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 "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRF_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 104     real             *vftab;
 105     _fjsp_v2r8       itab_tmp;
 106     _fjsp_v2r8       dummy_mask,cutoff_mask;
 107     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 108     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 109     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 110
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 125     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 126     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 127     nvdwtype         = fr->ntype;
 128     vdwparam         = fr->nbfp;
 129     vdwtype          = mdatoms->typeA;
 130
 131     vftab            = kernel_data->table_vdw->data;
 132     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 133
 134     /* Setup water-specific parameters */
 135     inr              = nlist->iinr[0];
 136     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 137     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 138     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 139     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 140
 141     /* Avoid stupid compiler warnings */
 142     jnrA = jnrB = 0;
 143     j_coord_offsetA = 0;
 144     j_coord_offsetB = 0;
 145
 146     outeriter        = 0;
 147     inneriter        = 0;
 148
 149     /* Start outer loop over neighborlists */
 150     for(iidx=0; iidx<nri; iidx++)
 151     {
 152         /* Load shift vector for this list */
 153         i_shift_offset   = DIM*shiftidx[iidx];
 154
 155         /* Load limits for loop over neighbors */
 156         j_index_start    = jindex[iidx];
 157         j_index_end      = jindex[iidx+1];
 158
 159         /* Get outer coordinate index */
 160         inr              = iinr[iidx];
 161         i_coord_offset   = DIM*inr;
 162
 163         /* Load i particle coords and add shift vector */
 164         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 165                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 166
 167         fix0             = _fjsp_setzero_v2r8();
 168         fiy0             = _fjsp_setzero_v2r8();
 169         fiz0             = _fjsp_setzero_v2r8();
 170         fix1             = _fjsp_setzero_v2r8();
 171         fiy1             = _fjsp_setzero_v2r8();
 172         fiz1             = _fjsp_setzero_v2r8();
 173         fix2             = _fjsp_setzero_v2r8();
 174         fiy2             = _fjsp_setzero_v2r8();
 175         fiz2             = _fjsp_setzero_v2r8();
 176         fix3             = _fjsp_setzero_v2r8();
 177         fiy3             = _fjsp_setzero_v2r8();
 178         fiz3             = _fjsp_setzero_v2r8();
 179
 180         /* Reset potential sums */
 181         velecsum         = _fjsp_setzero_v2r8();
 182         vvdwsum          = _fjsp_setzero_v2r8();
 183
 184         /* Start inner kernel loop */
 185         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 186         {
 187
 188             /* Get j neighbor index, and coordinate index */
 189             jnrA             = jjnr[jidx];
 190             jnrB             = jjnr[jidx+1];
 191             j_coord_offsetA  = DIM*jnrA;
 192             j_coord_offsetB  = DIM*jnrB;
 193
 194             /* load j atom coordinates */
 195             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 196                                               &jx0,&jy0,&jz0);
 197
 198             /* Calculate displacement vector */
 199             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 200             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 201             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 202             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 203             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 204             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 205             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 206             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 207             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 208             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 209             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 210             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 211
 212             /* Calculate squared distance and things based on it */
 213             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 214             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 215             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 216             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 217
 218             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 219             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 220             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 221             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 222
 223             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 224             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 225             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 226
 227             /* Load parameters for j particles */
 228             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 229             vdwjidx0A        = 2*vdwtype[jnrA+0];
 230             vdwjidx0B        = 2*vdwtype[jnrB+0];
 231
 232             fjx0             = _fjsp_setzero_v2r8();
 233             fjy0             = _fjsp_setzero_v2r8();
 234             fjz0             = _fjsp_setzero_v2r8();
 235
 236             /**************************
 237              * CALCULATE INTERACTIONS *
 238              **************************/
 239
 240             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 241
 242             /* Compute parameters for interactions between i and j atoms */
 243             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 244                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 245
 246             /* Calculate table index by multiplying r with table scale and truncate to integer */
 247             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 248             itab_tmp         = _fjsp_dtox_v2r8(rt);
 249             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 250             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 251             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 252
 253             vfconv.i[0]     *= 8;
 254             vfconv.i[1]     *= 8;
 255
 256             /* CUBIC SPLINE TABLE DISPERSION */
 257             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 258             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 259             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 260             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 261             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 262             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 263             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 264             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 265             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 266             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 267             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 268
 269             /* CUBIC SPLINE TABLE REPULSION */
 270             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 271             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 272             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 273             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 274             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 276             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 277             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 278             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 279             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 280             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 281             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 282             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 283
 284             /* Update potential sum for this i atom from the interaction with this j atom. */
 285             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 286
 287             fscal            = fvdw;
 288
 289             /* Update vectorial force */
 290             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 291             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 292             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 293
 294             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 295             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 296             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 297
 298             /**************************
 299              * CALCULATE INTERACTIONS *
 300              **************************/
 301
 302             /* Compute parameters for interactions between i and j atoms */
 303             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 304
