src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSh_VdwNone_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_ElecEwSh_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            None
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwNone_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     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  93     real             *ewtab;
  94     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113
 114     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 115     ewtab            = fr->ic->tabq_coul_FDV0;
 116     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 117     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 118
 119     /* Setup water-specific parameters */
 120     inr              = nlist->iinr[0];
 121     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 122     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 123     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 124
 125     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 126     rcutoff_scalar   = fr->rcoulomb;
 127     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 128     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 129
 130     /* Avoid stupid compiler warnings */
 131     jnrA = jnrB = 0;
 132     j_coord_offsetA = 0;
 133     j_coord_offsetB = 0;
 134
 135     outeriter        = 0;
 136     inneriter        = 0;
 137
 138     /* Start outer loop over neighborlists */
 139     for(iidx=0; iidx<nri; iidx++)
 140     {
 141         /* Load shift vector for this list */
 142         i_shift_offset   = DIM*shiftidx[iidx];
 143
 144         /* Load limits for loop over neighbors */
 145         j_index_start    = jindex[iidx];
 146         j_index_end      = jindex[iidx+1];
 147
 148         /* Get outer coordinate index */
 149         inr              = iinr[iidx];
 150         i_coord_offset   = DIM*inr;
 151
 152         /* Load i particle coords and add shift vector */
 153         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 154                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 155
 156         fix0             = _fjsp_setzero_v2r8();
 157         fiy0             = _fjsp_setzero_v2r8();
 158         fiz0             = _fjsp_setzero_v2r8();
 159         fix1             = _fjsp_setzero_v2r8();
 160         fiy1             = _fjsp_setzero_v2r8();
 161         fiz1             = _fjsp_setzero_v2r8();
 162         fix2             = _fjsp_setzero_v2r8();
 163         fiy2             = _fjsp_setzero_v2r8();
 164         fiz2             = _fjsp_setzero_v2r8();
 165
 166         /* Reset potential sums */
 167         velecsum         = _fjsp_setzero_v2r8();
 168
 169         /* Start inner kernel loop */
 170         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 171         {
 172
 173             /* Get j neighbor index, and coordinate index */
 174             jnrA             = jjnr[jidx];
 175             jnrB             = jjnr[jidx+1];
 176             j_coord_offsetA  = DIM*jnrA;
 177             j_coord_offsetB  = DIM*jnrB;
 178
 179             /* load j atom coordinates */
 180             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 181                                               &jx0,&jy0,&jz0);
 182
 183             /* Calculate displacement vector */
 184             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 185             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 186             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 187             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 188             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 189             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 190             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 191             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 192             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 193
 194             /* Calculate squared distance and things based on it */
 195             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 196             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 197             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 198
 199             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 200             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 201             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 202
 203             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 204             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 205             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 206
 207             /* Load parameters for j particles */
 208             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 209
 210             fjx0             = _fjsp_setzero_v2r8();
 211             fjy0             = _fjsp_setzero_v2r8();
 212             fjz0             = _fjsp_setzero_v2r8();
 213
 214             /**************************
 215              * CALCULATE INTERACTIONS *
 216              **************************/
 217
 218             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 219             {
 220
 221             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 222
 223             /* Compute parameters for interactions between i and j atoms */
 224             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 225
 226             /* EWALD ELECTROSTATICS */
 227
 228             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 229             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 230             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 231             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 232             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 233
 234             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 235             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 236             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 237             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 238             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 239             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 240             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 241             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 242             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 243             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 244
 245             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 246
 247             /* Update potential sum for this i atom from the interaction with this j atom. */
 248             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 249             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 250
 251             fscal            = felec;
 252
 253             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 254
 255             /* Update vectorial force */
 256             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 257             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 258             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 259
 260             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 261             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 262             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 263
 264             }
 265
 266             /**************************
 267              * CALCULATE INTERACTIONS *
 268              **************************/
 269
 270             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 271             {
 272
 273             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 274
 275             /* Compute parameters for interactions between i and j atoms */
 276             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 277
 278             /* EWALD ELECTROSTATICS */
 279
 280             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 281             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 282             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 283             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 284             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 285
 286             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 287             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 288             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 289             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 290             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 291             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 292             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 293             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 294             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 295             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 296
 297             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 298
 299             /* Update potential sum for this i atom from the interaction with this j atom. */
 300             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 301             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 302
 303             fscal            = felec;
 304
 305             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 306
 307             /* Update vectorial force */
 308             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 309             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 310             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 311
 312             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 313             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 314             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 315
 316             }
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 323             {
 324
 325             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 329
 330             /* EWALD ELECTROSTATICS */
 331
 332             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 333             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 334             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 335             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 336             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 337
 338             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 339             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 340             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 341             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 342             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 343             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 344             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 345             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 346             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 347             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 348
 349             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 350
 351             /* Update potential sum for this i atom from the interaction with this j atom. */
 352             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 353             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 354
 355             fscal            = felec;
 356
 357             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 358
 359             /* Update vectorial force */
 360             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 361             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 362             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 363
