src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_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_ElecRFCut_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 121     charge           = mdatoms->chargeA;
 122     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 123     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 124     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 125     nvdwtype         = fr->ntype;
 126     vdwparam         = fr->nbfp;
 127     vdwtype          = mdatoms->typeA;
 128
 129     vftab            = kernel_data->table_vdw->data;
 130     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 131
 132     /* Setup water-specific parameters */
 133     inr              = nlist->iinr[0];
 134     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 135     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 136     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 137     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 138
 139     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 140     rcutoff_scalar   = fr->ic->rcoulomb;
 141     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 142     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 143
 144     /* Avoid stupid compiler warnings */
 145     jnrA = jnrB = 0;
 146     j_coord_offsetA = 0;
 147     j_coord_offsetB = 0;
 148
 149     outeriter        = 0;
 150     inneriter        = 0;
 151
 152     /* Start outer loop over neighborlists */
 153     for(iidx=0; iidx<nri; iidx++)
 154     {
 155         /* Load shift vector for this list */
 156         i_shift_offset   = DIM*shiftidx[iidx];
 157
 158         /* Load limits for loop over neighbors */
 159         j_index_start    = jindex[iidx];
 160         j_index_end      = jindex[iidx+1];
 161
 162         /* Get outer coordinate index */
 163         inr              = iinr[iidx];
 164         i_coord_offset   = DIM*inr;
 165
 166         /* Load i particle coords and add shift vector */
 167         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 168                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 169
 170         fix0             = _fjsp_setzero_v2r8();
 171         fiy0             = _fjsp_setzero_v2r8();
 172         fiz0             = _fjsp_setzero_v2r8();
 173         fix1             = _fjsp_setzero_v2r8();
 174         fiy1             = _fjsp_setzero_v2r8();
 175         fiz1             = _fjsp_setzero_v2r8();
 176         fix2             = _fjsp_setzero_v2r8();
 177         fiy2             = _fjsp_setzero_v2r8();
 178         fiz2             = _fjsp_setzero_v2r8();
 179         fix3             = _fjsp_setzero_v2r8();
 180         fiy3             = _fjsp_setzero_v2r8();
 181         fiz3             = _fjsp_setzero_v2r8();
 182
 183         /* Reset potential sums */
 184         velecsum         = _fjsp_setzero_v2r8();
 185         vvdwsum          = _fjsp_setzero_v2r8();
 186
 187         /* Start inner kernel loop */
 188         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 189         {
 190
 191             /* Get j neighbor index, and coordinate index */
 192             jnrA             = jjnr[jidx];
 193             jnrB             = jjnr[jidx+1];
 194             j_coord_offsetA  = DIM*jnrA;
 195             j_coord_offsetB  = DIM*jnrB;
 196
 197             /* load j atom coordinates */
 198             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 199                                               &jx0,&jy0,&jz0);
 200
 201             /* Calculate displacement vector */
 202             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 203             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 204             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 205             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 206             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 207             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 208             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 209             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 210             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 211             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 212             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 213             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 214
 215             /* Calculate squared distance and things based on it */
 216             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 217             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 218             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 219             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 220
 221             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 222             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 223             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 224             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 225
 226             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 227             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 228             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 229
 230             /* Load parameters for j particles */
 231             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 232             vdwjidx0A        = 2*vdwtype[jnrA+0];
 233             vdwjidx0B        = 2*vdwtype[jnrB+0];
 234
 235             fjx0             = _fjsp_setzero_v2r8();
 236             fjy0             = _fjsp_setzero_v2r8();
 237             fjz0             = _fjsp_setzero_v2r8();
 238
 239             /**************************
 240              * CALCULATE INTERACTIONS *
 241              **************************/
 242
 243             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 244
 245             /* Compute parameters for interactions between i and j atoms */
 246             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 247                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 248
 249             /* Calculate table index by multiplying r with table scale and truncate to integer */
 250             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 251             itab_tmp         = _fjsp_dtox_v2r8(rt);
 252             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 253             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 254             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 255
 256             vfconv.i[0]     *= 8;
 257             vfconv.i[1]     *= 8;
 258
 259             /* CUBIC SPLINE TABLE DISPERSION */
 260             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 261             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 262             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 263             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 264             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 265             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 266             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 267             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 268             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 269             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 270             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 271
 272             /* CUBIC SPLINE TABLE REPULSION */
 273             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 274             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 276             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 277             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 278             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 279             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 280             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 281             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 282             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 283             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 284             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 285             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 286
 287             /* Update potential sum for this i atom from the interaction with this j atom. */
 288             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 289
 290             fscal            = fvdw;
 291
 292             /* Update vectorial force */
 293             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 294             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 295             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 296
