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, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "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                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     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_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 431
 432             /* Calculate table index by multiplying r with table scale and truncate to integer */
 433             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 434             itab_tmp         = _fjsp_dtox_v2r8(rt);
 435             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 436             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 437             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 438
 439             vfconv.i[0]     *= 8;
 440             vfconv.i[1]     *= 8;
 441
 442             /* CUBIC SPLINE TABLE DISPERSION */
 443             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 444             F                = _fjsp_setzero_v2r8();
 445             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 446             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 447             H                = _fjsp_setzero_v2r8();
 448             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 449             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 450             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 451             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 452             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 453             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 454
 455             /* CUBIC SPLINE TABLE REPULSION */
 456             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 457             F                = _fjsp_setzero_v2r8();
 458             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 459             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 460             H                = _fjsp_setzero_v2r8();
 461             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 462             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 463             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 464             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 465             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 466             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 467             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 468             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 472             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 473
 474             fscal            = fvdw;
 475
 476             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 477
 478             /* Update vectorial force */
 479             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 480             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 481             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 482
 483             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 484             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 485             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             /* Compute parameters for interactions between i and j atoms */
 492             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 493
 494             /* REACTION-FIELD ELECTROSTATICS */
 495             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 496             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 500             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 501
 502             fscal            = felec;
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 508             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 509             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             /* Compute parameters for interactions between i and j atoms */
 520             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 521
 522             /* REACTION-FIELD ELECTROSTATICS */
 523             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 524             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 525
 526             /* Update potential sum for this i atom from the interaction with this j atom. */
 527             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 528             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 533
 534             /* Update vectorial force */
 535             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 536             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 537             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 538
 539             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 540             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 541             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 542
 543             /**************************
 544              * CALCULATE INTERACTIONS *
 545              **************************/
 546
 547             /* Compute parameters for interactions between i and j atoms */
 548             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 549
 550             /* REACTION-FIELD ELECTROSTATICS */
 551             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 552             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 556             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 561
 562             /* Update vectorial force */
 563             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 564             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 565             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 566
 567             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 568             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 569             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 570
 571             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 572
 573             /* Inner loop uses 167 flops */
 574         }
 575
 576         /* End of innermost loop */
 577
 578         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 579                                               f+i_coord_offset,fshift+i_shift_offset);
 580
 581         ggid                        = gid[iidx];
 582         /* Update potential energies */
 583         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 584         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 585
 586         /* Increment number of inner iterations */
 587         inneriter                  += j_index_end - j_index_start;
 588
 589         /* Outer loop uses 26 flops */
 590     }
 591
 592     /* Increment number of outer iterations */
 593     outeriter        += nri;
 594
 595     /* Update outer/inner flops */
 596
 597     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*167);
 598 }
 599 /*
 600  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 601  * Electrostatics interaction: ReactionField
 602  * VdW interaction:            CubicSplineTable
 603  * Geometry:                   Water4-Particle
 604  * Calculate force/pot:        Force
 605  */
 606 void
 607 nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 608                     (t_nblist                    * gmx_restrict       nlist,
 609                      rvec                        * gmx_restrict          xx,
 610                      rvec                        * gmx_restrict          ff,
 611                      t_forcerec                  * gmx_restrict          fr,
 612                      t_mdatoms                   * gmx_restrict     mdatoms,
 613                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 614                      t_nrnb                      * gmx_restrict        nrnb)
 615 {
 616     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 617      * just 0 for non-waters.
 618      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 619      * jnr indices corresponding to data put in the four positions in the SIMD register.
