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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 120     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 121     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 132     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 133     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 137     rcutoff_scalar   = fr->rcoulomb;
 138     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 139     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 140
 141     /* Avoid stupid compiler warnings */
 142     jnrA = jnrB = 0;
 143     j_coord_offsetA = 0;
 144     j_coord_offsetB = 0;
 145
 146     outeriter        = 0;
 147     inneriter        = 0;
 148
 149     /* Start outer loop over neighborlists */
 150     for(iidx=0; iidx<nri; iidx++)
 151     {
 152         /* Load shift vector for this list */
 153         i_shift_offset   = DIM*shiftidx[iidx];
 154
 155         /* Load limits for loop over neighbors */
 156         j_index_start    = jindex[iidx];
 157         j_index_end      = jindex[iidx+1];
 158
 159         /* Get outer coordinate index */
 160         inr              = iinr[iidx];
 161         i_coord_offset   = DIM*inr;
 162
 163         /* Load i particle coords and add shift vector */
 164         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 165                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 166
 167         fix0             = _fjsp_setzero_v2r8();
 168         fiy0             = _fjsp_setzero_v2r8();
 169         fiz0             = _fjsp_setzero_v2r8();
 170         fix1             = _fjsp_setzero_v2r8();
 171         fiy1             = _fjsp_setzero_v2r8();
 172         fiz1             = _fjsp_setzero_v2r8();
 173         fix2             = _fjsp_setzero_v2r8();
 174         fiy2             = _fjsp_setzero_v2r8();
 175         fiz2             = _fjsp_setzero_v2r8();
 176
 177         /* Reset potential sums */
 178         velecsum         = _fjsp_setzero_v2r8();
 179         vvdwsum          = _fjsp_setzero_v2r8();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 193                                               &jx0,&jy0,&jz0);
 194
 195             /* Calculate displacement vector */
 196             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 197             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 198             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 199             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 200             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 201             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 202             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 203             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 204             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 205
 206             /* Calculate squared distance and things based on it */
 207             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 208             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 209             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 210
 211             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 212             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 213             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 214
 215             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 216             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 217             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 218
 219             /* Load parameters for j particles */
 220             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 221             vdwjidx0A        = 2*vdwtype[jnrA+0];
 222             vdwjidx0B        = 2*vdwtype[jnrB+0];
 223
 224             fjx0             = _fjsp_setzero_v2r8();
 225             fjy0             = _fjsp_setzero_v2r8();
 226             fjz0             = _fjsp_setzero_v2r8();
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 233             {
 234
 235             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 239             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 240                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 241
 242             /* Calculate table index by multiplying r with table scale and truncate to integer */
 243             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 244             itab_tmp         = _fjsp_dtox_v2r8(rt);
 245             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 246             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 247             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 248
 249             vfconv.i[0]     *= 8;
 250             vfconv.i[1]     *= 8;
 251
 252             /* REACTION-FIELD ELECTROSTATICS */
 253             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 254             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 255
 256             /* CUBIC SPLINE TABLE DISPERSION */
 257             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 258             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 259             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 260             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 261             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 262             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 263             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 264             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 265             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 266             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 267             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 268
 269             /* CUBIC SPLINE TABLE REPULSION */
 270             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 271             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 272             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 273             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 274             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 276             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 277             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 278             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 279             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 280             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 281             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 282             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 283
 284             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 285
 286             /* Update potential sum for this i atom from the interaction with this j atom. */
 287             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 288             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 289             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 290             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 291
 292             fscal            = _fjsp_add_v2r8(felec,fvdw);
 293
 294             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 295
 296             /* Update vectorial force */
 297             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 298             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 299             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 300
 301             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 302             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 303             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 304
 305             }
 306
 307             /**************************
 308              * CALCULATE INTERACTIONS *
 309              **************************/
 310
 311             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 312             {
 313
 314             /* Compute parameters for interactions between i and j atoms */
 315             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 316
 317             /* REACTION-FIELD ELECTROSTATICS */
 318             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 319             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 320
 321             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 322
 323             /* Update potential sum for this i atom from the interaction with this j atom. */
 324             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 325             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 326
 327             fscal            = felec;
 328
 329             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 330