 297             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 298             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 299             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 300
 301             /**************************
 302              * CALCULATE INTERACTIONS *
 303              **************************/
 304
 305             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 306             {
 307
 308             /* Compute parameters for interactions between i and j atoms */
 309             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 310
 311             /* REACTION-FIELD ELECTROSTATICS */
 312             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 313             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 314
 315             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 316
 317             /* Update potential sum for this i atom from the interaction with this j atom. */
 318             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 319             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 320
 321             fscal            = felec;
 322
 323             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 324
 325             /* Update vectorial force */
 326             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 327             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 328             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 329
 330             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 331             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 332             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 333
 334             }
 335
 336             /**************************
 337              * CALCULATE INTERACTIONS *
 338              **************************/
 339
 340             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 341             {
 342
 343             /* Compute parameters for interactions between i and j atoms */
 344             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 345
 346             /* REACTION-FIELD ELECTROSTATICS */
 347             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 348             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 349
 350             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 354             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 355
 356             fscal            = felec;
 357
 358             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 359
 360             /* Update vectorial force */
 361             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 362             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 363             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 364
 365             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 366             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 367             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 368
 369             }
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 376             {
 377
 378             /* Compute parameters for interactions between i and j atoms */
 379             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 380
 381             /* REACTION-FIELD ELECTROSTATICS */
 382             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 383             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 384
 385             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 386
 387             /* Update potential sum for this i atom from the interaction with this j atom. */
 388             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 389             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 390
 391             fscal            = felec;
 392
 393             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 394
 395             /* Update vectorial force */
 396             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 397             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 398             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 399
 400             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 401             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 402             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 403
 404             }
 405
 406             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 407
 408             /* Inner loop uses 179 flops */
 409         }
 410
 411         if(jidx<j_index_end)
 412         {
 413
 414             jnrA             = jjnr[jidx];
 415             j_coord_offsetA  = DIM*jnrA;
 416
 417             /* load j atom coordinates */
 418             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 419                                               &jx0,&jy0,&jz0);
 420
 421             /* Calculate displacement vector */
 422             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 423             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 424             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 425             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 426             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 427             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 428             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 429             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 430             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 431             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 432             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 433             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 434
 435             /* Calculate squared distance and things based on it */
 436             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 437             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 438             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 439             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 440
 441             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 442             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 443             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 444             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 445
 446             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 447             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 448             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 449
 450             /* Load parameters for j particles */
 451             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 452             vdwjidx0A        = 2*vdwtype[jnrA+0];
 453
 454             fjx0             = _fjsp_setzero_v2r8();
 455             fjy0             = _fjsp_setzero_v2r8();
 456             fjz0             = _fjsp_setzero_v2r8();
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 463
 464             /* Compute parameters for interactions between i and j atoms */
 465             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 466                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 467
 468             /* Calculate table index by multiplying r with table scale and truncate to integer */
 469             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 470             itab_tmp         = _fjsp_dtox_v2r8(rt);
 471             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 472             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 473             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 474
 475             vfconv.i[0]     *= 8;
 476             vfconv.i[1]     *= 8;
 477
 478             /* CUBIC SPLINE TABLE DISPERSION */
 479             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 480             F                = _fjsp_setzero_v2r8();
 481             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 482             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 483             H                = _fjsp_setzero_v2r8();
 484             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 485             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 486             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 487             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 488             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 489             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 490
