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 263             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 264             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 265             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 266
 267             /* CUBIC SPLINE TABLE REPULSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 273             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 274             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 275             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 276             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 279             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 280             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 281
 282             /* Update potential sum for this i atom from the interaction with this j atom. */
 283             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 284
 285             fscal            = fvdw;
 286
 287             /* Update vectorial force */
 288             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 289             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 290             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 291
 292             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 293             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 294             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 295
 296             /**************************
 297              * CALCULATE INTERACTIONS *
 298              **************************/
 299
 300             /* Compute parameters for interactions between i and j atoms */
 301             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 302
 303             /* REACTION-FIELD ELECTROSTATICS */
 304             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 305             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 306
 307             /* Update potential sum for this i atom from the interaction with this j atom. */
 308             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 309
 310             fscal            = felec;
 311
 312             /* Update vectorial force */
 313             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 314             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 315             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 316
 317             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 318             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 319             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 320
 321             /**************************
 322              * CALCULATE INTERACTIONS *
 323              **************************/
 324
 325             /* Compute parameters for interactions between i and j atoms */
 326             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 327
 328             /* REACTION-FIELD ELECTROSTATICS */
 329             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 330             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 331
 332             /* Update potential sum for this i atom from the interaction with this j atom. */
 333             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 334
 335             fscal            = felec;
 336
 337             /* Update vectorial force */
 338             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 339             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 340             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 341
 342             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 343             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 344             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 345
 346             /**************************
 347              * CALCULATE INTERACTIONS *
 348              **************************/
 349
 350             /* Compute parameters for interactions between i and j atoms */
 351             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 352
 353             /* REACTION-FIELD ELECTROSTATICS */
 354             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 355             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 356
 357             /* Update potential sum for this i atom from the interaction with this j atom. */
 358             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 359
 360             fscal            = felec;
 361
 362             /* Update vectorial force */
 363             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 364             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 365             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 366
 367             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 368             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 369             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 370
 371             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 372
 373             /* Inner loop uses 167 flops */
 374         }
 375
 376         if(jidx<j_index_end)
 377         {
 378
 379             jnrA             = jjnr[jidx];
 380             j_coord_offsetA  = DIM*jnrA;
 381
 382             /* load j atom coordinates */
 383             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 384                                               &jx0,&jy0,&jz0);
 385
 386             /* Calculate displacement vector */
 387             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 388             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 389             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 390             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 391             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 392             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 393             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 394             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 395             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 396             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 397             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 398             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 399
 400             /* Calculate squared distance and things based on it */
 401             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 402             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 403             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 404             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 405
 406             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 407             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 408             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 409             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 410
 411             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 412             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 413             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 414
 415             /* Load parameters for j particles */
 416             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 417             vdwjidx0A        = 2*vdwtype[jnrA+0];
 418
 419             fjx0             = _fjsp_setzero_v2r8();
 420             fjy0             = _fjsp_setzero_v2r8();
 421             fjz0             = _fjsp_setzero_v2r8();
 422
 423             /**************************
 424              * CALCULATE INTERACTIONS *
 425              **************************/
 426
 427             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 428
 429             /* Compute parameters for interactions between i and j atoms */
 430             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 431                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 432
 433             /* Calculate table index by multiplying r with table scale and truncate to integer */
 434             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 435             itab_tmp         = _fjsp_dtox_v2r8(rt);
 436             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 437             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 438             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 439
 440             vfconv.i[0]     *= 8;
 441             vfconv.i[1]     *= 8;
 442
 443             /* CUBIC SPLINE TABLE DISPERSION */
 444             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 445             F                = _fjsp_setzero_v2r8();
 446             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 447             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 448             H                = _fjsp_setzero_v2r8();
 449             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 450             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 451             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 452             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 453             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 454             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 455
 456             /* CUBIC SPLINE TABLE REPULSION */
 457             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 458             F                = _fjsp_setzero_v2r8();
 459             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 460             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 461             H                = _fjsp_setzero_v2r8();