 364             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 365             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 366             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 367
 368             }
 369
 370             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 371
 372             /* Inner loop uses 150 flops */
 373         }
 374
 375         if(jidx<j_index_end)
 376         {
 377
 378             jnrA             = jjnr[jidx];
 379             j_coord_offsetA  = DIM*jnrA;
 380
 381             /* load j atom coordinates */
 382             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 383                                               &jx0,&jy0,&jz0);
 384
 385             /* Calculate displacement vector */
 386             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 387             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 388             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 389             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 390             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 391             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 392             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 393             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 394             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 395
 396             /* Calculate squared distance and things based on it */
 397             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 398             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 399             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 400
 401             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 402             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 403             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 404
 405             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 406             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 407             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 408
 409             /* Load parameters for j particles */
 410             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 411
 412             fjx0             = _fjsp_setzero_v2r8();
 413             fjy0             = _fjsp_setzero_v2r8();
 414             fjz0             = _fjsp_setzero_v2r8();
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 421             {
 422
 423             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 424
 425             /* Compute parameters for interactions between i and j atoms */
 426             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 427
 428             /* EWALD ELECTROSTATICS */
 429
 430             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 431             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 432             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 433             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 434             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 435
 436             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 437             ewtabD           = _fjsp_setzero_v2r8();
 438             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 439             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 440             ewtabFn          = _fjsp_setzero_v2r8();
 441             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 442             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 443             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 444             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 445             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 446
 447             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 448
 449             /* Update potential sum for this i atom from the interaction with this j atom. */
 450             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 451             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 452             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 457
 458             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 459
 460             /* Update vectorial force */
 461             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 462             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 463             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 464
 465             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 466             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 467             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 468
 469             }
 470
 471             /**************************
 472              * CALCULATE INTERACTIONS *
 473              **************************/
 474
 475             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 476             {
 477
 478             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 479
 480             /* Compute parameters for interactions between i and j atoms */
 481             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 482
 483             /* EWALD ELECTROSTATICS */
 484
 485             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 486             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 487             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 488             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 489             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 490
 491             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 492             ewtabD           = _fjsp_setzero_v2r8();
 493             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 494             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 495             ewtabFn          = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 497             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 498             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 499             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 500             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 501
 502             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 503
 504             /* Update potential sum for this i atom from the interaction with this j atom. */
 505             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 506             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 507             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 508
 509             fscal            = felec;
 510
 511             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 512
 513             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 514
 515             /* Update vectorial force */
 516             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 517             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 518             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 519
 520             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 521             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 522             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 523
 524             }
 525
 526             /**************************
 527              * CALCULATE INTERACTIONS *
 528              **************************/
 529
 530             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 531             {
 532
 533             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 534
 535             /* Compute parameters for interactions between i and j atoms */
 536             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 537
 538             /* EWALD ELECTROSTATICS */
 539
 540             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 541             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 542             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 543             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 544             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 545
 546             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 547             ewtabD           = _fjsp_setzero_v2r8();
 548             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 549             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 550             ewtabFn          = _fjsp_setzero_v2r8();
 551             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 552             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 553             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 554             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 555             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 556
 557             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 558
 559             /* Update potential sum for this i atom from the interaction with this j atom. */
 560             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 561             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 562             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 563
 564             fscal            = felec;
 565
 566             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 567
 568             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 569
 570             /* Update vectorial force */
 571             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 572             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 573             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 574
 575             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 576             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 577             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 578
 579             }
 580
 581             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 582
 583             /* Inner loop uses 150 flops */
 584         }
 585
 586         /* End of innermost loop */
 587
 588         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 589                                               f+i_coord_offset,fshift+i_shift_offset);
 590
 591         ggid                        = gid[iidx];
 592         /* Update potential energies */
 593         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 594
 595         /* Increment number of inner iterations */
 596         inneriter                  += j_index_end - j_index_start;
 597
 598         /* Outer loop uses 19 flops */
 599     }
 600
 601     /* Increment number of outer iterations */
 602     outeriter        += nri;
 603
 604     /* Update outer/inner flops */
 605
 606     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*150);
 607 }
 608 /*
 609  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 610  * Electrostatics interaction: Ewald
 611  * VdW interaction:            None
 612  * Geometry:                   Water3-Particle
 613  * Calculate force/pot:        Force
 614  */
 615 void
 616 nb_kernel_ElecEwSh_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 617                     (t_nblist                    * gmx_restrict       nlist,
 618                      rvec                        * gmx_restrict          xx,
 619                      rvec                        * gmx_restrict          ff,
 620                      t_forcerec                  * gmx_restrict          fr,
 621                      t_mdatoms                   * gmx_restrict     mdatoms,
 622                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 623                      t_nrnb                      * gmx_restrict        nrnb)
 624 {
 625     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 626      * just 0 for non-waters.