 491             /* CUBIC SPLINE TABLE REPULSION */
 492             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 493             F                = _fjsp_setzero_v2r8();
 494             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 495             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 496             H                = _fjsp_setzero_v2r8();
 497             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 498             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 499             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 500             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 501             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 502             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 503             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 504             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 505
 506             /* Update potential sum for this i atom from the interaction with this j atom. */
 507             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 508             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 509
 510             fscal            = fvdw;
 511
 512             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 513
 514             /* Update vectorial force */
 515             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 516             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 517             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 518
 519             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 520             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 521             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 522
 523             /**************************
 524              * CALCULATE INTERACTIONS *
 525              **************************/
 526
 527             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 528             {
 529
 530             /* Compute parameters for interactions between i and j atoms */
 531             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 532
 533             /* REACTION-FIELD ELECTROSTATICS */
 534             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 535             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 536
 537             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 538
 539             /* Update potential sum for this i atom from the interaction with this j atom. */
 540             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 541             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 542             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 543
 544             fscal            = felec;
 545
 546             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 547
 548             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 549
 550             /* Update vectorial force */
 551             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 552             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 553             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 554
 555             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 556             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 557             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 558
 559             }
 560
 561             /**************************
 562              * CALCULATE INTERACTIONS *
 563              **************************/
 564
 565             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 566             {
 567
 568             /* Compute parameters for interactions between i and j atoms */
 569             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 570
 571             /* REACTION-FIELD ELECTROSTATICS */
 572             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 573             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 574
 575             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 576
 577             /* Update potential sum for this i atom from the interaction with this j atom. */
 578             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 579             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 580             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 581
 582             fscal            = felec;
 583
 584             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 585
 586             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 587
 588             /* Update vectorial force */
 589             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 590             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 591             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 592
 593             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 594             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 595             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 596
 597             }
 598
 599             /**************************
 600              * CALCULATE INTERACTIONS *
 601              **************************/
 602
 603             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 604             {
 605
 606             /* Compute parameters for interactions between i and j atoms */
 607             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 608
 609             /* REACTION-FIELD ELECTROSTATICS */
 610             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 611             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 612
 613             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 614
 615             /* Update potential sum for this i atom from the interaction with this j atom. */
 616             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 617             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 618             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 619
 620             fscal            = felec;
 621
 622             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 623
 624             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 625
 626             /* Update vectorial force */
 627             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 628             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 629             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 630
 631             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 632             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 633             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 634
 635             }
 636
 637             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 638
 639             /* Inner loop uses 179 flops */
 640         }
 641
 642         /* End of innermost loop */
 643
 644         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 645                                               f+i_coord_offset,fshift+i_shift_offset);
 646
 647         ggid                        = gid[iidx];
 648         /* Update potential energies */
 649         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 650         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 651
 652         /* Increment number of inner iterations */
 653         inneriter                  += j_index_end - j_index_start;
 654
 655         /* Outer loop uses 26 flops */
 656     }
 657
 658     /* Increment number of outer iterations */
 659     outeriter        += nri;
 660
 661     /* Update outer/inner flops */
 662
 663     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*179);
 664 }
 665 /*
 666  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 667  * Electrostatics interaction: ReactionField
 668  * VdW interaction:            CubicSplineTable
 669  * Geometry:                   Water4-Particle
 670  * Calculate force/pot:        Force
 671  */
 672 void
 673 nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 674                     (t_nblist                    * gmx_restrict       nlist,
 675                      rvec                        * gmx_restrict          xx,
 676                      rvec                        * gmx_restrict          ff,
 677                      struct t_forcerec           * gmx_restrict          fr,
 678                      t_mdatoms                   * gmx_restrict     mdatoms,
 679                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 680                      t_nrnb                      * gmx_restrict        nrnb)
 681 {
 682     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 683      * just 0 for non-waters.