 462             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 463             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 464             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 465             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 466             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 467             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 468             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 469             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 470
 471             /* Update potential sum for this i atom from the interaction with this j atom. */
 472             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 473             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 474
 475             fscal            = fvdw;
 476
 477             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 478
 479             /* Update vectorial force */
 480             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 481             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 482             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 483
 484             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 485             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 486             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 487
 488             /**************************
 489              * CALCULATE INTERACTIONS *
 490              **************************/
 491
 492             /* Compute parameters for interactions between i and j atoms */
 493             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 494
 495             /* REACTION-FIELD ELECTROSTATICS */
 496             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 497             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 498
 499             /* Update potential sum for this i atom from the interaction with this j atom. */
 500             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 501             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 502
 503             fscal            = felec;
 504
 505             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 506
 507             /* Update vectorial force */
 508             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 509             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 510             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 511
 512             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 513             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 514             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             /* Compute parameters for interactions between i and j atoms */
 521             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 522
 523             /* REACTION-FIELD ELECTROSTATICS */
 524             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 525             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 529             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 530
 531             fscal            = felec;
 532
 533             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 534
 535             /* Update vectorial force */
 536             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 537             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 538             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 539
 540             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 541             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 542             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             /* Compute parameters for interactions between i and j atoms */
 549             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 550
 551             /* REACTION-FIELD ELECTROSTATICS */
 552             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 553             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 557             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 562
 563             /* Update vectorial force */
 564             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 565             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 566             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 567
 568             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 569             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 570             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 571
 572             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 573
 574             /* Inner loop uses 167 flops */
 575         }
 576
 577         /* End of innermost loop */
 578
 579         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 580                                               f+i_coord_offset,fshift+i_shift_offset);
 581
 582         ggid                        = gid[iidx];
 583         /* Update potential energies */
 584         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 585         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 586
 587         /* Increment number of inner iterations */
 588         inneriter                  += j_index_end - j_index_start;
 589
 590         /* Outer loop uses 26 flops */
 591     }
 592
 593     /* Increment number of outer iterations */
 594     outeriter        += nri;
 595
 596     /* Update outer/inner flops */
 597
 598     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*167);
 599 }
 600 /*
 601  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 602  * Electrostatics interaction: ReactionField
 603  * VdW interaction:            CubicSplineTable
 604  * Geometry:                   Water4-Particle
 605  * Calculate force/pot:        Force
 606  */
 607 void
 608 nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 609                     (t_nblist                    * gmx_restrict       nlist,
 610                      rvec                        * gmx_restrict          xx,
 611                      rvec                        * gmx_restrict          ff,
 612                      struct t_forcerec           * gmx_restrict          fr,
 613                      t_mdatoms                   * gmx_restrict     mdatoms,
 614                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 615                      t_nrnb                      * gmx_restrict        nrnb)
 616 {
 617     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 618      * just 0 for non-waters.
 619      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 620      * jnr indices corresponding to data put in the four positions in the SIMD register.
 621      */
 622     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 623     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 624     int              jnrA,jnrB;
 625     int              j_coord_offsetA,j_coord_offsetB;
 626     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 627     real             rcutoff_scalar;
 628     real             *shiftvec,*fshift,*x,*f;
 629     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 630     int              vdwioffset0;
 631     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 632     int              vdwioffset1;
 633     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 634     int              vdwioffset2;
 635     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 636     int              vdwioffset3;
 637     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 638     int              vdwjidx0A,vdwjidx0B;
 639     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 640     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 641     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 642     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 643     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 644     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 645     real             *charge;
 646     int              nvdwtype;
 647     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 648     int              *vdwtype;
 649     real             *vdwparam;
 650     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 651     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 652     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 653     real             *vftab;
 654     _fjsp_v2r8       itab_tmp;
 655     _fjsp_v2r8       dummy_mask,cutoff_mask;
 656     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 657     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 658     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 659
 660     x                = xx[0];
 661     f                = ff[0];
 662
 663     nri              = nlist->nri;
 664     iinr             = nlist->iinr;
 665     jindex           = nlist->jindex;
 666     jjnr             = nlist->jjnr;
 667     shiftidx         = nlist->shift;
 668     gid              = nlist->gid;
 669     shiftvec         = fr->shift_vec[0];
 670     fshift           = fr->fshift[0];
 671     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 672     charge           = mdatoms->chargeA;
 673     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 674     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 675     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 676     nvdwtype         = fr->ntype;
 677     vdwparam         = fr->nbfp;
 678     vdwtype          = mdatoms->typeA;
 679
 680     vftab            = kernel_data->table_vdw->data;
 681     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 682
 683     /* Setup water-specific parameters */
 684     inr              = nlist->iinr[0];