 331             /* Update vectorial force */
 332             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 333             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 334             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 335
 336             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 337             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 338             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 339
 340             }
 341
 342             /**************************
 343              * CALCULATE INTERACTIONS *
 344              **************************/
 345
 346             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 347             {
 348
 349             /* Compute parameters for interactions between i and j atoms */
 350             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 351
 352             /* REACTION-FIELD ELECTROSTATICS */
 353             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 354             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 355
 356             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 357
 358             /* Update potential sum for this i atom from the interaction with this j atom. */
 359             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 360             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 361
 362             fscal            = felec;
 363
 364             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 365
 366             /* Update vectorial force */
 367             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 368             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 369             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 370
 371             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 372             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 373             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 374
 375             }
 376
 377             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 378
 379             /* Inner loop uses 156 flops */
 380         }
 381
 382         if(jidx<j_index_end)
 383         {
 384
 385             jnrA             = jjnr[jidx];
 386             j_coord_offsetA  = DIM*jnrA;
 387
 388             /* load j atom coordinates */
 389             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 390                                               &jx0,&jy0,&jz0);
 391
 392             /* Calculate displacement vector */
 393             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 394             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 395             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 396             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 397             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 398             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 399             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 400             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 401             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 402
 403             /* Calculate squared distance and things based on it */
 404             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 405             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 406             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 407
 408             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 409             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 410             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 411
 412             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 413             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 414             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 415
 416             /* Load parameters for j particles */
 417             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 418             vdwjidx0A        = 2*vdwtype[jnrA+0];
 419
 420             fjx0             = _fjsp_setzero_v2r8();
 421             fjy0             = _fjsp_setzero_v2r8();
 422             fjz0             = _fjsp_setzero_v2r8();
 423
 424             /**************************
 425              * CALCULATE INTERACTIONS *
 426              **************************/
 427
 428             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 429             {
 430
 431             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 432
 433             /* Compute parameters for interactions between i and j atoms */
 434             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 435             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 436
 437             /* Calculate table index by multiplying r with table scale and truncate to integer */
 438             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 439             itab_tmp         = _fjsp_dtox_v2r8(rt);
 440             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 441             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 442             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 443
 444             vfconv.i[0]     *= 8;
 445             vfconv.i[1]     *= 8;
 446
 447             /* REACTION-FIELD ELECTROSTATICS */
 448             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 449             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 450
 451             /* CUBIC SPLINE TABLE DISPERSION */
 452             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 453             F                = _fjsp_setzero_v2r8();
 454             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 455             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 456             H                = _fjsp_setzero_v2r8();
 457             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 458             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 459             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 460             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 461             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 462             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 463
 464             /* CUBIC SPLINE TABLE REPULSION */
 465             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 466             F                = _fjsp_setzero_v2r8();
 467             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 468             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 469             H                = _fjsp_setzero_v2r8();
 470             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 471             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 472             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 473             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 474             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 475             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 476             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 477             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 478
 479             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 480
 481             /* Update potential sum for this i atom from the interaction with this j atom. */
 482             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 483             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 484             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 485             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 486             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 487             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 488
 489             fscal            = _fjsp_add_v2r8(felec,fvdw);
 490
 491             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 492
 493             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 494
 495             /* Update vectorial force */
 496             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 497             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 498             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 499
 500             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 501             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 502             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 503
 504             }
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 511             {
 512
 513             /* Compute parameters for interactions between i and j atoms */
 514             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 515