 684      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 685      * jnr indices corresponding to data put in the four positions in the SIMD register.
 686      */
 687     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 688     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 689     int              jnrA,jnrB;
 690     int              j_coord_offsetA,j_coord_offsetB;
 691     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 692     real             rcutoff_scalar;
 693     real             *shiftvec,*fshift,*x,*f;
 694     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 695     int              vdwioffset0;
 696     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 697     int              vdwioffset1;
 698     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 699     int              vdwioffset2;
 700     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 701     int              vdwioffset3;
 702     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 703     int              vdwjidx0A,vdwjidx0B;
 704     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 705     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 706     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 707     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 708     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 709     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 710     real             *charge;
 711     int              nvdwtype;
 712     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 713     int              *vdwtype;
 714     real             *vdwparam;
 715     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 716     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 717     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 718     real             *vftab;
 719     _fjsp_v2r8       itab_tmp;
 720     _fjsp_v2r8       dummy_mask,cutoff_mask;
 721     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 722     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 723     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 724
 725     x                = xx[0];
 726     f                = ff[0];
 727
 728     nri              = nlist->nri;
 729     iinr             = nlist->iinr;
 730     jindex           = nlist->jindex;
 731     jjnr             = nlist->jjnr;
 732     shiftidx         = nlist->shift;
 733     gid              = nlist->gid;
 734     shiftvec         = fr->shift_vec[0];
 735     fshift           = fr->fshift[0];
 736     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 737     charge           = mdatoms->chargeA;
 738     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 739     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 740     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 741     nvdwtype         = fr->ntype;
 742     vdwparam         = fr->nbfp;
 743     vdwtype          = mdatoms->typeA;
 744
 745     vftab            = kernel_data->table_vdw->data;
 746     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 747
 748     /* Setup water-specific parameters */
 749     inr              = nlist->iinr[0];
 750     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 751     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 752     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 753     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 754
 755     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 756     rcutoff_scalar   = fr->ic->rcoulomb;
 757     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 758     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 759
 760     /* Avoid stupid compiler warnings */
 761     jnrA = jnrB = 0;
 762     j_coord_offsetA = 0;
 763     j_coord_offsetB = 0;
 764
 765     outeriter        = 0;
 766     inneriter        = 0;
 767
 768     /* Start outer loop over neighborlists */
 769     for(iidx=0; iidx<nri; iidx++)
 770     {
 771         /* Load shift vector for this list */
 772         i_shift_offset   = DIM*shiftidx[iidx];
 773
 774         /* Load limits for loop over neighbors */
 775         j_index_start    = jindex[iidx];
 776         j_index_end      = jindex[iidx+1];
 777
 778         /* Get outer coordinate index */
 779         inr              = iinr[iidx];
 780         i_coord_offset   = DIM*inr;
 781
 782         /* Load i particle coords and add shift vector */
 783         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 784                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 785
 786         fix0             = _fjsp_setzero_v2r8();
 787         fiy0             = _fjsp_setzero_v2r8();
 788         fiz0             = _fjsp_setzero_v2r8();
 789         fix1             = _fjsp_setzero_v2r8();
 790         fiy1             = _fjsp_setzero_v2r8();
 791         fiz1             = _fjsp_setzero_v2r8();
 792         fix2             = _fjsp_setzero_v2r8();
 793         fiy2             = _fjsp_setzero_v2r8();
 794         fiz2             = _fjsp_setzero_v2r8();
 795         fix3             = _fjsp_setzero_v2r8();
 796         fiy3             = _fjsp_setzero_v2r8();
 797         fiz3             = _fjsp_setzero_v2r8();
 798
 799         /* Start inner kernel loop */
 800         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 801         {
 802
 803             /* Get j neighbor index, and coordinate index */
 804             jnrA             = jjnr[jidx];
 805             jnrB             = jjnr[jidx+1];
 806             j_coord_offsetA  = DIM*jnrA;
 807             j_coord_offsetB  = DIM*jnrB;
 808
 809             /* load j atom coordinates */
 810             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 811                                               &jx0,&jy0,&jz0);
 812
 813             /* Calculate displacement vector */
 814             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 815             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 816             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 817             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 818             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 819             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 820             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 821             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 822             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 823             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 824             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 825             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 826
 827             /* Calculate squared distance and things based on it */
 828             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 829             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 830             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 831             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 832
 833             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 834             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 835             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 836             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 837
 838             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 839             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 840             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 841
 842             /* Load parameters for j particles */
 843             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 844             vdwjidx0A        = 2*vdwtype[jnrA+0];
 845             vdwjidx0B        = 2*vdwtype[jnrB+0];
 846
 847             fjx0             = _fjsp_setzero_v2r8();
 848             fjy0             = _fjsp_setzero_v2r8();
 849             fjz0             = _fjsp_setzero_v2r8();
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 856
 857             /* Compute parameters for interactions between i and j atoms */