 685     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 686     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 687     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 688     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 689
 690     /* Avoid stupid compiler warnings */
 691     jnrA = jnrB = 0;
 692     j_coord_offsetA = 0;
 693     j_coord_offsetB = 0;
 694
 695     outeriter        = 0;
 696     inneriter        = 0;
 697
 698     /* Start outer loop over neighborlists */
 699     for(iidx=0; iidx<nri; iidx++)
 700     {
 701         /* Load shift vector for this list */
 702         i_shift_offset   = DIM*shiftidx[iidx];
 703
 704         /* Load limits for loop over neighbors */
 705         j_index_start    = jindex[iidx];
 706         j_index_end      = jindex[iidx+1];
 707
 708         /* Get outer coordinate index */
 709         inr              = iinr[iidx];
 710         i_coord_offset   = DIM*inr;
 711
 712         /* Load i particle coords and add shift vector */
 713         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 714                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 715
 716         fix0             = _fjsp_setzero_v2r8();
 717         fiy0             = _fjsp_setzero_v2r8();
 718         fiz0             = _fjsp_setzero_v2r8();
 719         fix1             = _fjsp_setzero_v2r8();
 720         fiy1             = _fjsp_setzero_v2r8();
 721         fiz1             = _fjsp_setzero_v2r8();
 722         fix2             = _fjsp_setzero_v2r8();
 723         fiy2             = _fjsp_setzero_v2r8();
 724         fiz2             = _fjsp_setzero_v2r8();
 725         fix3             = _fjsp_setzero_v2r8();
 726         fiy3             = _fjsp_setzero_v2r8();
 727         fiz3             = _fjsp_setzero_v2r8();
 728
 729         /* Start inner kernel loop */
 730         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 731         {
 732
 733             /* Get j neighbor index, and coordinate index */
 734             jnrA             = jjnr[jidx];
 735             jnrB             = jjnr[jidx+1];
 736             j_coord_offsetA  = DIM*jnrA;
 737             j_coord_offsetB  = DIM*jnrB;
 738
 739             /* load j atom coordinates */
 740             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 741                                               &jx0,&jy0,&jz0);
 742
 743             /* Calculate displacement vector */
 744             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 745             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 746             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 747             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 748             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 749             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 750             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 751             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 752             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 753             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 754             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 755             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 756
 757             /* Calculate squared distance and things based on it */
 758             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 759             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 760             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 761             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 762
 763             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 764             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 765             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 766             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 767
 768             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 769             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 770             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 771
 772             /* Load parameters for j particles */
 773             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 774             vdwjidx0A        = 2*vdwtype[jnrA+0];
 775             vdwjidx0B        = 2*vdwtype[jnrB+0];
 776
 777             fjx0             = _fjsp_setzero_v2r8();
 778             fjy0             = _fjsp_setzero_v2r8();
 779             fjz0             = _fjsp_setzero_v2r8();
 780
 781             /**************************
 782              * CALCULATE INTERACTIONS *
 783              **************************/
 784
 785             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 786
 787             /* Compute parameters for interactions between i and j atoms */
 788             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 789                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 790
 791             /* Calculate table index by multiplying r with table scale and truncate to integer */
 792             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 793             itab_tmp         = _fjsp_dtox_v2r8(rt);
 794             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 795             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 796             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 797
 798             vfconv.i[0]     *= 8;
 799             vfconv.i[1]     *= 8;
 800
 801             /* CUBIC SPLINE TABLE DISPERSION */
 802             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 803             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 804             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 805             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 806             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 807             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 808             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 809             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 810             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 811
 812             /* CUBIC SPLINE TABLE REPULSION */
 813             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 814             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 815             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 816             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 817             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 818             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 819             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 820             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 821             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 822             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 823
 824             fscal            = fvdw;
 825
 826             /* Update vectorial force */
 827             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 828             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 829             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 830
 831             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 832             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 833             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 834
 835             /**************************
 836              * CALCULATE INTERACTIONS *
 837              **************************/
 838
 839             /* Compute parameters for interactions between i and j atoms */
 840             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 841
 842             /* REACTION-FIELD ELECTROSTATICS */
 843             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 844
 845             fscal            = felec;
 846
 847             /* Update vectorial force */
 848             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 849             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 850             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 851
 852             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 853             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 854             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 855
 856             /**************************
 857              * CALCULATE INTERACTIONS *
 858              **************************/
 859
 860             /* Compute parameters for interactions between i and j atoms */
 861             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 862
 863             /* REACTION-FIELD ELECTROSTATICS */
 864             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 865
 866             fscal            = felec;
 867
 868             /* Update vectorial force */
 869             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 870             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 871             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 872
 873             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 874             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 875             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 876
 877             /**************************
 878              * CALCULATE INTERACTIONS *
 879              **************************/
 880
 881             /* Compute parameters for interactions between i and j atoms */
 882             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 883
 884             /* REACTION-FIELD ELECTROSTATICS */
 885             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 886
 887             fscal            = felec;
 888