 516             /* REACTION-FIELD ELECTROSTATICS */
 517             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 518             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 519
 520             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 521
 522             /* Update potential sum for this i atom from the interaction with this j atom. */
 523             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 524             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 525             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 526
 527             fscal            = felec;
 528
 529             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 530
 531             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 532
 533             /* Update vectorial force */
 534             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 535             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 536             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 537
 538             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 539             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 540             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 541
 542             }
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 549             {
 550
 551             /* Compute parameters for interactions between i and j atoms */
 552             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 553
 554             /* REACTION-FIELD ELECTROSTATICS */
 555             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 556             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 557
 558             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 559
 560             /* Update potential sum for this i atom from the interaction with this j atom. */
 561             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 562             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 563             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 568
 569             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 570
 571             /* Update vectorial force */
 572             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 573             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 574             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 575
 576             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 577             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 578             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 579
 580             }
 581
 582             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 583
 584             /* Inner loop uses 156 flops */
 585         }
 586
 587         /* End of innermost loop */
 588
 589         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 590                                               f+i_coord_offset,fshift+i_shift_offset);
 591
 592         ggid                        = gid[iidx];
 593         /* Update potential energies */
 594         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 595         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 596
 597         /* Increment number of inner iterations */
 598         inneriter                  += j_index_end - j_index_start;
 599
 600         /* Outer loop uses 20 flops */
 601     }
 602
 603     /* Increment number of outer iterations */
 604     outeriter        += nri;
 605
 606     /* Update outer/inner flops */
 607
 608     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*156);
 609 }
 610 /*
 611  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 612  * Electrostatics interaction: ReactionField
 613  * VdW interaction:            CubicSplineTable
 614  * Geometry:                   Water3-Particle
 615  * Calculate force/pot:        Force
 616  */
 617 void
 618 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 619                     (t_nblist                    * gmx_restrict       nlist,
 620                      rvec                        * gmx_restrict          xx,
 621                      rvec                        * gmx_restrict          ff,
 622                      t_forcerec                  * gmx_restrict          fr,
 623                      t_mdatoms                   * gmx_restrict     mdatoms,
 624                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 625                      t_nrnb                      * gmx_restrict        nrnb)
 626 {
 627     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 628      * just 0 for non-waters.
 629      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 630      * jnr indices corresponding to data put in the four positions in the SIMD register.
 631      */
 632     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 633     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 634     int              jnrA,jnrB;
 635     int              j_coord_offsetA,j_coord_offsetB;
 636     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 637     real             rcutoff_scalar;
 638     real             *shiftvec,*fshift,*x,*f;
 639     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 640     int              vdwioffset0;
 641     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 642     int              vdwioffset1;
 643     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 644     int              vdwioffset2;
 645     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 646     int              vdwjidx0A,vdwjidx0B;
 647     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 648     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 649     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 650     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 651     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 652     real             *charge;
 653     int              nvdwtype;
 654     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 655     int              *vdwtype;
 656     real             *vdwparam;
 657     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 658     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 659     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 660     real             *vftab;
 661     _fjsp_v2r8       itab_tmp;
 662     _fjsp_v2r8       dummy_mask,cutoff_mask;
 663     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 664     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 665     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 666
 667     x                = xx[0];
 668     f                = ff[0];
 669
 670     nri              = nlist->nri;
 671     iinr             = nlist->iinr;
 672     jindex           = nlist->jindex;
 673     jjnr             = nlist->jjnr;
 674     shiftidx         = nlist->shift;
 675     gid              = nlist->gid;
 676     shiftvec         = fr->shift_vec[0];
 677     fshift           = fr->fshift[0];
 678     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 679     charge           = mdatoms->chargeA;
 680     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 681     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 682     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 683     nvdwtype         = fr->ntype;
 684     vdwparam         = fr->nbfp;
 685     vdwtype          = mdatoms->typeA;
 686
 687     vftab            = kernel_data->table_vdw->data;
 688     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 689
 690     /* Setup water-specific parameters */
 691     inr              = nlist->iinr[0];
 692     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 693     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 694     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 695     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 696
 697     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 698     rcutoff_scalar   = fr->rcoulomb;
 699     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 700     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 701
 702     /* Avoid stupid compiler warnings */
 703     jnrA = jnrB = 0;
 704     j_coord_offsetA = 0;
 705     j_coord_offsetB = 0;
 706
 707     outeriter        = 0;
 708     inneriter        = 0;
 709
 710     /* Start outer loop over neighborlists */
 711     for(iidx=0; iidx<nri; iidx++)
 712     {
 713         /* Load shift vector for this list */
 714         i_shift_offset   = DIM*shiftidx[iidx];
 715
 716         /* Load limits for loop over neighbors */
 717         j_index_start    = jindex[iidx];
 718         j_index_end      = jindex[iidx+1];
 719
 720         /* Get outer coordinate index */
 721         inr              = iinr[iidx];
 722         i_coord_offset   = DIM*inr;
 723