 858             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 859                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 860
 861             /* Calculate table index by multiplying r with table scale and truncate to integer */
 862             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 863             itab_tmp         = _fjsp_dtox_v2r8(rt);
 864             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 865             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 866             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 867
 868             vfconv.i[0]     *= 8;
 869             vfconv.i[1]     *= 8;
 870
 871             /* CUBIC SPLINE TABLE DISPERSION */
 872             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 873             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 874             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 875             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 876             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 877             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 878             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 879             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 880             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 881
 882             /* CUBIC SPLINE TABLE REPULSION */
 883             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 884             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 885             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 886             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 887             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 888             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 889             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 890             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 891             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 892             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 893
 894             fscal            = fvdw;
 895
 896             /* Update vectorial force */
 897             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 898             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 899             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 900
 901             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 902             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 903             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 904
 905             /**************************
 906              * CALCULATE INTERACTIONS *
 907              **************************/
 908
 909             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 910             {
 911
 912             /* Compute parameters for interactions between i and j atoms */
 913             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 914
 915             /* REACTION-FIELD ELECTROSTATICS */
 916             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 917
 918             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 919
 920             fscal            = felec;
 921
 922             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 923
 924             /* Update vectorial force */
 925             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 926             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 927             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 928
 929             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 930             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 931             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 932
 933             }
 934
 935             /**************************
 936              * CALCULATE INTERACTIONS *
 937              **************************/
 938
 939             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 940             {
 941
 942             /* Compute parameters for interactions between i and j atoms */
 943             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 944
 945             /* REACTION-FIELD ELECTROSTATICS */
 946             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 947
 948             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 949
 950             fscal            = felec;
 951
 952             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 953
 954             /* Update vectorial force */
 955             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 956             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 957             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 958
 959             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 960             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 961             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 962
 963             }
 964
 965             /**************************
 966              * CALCULATE INTERACTIONS *
 967              **************************/
 968
 969             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 970             {
 971
 972             /* Compute parameters for interactions between i and j atoms */
 973             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 974
 975             /* REACTION-FIELD ELECTROSTATICS */
 976             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 977
 978             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 979
 980             fscal            = felec;
 981
 982             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 983
 984             /* Update vectorial force */
 985             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 986             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 987             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 988
 989             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 990             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 991             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 992
 993             }
 994
 995             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 996
 997             /* Inner loop uses 153 flops */
 998         }
 999
1000         if(jidx<j_index_end)
1001         {
1002
1003             jnrA             = jjnr[jidx];
1004             j_coord_offsetA  = DIM*jnrA;
1005
1006             /* load j atom coordinates */
1007             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1008                                               &jx0,&jy0,&jz0);
1009
1010             /* Calculate displacement vector */
1011             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1012             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1013             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1014             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1015             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1016             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1017             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1018             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1019             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1020             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1021             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1022             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1023
1024             /* Calculate squared distance and things based on it */
1025             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1026             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1027             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1028             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1029