 889             /* Update vectorial force */
 890             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 891             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 892             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 893
 894             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 895             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 896             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 897
 898             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 899
 900             /* Inner loop uses 144 flops */
 901         }
 902
 903         if(jidx<j_index_end)
 904         {
 905
 906             jnrA             = jjnr[jidx];
 907             j_coord_offsetA  = DIM*jnrA;
 908
 909             /* load j atom coordinates */
 910             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 911                                               &jx0,&jy0,&jz0);
 912
 913             /* Calculate displacement vector */
 914             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 915             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 916             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 917             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 918             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 919             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 920             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 921             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 922             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 923             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 924             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 925             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 926
 927             /* Calculate squared distance and things based on it */
 928             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 929             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 930             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 931             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 932
 933             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 934             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 935             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 936             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 937
 938             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 939             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 940             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 941
 942             /* Load parameters for j particles */
 943             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 944             vdwjidx0A        = 2*vdwtype[jnrA+0];
 945
 946             fjx0             = _fjsp_setzero_v2r8();
 947             fjy0             = _fjsp_setzero_v2r8();
 948             fjz0             = _fjsp_setzero_v2r8();
 949
 950             /**************************
 951              * CALCULATE INTERACTIONS *
 952              **************************/
 953
 954             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 955
 956             /* Compute parameters for interactions between i and j atoms */
 957             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 958                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 959
 960             /* Calculate table index by multiplying r with table scale and truncate to integer */
 961             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 962             itab_tmp         = _fjsp_dtox_v2r8(rt);
 963             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 964             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 965             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 966
 967             vfconv.i[0]     *= 8;
 968             vfconv.i[1]     *= 8;
 969
 970             /* CUBIC SPLINE TABLE DISPERSION */
 971             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 972             F                = _fjsp_setzero_v2r8();
 973             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 974             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 975             H                = _fjsp_setzero_v2r8();
 976             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 977             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 978             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 979             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 980
 981             /* CUBIC SPLINE TABLE REPULSION */
 982             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 983             F                = _fjsp_setzero_v2r8();
 984             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 985             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 986             H                = _fjsp_setzero_v2r8();
 987             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 988             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 989             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 990             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 991             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 992
 993             fscal            = fvdw;
 994
 995             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 996
 997             /* Update vectorial force */
 998             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 999             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1000             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1001
1002             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1003             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1004             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1005
1006             /**************************
1007              * CALCULATE INTERACTIONS *
1008              **************************/
1009
1010             /* Compute parameters for interactions between i and j atoms */
1011             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1012
1013             /* REACTION-FIELD ELECTROSTATICS */
1014             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1015
1016             fscal            = felec;
1017
1018             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1019
1020             /* Update vectorial force */
1021             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1022             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1023             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1024
1025             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1026             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1027             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1028
1029             /**************************
1030              * CALCULATE INTERACTIONS *
1031              **************************/
1032
1033             /* Compute parameters for interactions between i and j atoms */
1034             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1035
1036             /* REACTION-FIELD ELECTROSTATICS */
1037             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1038
1039             fscal            = felec;
1040
1041             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1042
1043             /* Update vectorial force */
1044             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1045             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1046             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1047
1048             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1049             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1050             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1051
1052             /**************************
1053              * CALCULATE INTERACTIONS *
1054              **************************/
1055
1056             /* Compute parameters for interactions between i and j atoms */
1057             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1058
1059             /* REACTION-FIELD ELECTROSTATICS */
1060             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1061
1062             fscal            = felec;
1063
1064             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1065
1066             /* Update vectorial force */
1067             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1068             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1069             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1070
1071             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1072             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1073             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1074
1075             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1076
1077             /* Inner loop uses 144 flops */
1078         }
1079
1080         /* End of innermost loop */
1081
1082         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1083                                               f+i_coord_offset,fshift+i_shift_offset);
1084
1085         /* Increment number of inner iterations */
1086         inneriter                  += j_index_end - j_index_start;
1087
1088         /* Outer loop uses 24 flops */
1089     }
1090
1091     /* Increment number of outer iterations */
1092     outeriter        += nri;
1093
1094     /* Update outer/inner flops */
1095
1096     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1097 }