 305             /* REACTION-FIELD ELECTROSTATICS */
 306             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 307             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 308
 309             /* Update potential sum for this i atom from the interaction with this j atom. */
 310             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 311
 312             fscal            = felec;
 313
 314             /* Update vectorial force */
 315             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 316             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 317             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 318
 319             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 320             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 321             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 329
 330             /* REACTION-FIELD ELECTROSTATICS */
 331             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 332             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 333
 334             /* Update potential sum for this i atom from the interaction with this j atom. */
 335             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 336
 337             fscal            = felec;
 338
 339             /* Update vectorial force */
 340             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 341             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 342             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 343
 344             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 345             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 346             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 347
 348             /**************************
 349              * CALCULATE INTERACTIONS *
 350              **************************/
 351
 352             /* Compute parameters for interactions between i and j atoms */
 353             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 354
 355             /* REACTION-FIELD ELECTROSTATICS */
 356             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 357             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 358
 359             /* Update potential sum for this i atom from the interaction with this j atom. */
 360             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 361
 362             fscal            = felec;
 363
 364             /* Update vectorial force */
 365             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 366             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 367             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 368
 369             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 370             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 371             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 372
 373             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 374
 375             /* Inner loop uses 167 flops */
 376         }
 377
 378         if(jidx<j_index_end)
 379         {
 380
 381             jnrA             = jjnr[jidx];
 382             j_coord_offsetA  = DIM*jnrA;
 383
 384             /* load j atom coordinates */
 385             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 386                                               &jx0,&jy0,&jz0);
 387
 388             /* Calculate displacement vector */
 389             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 390             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 391             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 392             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 393             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 394             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 395             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 396             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 397             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 398             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 399             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 400             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 401
 402             /* Calculate squared distance and things based on it */
 403             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 404             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 405             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 406             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 407
 408             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 409             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 410             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 411             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 412
 413             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 414             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 415             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 416
 417             /* Load parameters for j particles */
 418             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 419             vdwjidx0A        = 2*vdwtype[jnrA+0];
 420
 421             fjx0             = _fjsp_setzero_v2r8();
 422             fjy0             = _fjsp_setzero_v2r8();
 423             fjz0             = _fjsp_setzero_v2r8();
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 430
 431             /* Compute parameters for interactions between i and j atoms */
 432             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 433                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 434
 435             /* Calculate table index by multiplying r with table scale and truncate to integer */
 436             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 437             itab_tmp         = _fjsp_dtox_v2r8(rt);
 438             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 439             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 440             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 441
 442             vfconv.i[0]     *= 8;
 443             vfconv.i[1]     *= 8;
 444
 445             /* CUBIC SPLINE TABLE DISPERSION */
 446             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 447             F                = _fjsp_setzero_v2r8();
 448             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 449             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 450             H                = _fjsp_setzero_v2r8();
 451             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 452             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 453             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 454             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 455             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 456             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 457
 458             /* CUBIC SPLINE TABLE REPULSION */
 459             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 460             F                = _fjsp_setzero_v2r8();
 461             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 462             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 463             H                = _fjsp_setzero_v2r8();
 464             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 465             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 466             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 467             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 468             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 469             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 470             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 471             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 472
 473             /* Update potential sum for this i atom from the interaction with this j atom. */
 474             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 475             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 476
 477             fscal            = fvdw;
 478
 479             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 480
 481             /* Update vectorial force */
 482             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 483             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 484             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 485
 486             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 487             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 488             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             /* Compute parameters for interactions between i and j atoms */
 495             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 496
 497             /* REACTION-FIELD ELECTROSTATICS */
 498             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 499             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 500