1030             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1031             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1032             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1033             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1034
1035             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1036             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1037             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1038
1039             /* Load parameters for j particles */
1040             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1041             vdwjidx0A        = 2*vdwtype[jnrA+0];
1042
1043             fjx0             = _fjsp_setzero_v2r8();
1044             fjy0             = _fjsp_setzero_v2r8();
1045             fjz0             = _fjsp_setzero_v2r8();
1046
1047             /**************************
1048              * CALCULATE INTERACTIONS *
1049              **************************/
1050
1051             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1052
1053             /* Compute parameters for interactions between i and j atoms */
1054             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1055                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1056
1057             /* Calculate table index by multiplying r with table scale and truncate to integer */
1058             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1059             itab_tmp         = _fjsp_dtox_v2r8(rt);
1060             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1061             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1062             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1063
1064             vfconv.i[0]     *= 8;
1065             vfconv.i[1]     *= 8;
1066
1067             /* CUBIC SPLINE TABLE DISPERSION */
1068             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1069             F                = _fjsp_setzero_v2r8();
1070             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1071             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1072             H                = _fjsp_setzero_v2r8();
1073             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1074             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1075             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1076             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1077
1078             /* CUBIC SPLINE TABLE REPULSION */
1079             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1080             F                = _fjsp_setzero_v2r8();
1081             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1082             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1083             H                = _fjsp_setzero_v2r8();
1084             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1085             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1086             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1087             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1088             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1089
1090             fscal            = fvdw;
1091
1092             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1093
1094             /* Update vectorial force */
1095             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1096             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1097             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1098
1099             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1100             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1101             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1108             {
1109
1110             /* Compute parameters for interactions between i and j atoms */
1111             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1112
1113             /* REACTION-FIELD ELECTROSTATICS */
1114             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1115
1116             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1117
1118             fscal            = felec;
1119
1120             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1121
1122             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1123
1124             /* Update vectorial force */
1125             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1126             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1127             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1128
1129             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1130             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1131             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1132
1133             }
1134
1135             /**************************
1136              * CALCULATE INTERACTIONS *
1137              **************************/
1138
1139             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1140             {
1141
1142             /* Compute parameters for interactions between i and j atoms */
1143             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1144
1145             /* REACTION-FIELD ELECTROSTATICS */
1146             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1147
1148             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1149
1150             fscal            = felec;
1151
1152             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1153
1154             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1155
1156             /* Update vectorial force */
1157             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1158             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1159             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1160
1161             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1162             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1163             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1164
1165             }
1166
1167             /**************************
1168              * CALCULATE INTERACTIONS *
1169              **************************/
1170
1171             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1172             {
1173
1174             /* Compute parameters for interactions between i and j atoms */
1175             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1176
1177             /* REACTION-FIELD ELECTROSTATICS */
1178             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1179
1180             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1181
1182             fscal            = felec;
1183
1184             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1185
1186             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1187
1188             /* Update vectorial force */
1189             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1190             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1191             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1192
1193             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1194             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1195             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1196
1197             }
1198
1199             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1200
1201             /* Inner loop uses 153 flops */
1202         }
1203
1204         /* End of innermost loop */
1205
1206         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1207                                               f+i_coord_offset,fshift+i_shift_offset);
1208
1209         /* Increment number of inner iterations */
1210         inneriter                  += j_index_end - j_index_start;
1211
1212         /* Outer loop uses 24 flops */
1213     }
1214
1215     /* Increment number of outer iterations */
1216     outeriter        += nri;
1217
1218     /* Update outer/inner flops */
1219
1220     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*153);
1221 }