 627      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 628      * jnr indices corresponding to data put in the four positions in the SIMD register.
 629      */
 630     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 631     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 632     int              jnrA,jnrB;
 633     int              j_coord_offsetA,j_coord_offsetB;
 634     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 635     real             rcutoff_scalar;
 636     real             *shiftvec,*fshift,*x,*f;
 637     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 638     int              vdwioffset0;
 639     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 640     int              vdwioffset1;
 641     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 642     int              vdwioffset2;
 643     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 644     int              vdwjidx0A,vdwjidx0B;
 645     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 646     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 647     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 648     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 649     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 650     real             *charge;
 651     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 652     real             *ewtab;
 653     _fjsp_v2r8       itab_tmp;
 654     _fjsp_v2r8       dummy_mask,cutoff_mask;
 655     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 656     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 657     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 658
 659     x                = xx[0];
 660     f                = ff[0];
 661
 662     nri              = nlist->nri;
 663     iinr             = nlist->iinr;
 664     jindex           = nlist->jindex;
 665     jjnr             = nlist->jjnr;
 666     shiftidx         = nlist->shift;
 667     gid              = nlist->gid;
 668     shiftvec         = fr->shift_vec[0];
 669     fshift           = fr->fshift[0];
 670     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 671     charge           = mdatoms->chargeA;
 672
 673     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 674     ewtab            = fr->ic->tabq_coul_F;
 675     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 676     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 677
 678     /* Setup water-specific parameters */
 679     inr              = nlist->iinr[0];
 680     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 681     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 682     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 683
 684     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 685     rcutoff_scalar   = fr->rcoulomb;
 686     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 687     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 688
 689     /* Avoid stupid compiler warnings */
 690     jnrA = jnrB = 0;
 691     j_coord_offsetA = 0;
 692     j_coord_offsetB = 0;
 693
 694     outeriter        = 0;
 695     inneriter        = 0;
 696
 697     /* Start outer loop over neighborlists */
 698     for(iidx=0; iidx<nri; iidx++)
 699     {
 700         /* Load shift vector for this list */
 701         i_shift_offset   = DIM*shiftidx[iidx];
 702
 703         /* Load limits for loop over neighbors */
 704         j_index_start    = jindex[iidx];
 705         j_index_end      = jindex[iidx+1];
 706
 707         /* Get outer coordinate index */
 708         inr              = iinr[iidx];
 709         i_coord_offset   = DIM*inr;
 710
 711         /* Load i particle coords and add shift vector */
 712         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 713                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 714
 715         fix0             = _fjsp_setzero_v2r8();
 716         fiy0             = _fjsp_setzero_v2r8();
 717         fiz0             = _fjsp_setzero_v2r8();
 718         fix1             = _fjsp_setzero_v2r8();
 719         fiy1             = _fjsp_setzero_v2r8();
 720         fiz1             = _fjsp_setzero_v2r8();
 721         fix2             = _fjsp_setzero_v2r8();
 722         fiy2             = _fjsp_setzero_v2r8();
 723         fiz2             = _fjsp_setzero_v2r8();
 724
 725         /* Start inner kernel loop */
 726         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 727         {
 728
 729             /* Get j neighbor index, and coordinate index */
 730             jnrA             = jjnr[jidx];
 731             jnrB             = jjnr[jidx+1];
 732             j_coord_offsetA  = DIM*jnrA;
 733             j_coord_offsetB  = DIM*jnrB;
 734
 735             /* load j atom coordinates */
 736             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 737                                               &jx0,&jy0,&jz0);
 738
 739             /* Calculate displacement vector */
 740             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 741             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 742             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 743             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 744             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 745             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 746             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 747             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 748             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 749
 750             /* Calculate squared distance and things based on it */
 751             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 752             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 753             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 754
 755             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 756             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 757             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 758
 759             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 760             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 761             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 762
 763             /* Load parameters for j particles */
 764             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 765
 766             fjx0             = _fjsp_setzero_v2r8();
 767             fjy0             = _fjsp_setzero_v2r8();
 768             fjz0             = _fjsp_setzero_v2r8();
 769
 770             /**************************
 771              * CALCULATE INTERACTIONS *
 772              **************************/
 773
 774             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 775             {
 776
 777             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 778
 779             /* Compute parameters for interactions between i and j atoms */
 780             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 781
 782             /* EWALD ELECTROSTATICS */
 783
 784             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 785             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 786             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 787             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 788             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 789
 790             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 791                                          &ewtabF,&ewtabFn);
 792             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 793             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 794
 795             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 796
 797             fscal            = felec;
 798
 799             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 800
 801             /* Update vectorial force */
 802             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 803             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 804             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 805
 806             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 807             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 808             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 809
 810             }
 811
 812             /**************************
 813              * CALCULATE INTERACTIONS *
 814              **************************/