 501             /* Update potential sum for this i atom from the interaction with this j atom. */
 502             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 503             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 504
 505             fscal            = felec;
 506
 507             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 508
 509             /* Update vectorial force */
 510             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 511             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 512             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 513
 514             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 515             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 516             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             /* Compute parameters for interactions between i and j atoms */
 523             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 524
 525             /* REACTION-FIELD ELECTROSTATICS */
 526             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 527             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 528
 529             /* Update potential sum for this i atom from the interaction with this j atom. */
 530             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 531             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 532
 533             fscal            = felec;
 534
 535             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 536
 537             /* Update vectorial force */
 538             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 539             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 540             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 541
 542             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 543             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 544             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             /* Compute parameters for interactions between i and j atoms */
 551             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 552
 553             /* REACTION-FIELD ELECTROSTATICS */
 554             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 555             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 556
 557             /* Update potential sum for this i atom from the interaction with this j atom. */
 558             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 559             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 560
 561             fscal            = felec;
 562
 563             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 564
 565             /* Update vectorial force */
 566             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 567             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 568             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 569
 570             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 571             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 572             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 573
 574             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 575
 576             /* Inner loop uses 167 flops */
 577         }
 578
 579         /* End of innermost loop */
 580
 581         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 582                                               f+i_coord_offset,fshift+i_shift_offset);
 583
 584         ggid                        = gid[iidx];
 585         /* Update potential energies */
 586         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 587         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 588
 589         /* Increment number of inner iterations */
 590         inneriter                  += j_index_end - j_index_start;
 591
 592         /* Outer loop uses 26 flops */
 593     }
 594
 595     /* Increment number of outer iterations */
 596     outeriter        += nri;
 597
 598     /* Update outer/inner flops */
 599
 600     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*167);
 601 }
 602 /*
 603  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 604  * Electrostatics interaction: ReactionField
 605  * VdW interaction:            CubicSplineTable
 606  * Geometry:                   Water4-Particle
 607  * Calculate force/pot:        Force
 608  */
 609 void
 610 nb_kernel_ElecRF_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 611                     (t_nblist                    * gmx_restrict       nlist,
 612                      rvec                        * gmx_restrict          xx,
 613                      rvec                        * gmx_restrict          ff,
 614                      t_forcerec                  * gmx_restrict          fr,
 615                      t_mdatoms                   * gmx_restrict     mdatoms,
 616                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 617                      t_nrnb                      * gmx_restrict        nrnb)
 618 {
 619     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 620      * just 0 for non-waters.
 621      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 622      * jnr indices corresponding to data put in the four positions in the SIMD register.
 623      */
 624     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 625     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 626     int              jnrA,jnrB;
 627     int              j_coord_offsetA,j_coord_offsetB;
 628     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 629     real             rcutoff_scalar;
 630     real             *shiftvec,*fshift,*x,*f;
 631     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 632     int              vdwioffset0;
 633     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 634     int              vdwioffset1;
 635     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 636     int              vdwioffset2;
 637     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 638     int              vdwioffset3;
 639     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 640     int              vdwjidx0A,vdwjidx0B;
 641     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 642     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 643     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 644     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 645     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 646     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 647     real             *charge;
 648     int              nvdwtype;
 649     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 650     int              *vdwtype;
 651     real             *vdwparam;
 652     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 653     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 654     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 655     real             *vftab;
 656     _fjsp_v2r8       itab_tmp;
 657     _fjsp_v2r8       dummy_mask,cutoff_mask;
 658     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 659     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 660     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 661
 662     x                = xx[0];
 663     f                = ff[0];
 664
 665     nri              = nlist->nri;
 666     iinr             = nlist->iinr;
 667     jindex           = nlist->jindex;
 668     jjnr             = nlist->jjnr;
 669     shiftidx         = nlist->shift;
 670     gid              = nlist->gid;
 671     shiftvec         = fr->shift_vec[0];
 672     fshift           = fr->fshift[0];
 673     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 674     charge           = mdatoms->chargeA;
 675     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 676     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 677     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 678     nvdwtype         = fr->ntype;
 679     vdwparam         = fr->nbfp;
 680     vdwtype          = mdatoms->typeA;
 681
 682     vftab            = kernel_data->table_vdw->data;
 683     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 684
 685     /* Setup water-specific parameters */
 686     inr              = nlist->iinr[0];
 687     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 688     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 689     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 690     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 691
 692     /* Avoid stupid compiler warnings */
 693     jnrA = jnrB = 0;
 694     j_coord_offsetA = 0;
 695     j_coord_offsetB = 0;
 696
 697     outeriter        = 0;
 698     inneriter        = 0;
 699
 700     /* Start outer loop over neighborlists */
 701     for(iidx=0; iidx<nri; iidx++)
 702     {
 703         /* Load shift vector for this list */
 704         i_shift_offset   = DIM*shiftidx[iidx];
 705
 706         /* Load limits for loop over neighbors */
 707         j_index_start    = jindex[iidx];
 708         j_index_end      = jindex[iidx+1];
 709
 710         /* Get outer coordinate index */
 711         inr              = iinr[iidx];
 712         i_coord_offset   = DIM*inr;
 713
 714         /* Load i particle coords and add shift vector */