 620      */
 621     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 622     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 623     int              jnrA,jnrB;
 624     int              j_coord_offsetA,j_coord_offsetB;
 625     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 626     real             rcutoff_scalar;
 627     real             *shiftvec,*fshift,*x,*f;
 628     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 629     int              vdwioffset0;
 630     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 631     int              vdwioffset1;
 632     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 633     int              vdwioffset2;
 634     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 635     int              vdwioffset3;
 636     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 637     int              vdwjidx0A,vdwjidx0B;
 638     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 639     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 640     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 641     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 642     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 643     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 644     real             *charge;
 645     int              nvdwtype;
 646     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 647     int              *vdwtype;
 648     real             *vdwparam;
 649     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 650     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 651     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 652     real             *vftab;
 653     _fjsp_v2r8       itab_tmp;
 654     _fjsp_v2r8       dummy_mask,cutoff_mask;
 655     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 656     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 657     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 658
 659     x                = xx[0];
 660     f                = ff[0];
 661
 662     nri              = nlist->nri;
 663     iinr             = nlist->iinr;
 664     jindex           = nlist->jindex;
 665     jjnr             = nlist->jjnr;
 666     shiftidx         = nlist->shift;
 667     gid              = nlist->gid;
 668     shiftvec         = fr->shift_vec[0];
 669     fshift           = fr->fshift[0];
 670     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 671     charge           = mdatoms->chargeA;
 672     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 673     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 674     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 675     nvdwtype         = fr->ntype;
 676     vdwparam         = fr->nbfp;
 677     vdwtype          = mdatoms->typeA;
 678
 679     vftab            = kernel_data->table_vdw->data;
 680     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 681
 682     /* Setup water-specific parameters */
 683     inr              = nlist->iinr[0];
 684     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 685     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 686     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 687     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 688
 689     /* Avoid stupid compiler warnings */
 690     jnrA = jnrB = 0;
 691     j_coord_offsetA = 0;
 692     j_coord_offsetB = 0;
 693
 694     outeriter        = 0;
 695     inneriter        = 0;
 696
 697     /* Start outer loop over neighborlists */
 698     for(iidx=0; iidx<nri; iidx++)
 699     {
 700         /* Load shift vector for this list */
 701         i_shift_offset   = DIM*shiftidx[iidx];
 702
 703         /* Load limits for loop over neighbors */
 704         j_index_start    = jindex[iidx];
 705         j_index_end      = jindex[iidx+1];
 706
 707         /* Get outer coordinate index */
 708         inr              = iinr[iidx];
 709         i_coord_offset   = DIM*inr;
 710
 711         /* Load i particle coords and add shift vector */
 712         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 713                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 714
 715         fix0             = _fjsp_setzero_v2r8();
 716         fiy0             = _fjsp_setzero_v2r8();
 717         fiz0             = _fjsp_setzero_v2r8();
 718         fix1             = _fjsp_setzero_v2r8();
 719         fiy1             = _fjsp_setzero_v2r8();
 720         fiz1             = _fjsp_setzero_v2r8();
 721         fix2             = _fjsp_setzero_v2r8();
 722         fiy2             = _fjsp_setzero_v2r8();
 723         fiz2             = _fjsp_setzero_v2r8();
 724         fix3             = _fjsp_setzero_v2r8();
 725         fiy3             = _fjsp_setzero_v2r8();
 726         fiz3             = _fjsp_setzero_v2r8();
 727
 728         /* Start inner kernel loop */
 729         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 730         {
 731
 732             /* Get j neighbor index, and coordinate index */
 733             jnrA             = jjnr[jidx];
 734             jnrB             = jjnr[jidx+1];
 735             j_coord_offsetA  = DIM*jnrA;
 736             j_coord_offsetB  = DIM*jnrB;
 737
 738             /* load j atom coordinates */
 739             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 740                                               &jx0,&jy0,&jz0);
 741
 742             /* Calculate displacement vector */
 743             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 744             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 745             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 746             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 747             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 748             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 749             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 750             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 751             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 752             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 753             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 754             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 755
 756             /* Calculate squared distance and things based on it */
 757             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 758             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 759             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 760             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 761
 762             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 763             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 764             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 765             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 766
 767             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 768             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 769             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 770
 771             /* Load parameters for j particles */
 772             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 773             vdwjidx0A        = 2*vdwtype[jnrA+0];
 774             vdwjidx0B        = 2*vdwtype[jnrB+0];
 775
 776             fjx0             = _fjsp_setzero_v2r8();
 777             fjy0             = _fjsp_setzero_v2r8();
 778             fjz0             = _fjsp_setzero_v2r8();
 779
 780             /**************************
 781              * CALCULATE INTERACTIONS *
 782              **************************/
 783
 784             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 785
 786             /* Compute parameters for interactions between i and j atoms */
 787             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 788                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 789
 790             /* Calculate table index by multiplying r with table scale and truncate to integer */
 791             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 792             itab_tmp         = _fjsp_dtox_v2r8(rt);
 793             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 794             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 795             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 796
 797             vfconv.i[0]     *= 8;
 798             vfconv.i[1]     *= 8;
 799
 800             /* CUBIC SPLINE TABLE DISPERSION */
 801             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 802             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 803             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 804             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 805             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 806             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 807             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 808             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 809             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 810
 811             /* CUBIC SPLINE TABLE REPULSION */
 812             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 813             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 814             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 815             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 816             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 817             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 818             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 819             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 820             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 821             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 822
 823             fscal            = fvdw;
 824
 825             /* Update vectorial force */
 826             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 827             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 828             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 829
 830             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 831             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 832             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 833
 834             /**************************
 835              * CALCULATE INTERACTIONS *
 836              **************************/
 837
 838             /* Compute parameters for interactions between i and j atoms */
 839             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 840
 841             /* REACTION-FIELD ELECTROSTATICS */
 842             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 843
 844             fscal            = felec;
 845
 846             /* Update vectorial force */
 847             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 848             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 849             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 850