 724         /* Load i particle coords and add shift vector */
 725         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 726                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 727
 728         fix0             = _fjsp_setzero_v2r8();
 729         fiy0             = _fjsp_setzero_v2r8();
 730         fiz0             = _fjsp_setzero_v2r8();
 731         fix1             = _fjsp_setzero_v2r8();
 732         fiy1             = _fjsp_setzero_v2r8();
 733         fiz1             = _fjsp_setzero_v2r8();
 734         fix2             = _fjsp_setzero_v2r8();
 735         fiy2             = _fjsp_setzero_v2r8();
 736         fiz2             = _fjsp_setzero_v2r8();
 737
 738         /* Start inner kernel loop */
 739         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 740         {
 741
 742             /* Get j neighbor index, and coordinate index */
 743             jnrA             = jjnr[jidx];
 744             jnrB             = jjnr[jidx+1];
 745             j_coord_offsetA  = DIM*jnrA;
 746             j_coord_offsetB  = DIM*jnrB;
 747
 748             /* load j atom coordinates */
 749             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 750                                               &jx0,&jy0,&jz0);
 751
 752             /* Calculate displacement vector */
 753             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 754             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 755             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 756             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 757             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 758             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 759             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 760             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 761             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 762
 763             /* Calculate squared distance and things based on it */
 764             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 765             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 766             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 767
 768             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 769             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 770             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 771
 772             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 773             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 774             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 775
 776             /* Load parameters for j particles */
 777             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 778             vdwjidx0A        = 2*vdwtype[jnrA+0];
 779             vdwjidx0B        = 2*vdwtype[jnrB+0];
 780
 781             fjx0             = _fjsp_setzero_v2r8();
 782             fjy0             = _fjsp_setzero_v2r8();
 783             fjz0             = _fjsp_setzero_v2r8();
 784
 785             /**************************
 786              * CALCULATE INTERACTIONS *
 787              **************************/
 788
 789             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 790             {
 791
 792             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 793
 794             /* Compute parameters for interactions between i and j atoms */
 795             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 796             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 797                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 798
 799             /* Calculate table index by multiplying r with table scale and truncate to integer */
 800             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 801             itab_tmp         = _fjsp_dtox_v2r8(rt);
 802             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 803             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 804             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 805
 806             vfconv.i[0]     *= 8;
 807             vfconv.i[1]     *= 8;
 808
 809             /* REACTION-FIELD ELECTROSTATICS */
 810             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 811
 812             /* CUBIC SPLINE TABLE DISPERSION */
 813             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 814             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 815             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 816             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 817             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 818             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 819             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 820             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 821             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 822
 823             /* CUBIC SPLINE TABLE REPULSION */
 824             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 825             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 826             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 827             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 828             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 829             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 830             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 831             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 832             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 833             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 834
 835             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 836
 837             fscal            = _fjsp_add_v2r8(felec,fvdw);
 838
 839             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 840
 841             /* Update vectorial force */
 842             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 843             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 844             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 845
 846             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 847             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 848             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 849
 850             }
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 857             {
 858
 859             /* Compute parameters for interactions between i and j atoms */
 860             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 861
 862             /* REACTION-FIELD ELECTROSTATICS */
 863             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 864
 865             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 866
 867             fscal            = felec;
 868
 869             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 870
 871             /* Update vectorial force */
 872             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 873             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 874             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 875
 876             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 877             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 878             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 879
 880             }
 881
 882             /**************************
 883              * CALCULATE INTERACTIONS *
 884              **************************/
 885
 886             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 887             {
 888
 889             /* Compute parameters for interactions between i and j atoms */
 890             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 891
 892             /* REACTION-FIELD ELECTROSTATICS */
 893             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 894
 895             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 896
 897             fscal            = felec;
 898
 899             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 900
 901             /* Update vectorial force */
 902             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 903             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 904             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 905
 906             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 907             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 908             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 909
 910             }
 911
 912             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 913