 715         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 716                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 717
 718         fix0             = _fjsp_setzero_v2r8();
 719         fiy0             = _fjsp_setzero_v2r8();
 720         fiz0             = _fjsp_setzero_v2r8();
 721         fix1             = _fjsp_setzero_v2r8();
 722         fiy1             = _fjsp_setzero_v2r8();
 723         fiz1             = _fjsp_setzero_v2r8();
 724         fix2             = _fjsp_setzero_v2r8();
 725         fiy2             = _fjsp_setzero_v2r8();
 726         fiz2             = _fjsp_setzero_v2r8();
 727         fix3             = _fjsp_setzero_v2r8();
 728         fiy3             = _fjsp_setzero_v2r8();
 729         fiz3             = _fjsp_setzero_v2r8();
 730
 731         /* Start inner kernel loop */
 732         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 733         {
 734
 735             /* Get j neighbor index, and coordinate index */
 736             jnrA             = jjnr[jidx];
 737             jnrB             = jjnr[jidx+1];
 738             j_coord_offsetA  = DIM*jnrA;
 739             j_coord_offsetB  = DIM*jnrB;
 740
 741             /* load j atom coordinates */
 742             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 743                                               &jx0,&jy0,&jz0);
 744
 745             /* Calculate displacement vector */
 746             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 747             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 748             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 749             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 750             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 751             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 752             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 753             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 754             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 755             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 756             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 757             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 758
 759             /* Calculate squared distance and things based on it */
 760             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 761             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 762             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 763             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 764
 765             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 766             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 767             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 768             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 769
 770             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 771             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 772             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 773
 774             /* Load parameters for j particles */
 775             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 776             vdwjidx0A        = 2*vdwtype[jnrA+0];
 777             vdwjidx0B        = 2*vdwtype[jnrB+0];
 778
 779             fjx0             = _fjsp_setzero_v2r8();
 780             fjy0             = _fjsp_setzero_v2r8();
 781             fjz0             = _fjsp_setzero_v2r8();
 782
 783             /**************************
 784              * CALCULATE INTERACTIONS *
 785              **************************/
 786
 787             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 788
 789             /* Compute parameters for interactions between i and j atoms */
 790             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 791                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 792
 793             /* Calculate table index by multiplying r with table scale and truncate to integer */
 794             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 795             itab_tmp         = _fjsp_dtox_v2r8(rt);
 796             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 797             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 798             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 799
 800             vfconv.i[0]     *= 8;
 801             vfconv.i[1]     *= 8;
 802
 803             /* CUBIC SPLINE TABLE DISPERSION */
 804             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 805             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 806             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 807             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 808             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 809             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 810             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 811             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 812             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 813
 814             /* CUBIC SPLINE TABLE REPULSION */
 815             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 816             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 817             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 818             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 819             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 820             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 821             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 822             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 823             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 824             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 825
 826             fscal            = fvdw;
 827
 828             /* Update vectorial force */
 829             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 830             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 831             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 832
 833             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 834             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 835             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 836
 837             /**************************
 838              * CALCULATE INTERACTIONS *
 839              **************************/
 840
 841             /* Compute parameters for interactions between i and j atoms */
 842             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 843
 844             /* REACTION-FIELD ELECTROSTATICS */
 845             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 846
 847             fscal            = felec;
 848
 849             /* Update vectorial force */
 850             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 851             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 852             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 853
 854             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 855             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 856             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 857
 858             /**************************
 859              * CALCULATE INTERACTIONS *
 860              **************************/
 861
 862             /* Compute parameters for interactions between i and j atoms */
 863             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 864
 865             /* REACTION-FIELD ELECTROSTATICS */
 866             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 867
 868             fscal            = felec;
 869
 870             /* Update vectorial force */
 871             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 872             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 873             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 874
 875             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 876             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 877             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 878
 879             /**************************
 880              * CALCULATE INTERACTIONS *
 881              **************************/
 882
 883             /* Compute parameters for interactions between i and j atoms */
 884             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 885
 886             /* REACTION-FIELD ELECTROSTATICS */
 887             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 888
 889             fscal            = felec;
 890
 891             /* Update vectorial force */
 892             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 893             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 894             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 895
 896             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 897             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 898             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 899