 851             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 852             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 853             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 854
 855             /**************************
 856              * CALCULATE INTERACTIONS *
 857              **************************/
 858
 859             /* Compute parameters for interactions between i and j atoms */
 860             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 861
 862             /* REACTION-FIELD ELECTROSTATICS */
 863             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 864
 865             fscal            = felec;
 866
 867             /* Update vectorial force */
 868             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 869             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 870             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 871
 872             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 873             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 874             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 875
 876             /**************************
 877              * CALCULATE INTERACTIONS *
 878              **************************/
 879
 880             /* Compute parameters for interactions between i and j atoms */
 881             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 882
 883             /* REACTION-FIELD ELECTROSTATICS */
 884             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 885
 886             fscal            = felec;
 887
 888             /* Update vectorial force */
 889             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 890             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 891             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 892
 893             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 894             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 895             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 896
 897             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 898
 899             /* Inner loop uses 144 flops */
 900         }
 901
 902         if(jidx<j_index_end)
 903         {
 904
 905             jnrA             = jjnr[jidx];
 906             j_coord_offsetA  = DIM*jnrA;
 907
 908             /* load j atom coordinates */
 909             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 910                                               &jx0,&jy0,&jz0);
 911
 912             /* Calculate displacement vector */
 913             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 914             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 915             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 916             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 917             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 918             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 919             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 920             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 921             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 922             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 923             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 924             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 925
 926             /* Calculate squared distance and things based on it */
 927             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 928             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 929             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 930             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 931
 932             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 933             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 934             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 935             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 936
 937             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 938             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 939             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 940
 941             /* Load parameters for j particles */
 942             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 943             vdwjidx0A        = 2*vdwtype[jnrA+0];
 944
 945             fjx0             = _fjsp_setzero_v2r8();
 946             fjy0             = _fjsp_setzero_v2r8();
 947             fjz0             = _fjsp_setzero_v2r8();
 948
 949             /**************************
 950              * CALCULATE INTERACTIONS *
 951              **************************/
 952
 953             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 954
 955             /* Compute parameters for interactions between i and j atoms */
 956             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 957
 958             /* Calculate table index by multiplying r with table scale and truncate to integer */
 959             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 960             itab_tmp         = _fjsp_dtox_v2r8(rt);
 961             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 962             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 963             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 964
 965             vfconv.i[0]     *= 8;
 966             vfconv.i[1]     *= 8;
 967
 968             /* CUBIC SPLINE TABLE DISPERSION */
 969             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 970             F                = _fjsp_setzero_v2r8();
 971             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 972             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 973             H                = _fjsp_setzero_v2r8();
 974             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 975             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 976             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 977             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 978
 979             /* CUBIC SPLINE TABLE REPULSION */
 980             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 981             F                = _fjsp_setzero_v2r8();
 982             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 983             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 984             H                = _fjsp_setzero_v2r8();
 985             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 986             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 987             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 988             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 989             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 990
 991             fscal            = fvdw;
 992
 993             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 994
 995             /* Update vectorial force */
 996             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 997             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 998             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 999
1000             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1001             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1002             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1003
1004             /**************************
1005              * CALCULATE INTERACTIONS *
1006              **************************/
1007
1008             /* Compute parameters for interactions between i and j atoms */
1009             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1010
1011             /* REACTION-FIELD ELECTROSTATICS */
1012             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1013
1014             fscal            = felec;
1015
1016             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1017
1018             /* Update vectorial force */
1019             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1020             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1021             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1022
1023             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1024             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1025             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1026
1027             /**************************
1028              * CALCULATE INTERACTIONS *
1029              **************************/
1030
1031             /* Compute parameters for interactions between i and j atoms */
1032             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1033
1034             /* REACTION-FIELD ELECTROSTATICS */
1035             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1036
1037             fscal            = felec;
1038
1039             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1040
1041             /* Update vectorial force */
1042             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1043             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1044             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1045
1046             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1047             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1048             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1049
1050             /**************************
1051              * CALCULATE INTERACTIONS *
1052              **************************/
1053
1054             /* Compute parameters for interactions between i and j atoms */
1055             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1056
1057             /* REACTION-FIELD ELECTROSTATICS */
1058             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1059
1060             fscal            = felec;
1061
1062             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1063
1064             /* Update vectorial force */
1065             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1066             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1067             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1068
1069             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1070             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1071             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1072
1073             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1074
1075             /* Inner loop uses 144 flops */
1076         }
1077
1078         /* End of innermost loop */
1079
1080         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1081                                               f+i_coord_offset,fshift+i_shift_offset);
1082
1083         /* Increment number of inner iterations */
1084         inneriter                  += j_index_end - j_index_start;
1085
1086         /* Outer loop uses 24 flops */
1087     }
1088
1089     /* Increment number of outer iterations */
1090     outeriter        += nri;
1091
1092     /* Update outer/inner flops */
1093
1094     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1095 }