 914             /* Inner loop uses 129 flops */
 915         }
 916
 917         if(jidx<j_index_end)
 918         {
 919
 920             jnrA             = jjnr[jidx];
 921             j_coord_offsetA  = DIM*jnrA;
 922
 923             /* load j atom coordinates */
 924             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 925                                               &jx0,&jy0,&jz0);
 926
 927             /* Calculate displacement vector */
 928             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 929             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 930             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 931             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 932             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 933             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 934             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 935             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 936             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 937
 938             /* Calculate squared distance and things based on it */
 939             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 940             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 941             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 942
 943             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 944             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 945             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 946
 947             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 948             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 949             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 950
 951             /* Load parameters for j particles */
 952             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 953             vdwjidx0A        = 2*vdwtype[jnrA+0];
 954
 955             fjx0             = _fjsp_setzero_v2r8();
 956             fjy0             = _fjsp_setzero_v2r8();
 957             fjz0             = _fjsp_setzero_v2r8();
 958
 959             /**************************
 960              * CALCULATE INTERACTIONS *
 961              **************************/
 962
 963             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 964             {
 965
 966             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 967
 968             /* Compute parameters for interactions between i and j atoms */
 969             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 970             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 971
 972             /* Calculate table index by multiplying r with table scale and truncate to integer */
 973             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 974             itab_tmp         = _fjsp_dtox_v2r8(rt);
 975             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 976             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 977             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 978
 979             vfconv.i[0]     *= 8;
 980             vfconv.i[1]     *= 8;
 981
 982             /* REACTION-FIELD ELECTROSTATICS */
 983             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 984
 985             /* CUBIC SPLINE TABLE DISPERSION */
 986             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 987             F                = _fjsp_setzero_v2r8();
 988             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 989             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 990             H                = _fjsp_setzero_v2r8();
 991             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 992             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 993             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 994             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 995
 996             /* CUBIC SPLINE TABLE REPULSION */
 997             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 998             F                = _fjsp_setzero_v2r8();
 999             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1000             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1001             H                = _fjsp_setzero_v2r8();
1002             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1003             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1004             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1005             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1006             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1007
1008             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1009
1010             fscal            = _fjsp_add_v2r8(felec,fvdw);
1011
1012             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1013
1014             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1015
1016             /* Update vectorial force */
1017             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1018             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1019             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1020
1021             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1022             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1023             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1024
1025             }
1026
1027             /**************************
1028              * CALCULATE INTERACTIONS *
1029              **************************/
1030
1031             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1032             {
1033
1034             /* Compute parameters for interactions between i and j atoms */
1035             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1036
1037             /* REACTION-FIELD ELECTROSTATICS */
1038             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1039
1040             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1041
1042             fscal            = felec;
1043
1044             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1045
1046             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1047
1048             /* Update vectorial force */
1049             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1050             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1051             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1052
1053             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1054             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1055             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1056
1057             }
1058
1059             /**************************
1060              * CALCULATE INTERACTIONS *
1061              **************************/
1062
1063             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1064             {
1065
1066             /* Compute parameters for interactions between i and j atoms */
1067             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1068
1069             /* REACTION-FIELD ELECTROSTATICS */
1070             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1071
1072             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1073
1074             fscal            = felec;
1075
1076             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1077
1078             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1079
1080             /* Update vectorial force */
1081             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1082             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1083             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1084
1085             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1086             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1087             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1088
1089             }
1090
1091             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1092
1093             /* Inner loop uses 129 flops */
1094         }
1095
1096         /* End of innermost loop */
1097
1098         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1099                                               f+i_coord_offset,fshift+i_shift_offset);
1100
1101         /* Increment number of inner iterations */
1102         inneriter                  += j_index_end - j_index_start;
1103
1104         /* Outer loop uses 18 flops */
1105     }
1106
1107     /* Increment number of outer iterations */
1108     outeriter        += nri;
1109
1110     /* Update outer/inner flops */
1111
1112     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*129);
1113 }