 900             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 901
 902             /* Inner loop uses 144 flops */
 903         }
 904
 905         if(jidx<j_index_end)
 906         {
 907
 908             jnrA             = jjnr[jidx];
 909             j_coord_offsetA  = DIM*jnrA;
 910
 911             /* load j atom coordinates */
 912             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 913                                               &jx0,&jy0,&jz0);
 914
 915             /* Calculate displacement vector */
 916             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 917             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 918             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 919             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 920             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 921             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 922             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 923             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 924             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 925             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 926             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 927             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 928
 929             /* Calculate squared distance and things based on it */
 930             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 931             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 932             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 933             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 934
 935             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 936             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 937             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 938             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 939
 940             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 941             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 942             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 943
 944             /* Load parameters for j particles */
 945             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 946             vdwjidx0A        = 2*vdwtype[jnrA+0];
 947
 948             fjx0             = _fjsp_setzero_v2r8();
 949             fjy0             = _fjsp_setzero_v2r8();
 950             fjz0             = _fjsp_setzero_v2r8();
 951
 952             /**************************
 953              * CALCULATE INTERACTIONS *
 954              **************************/
 955
 956             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 957
 958             /* Compute parameters for interactions between i and j atoms */
 959             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 960                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 961
 962             /* Calculate table index by multiplying r with table scale and truncate to integer */
 963             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 964             itab_tmp         = _fjsp_dtox_v2r8(rt);
 965             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 966             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 967             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 968
 969             vfconv.i[0]     *= 8;
 970             vfconv.i[1]     *= 8;
 971
 972             /* CUBIC SPLINE TABLE DISPERSION */
 973             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 974             F                = _fjsp_setzero_v2r8();
 975             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 976             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 977             H                = _fjsp_setzero_v2r8();
 978             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 979             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 980             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 981             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 982
 983             /* CUBIC SPLINE TABLE REPULSION */
 984             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 985             F                = _fjsp_setzero_v2r8();
 986             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 987             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 988             H                = _fjsp_setzero_v2r8();
 989             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 990             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 991             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 992             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 993             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 994
 995             fscal            = fvdw;
 996
 997             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 998
 999             /* Update vectorial force */
1000             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1001             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1002             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1003
1004             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1005             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1006             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1007
1008             /**************************
1009              * CALCULATE INTERACTIONS *
1010              **************************/
1011
1012             /* Compute parameters for interactions between i and j atoms */
1013             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1014
1015             /* REACTION-FIELD ELECTROSTATICS */
1016             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1017
1018             fscal            = felec;
1019
1020             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1021
1022             /* Update vectorial force */
1023             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1024             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1025             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1026
1027             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1028             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1029             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1030
1031             /**************************
1032              * CALCULATE INTERACTIONS *
1033              **************************/
1034
1035             /* Compute parameters for interactions between i and j atoms */
1036             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1037
1038             /* REACTION-FIELD ELECTROSTATICS */
1039             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1040
1041             fscal            = felec;
1042
1043             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1044
1045             /* Update vectorial force */
1046             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1047             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1048             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1049
1050             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1051             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1052             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1053
1054             /**************************
1055              * CALCULATE INTERACTIONS *
1056              **************************/
1057
1058             /* Compute parameters for interactions between i and j atoms */
1059             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1060
1061             /* REACTION-FIELD ELECTROSTATICS */
1062             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1063
1064             fscal            = felec;
1065
1066             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1067
1068             /* Update vectorial force */
1069             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1070             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1071             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1072
1073             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1074             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1075             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1076
1077             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1078
1079             /* Inner loop uses 144 flops */
1080         }
1081
1082         /* End of innermost loop */
1083
1084         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1085                                               f+i_coord_offset,fshift+i_shift_offset);
1086
1087         /* Increment number of inner iterations */
1088         inneriter                  += j_index_end - j_index_start;
1089
1090         /* Outer loop uses 24 flops */
1091     }
1092
1093     /* Increment number of outer iterations */
1094     outeriter        += nri;
1095
1096     /* Update outer/inner flops */
1097
1098     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*144);
1099 }