 815
 816             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 817             {
 818
 819             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 820
 821             /* Compute parameters for interactions between i and j atoms */
 822             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 823
 824             /* EWALD ELECTROSTATICS */
 825
 826             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 827             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 828             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 829             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 830             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 831
 832             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 833                                          &ewtabF,&ewtabFn);
 834             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 835             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 836
 837             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 838
 839             fscal            = felec;
 840
 841             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 842
 843             /* Update vectorial force */
 844             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 845             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 846             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 847
 848             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 849             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 850             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 851
 852             }
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 859             {
 860
 861             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 862
 863             /* Compute parameters for interactions between i and j atoms */
 864             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 865
 866             /* EWALD ELECTROSTATICS */
 867
 868             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 869             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 870             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 871             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 872             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 873
 874             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 875                                          &ewtabF,&ewtabFn);
 876             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 877             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 878
 879             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 880
 881             fscal            = felec;
 882
 883             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 884
 885             /* Update vectorial force */
 886             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 887             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 888             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 889
 890             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 891             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 892             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 893
 894             }
 895
 896             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 897
 898             /* Inner loop uses 129 flops */
 899         }
 900
 901         if(jidx<j_index_end)
 902         {
 903
 904             jnrA             = jjnr[jidx];
 905             j_coord_offsetA  = DIM*jnrA;
 906
 907             /* load j atom coordinates */
 908             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 909                                               &jx0,&jy0,&jz0);
 910
 911             /* Calculate displacement vector */
 912             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 913             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 914             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 915             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 916             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 917             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 918             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 919             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 920             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 921
 922             /* Calculate squared distance and things based on it */
 923             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 924             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 925             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 926
 927             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 928             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 929             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 930
 931             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 932             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 933             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 934
 935             /* Load parameters for j particles */
 936             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 937
 938             fjx0             = _fjsp_setzero_v2r8();
 939             fjy0             = _fjsp_setzero_v2r8();
 940             fjz0             = _fjsp_setzero_v2r8();
 941
 942             /**************************
 943              * CALCULATE INTERACTIONS *
 944              **************************/
 945
 946             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 947             {
 948
 949             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 950
 951             /* Compute parameters for interactions between i and j atoms */
 952             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 953
 954             /* EWALD ELECTROSTATICS */
 955
 956             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 957             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 958             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 959             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 960             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 961
 962             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 963             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 964             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 965
 966             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 967
 968             fscal            = felec;
 969
 970             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 971
 972             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 973
 974             /* Update vectorial force */
 975             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 976             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 977             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 978
 979             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 980             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 981             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 982
 983             }
 984
 985             /**************************
 986              * CALCULATE INTERACTIONS *
 987              **************************/
 988
 989             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 990             {
 991
 992             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 993
 994             /* Compute parameters for interactions between i and j atoms */
 995             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 996
 997             /* EWALD ELECTROSTATICS */
 998
 999             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1000             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1001             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1002             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1003             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1004
1005             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1006             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1007             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1008
1009             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1010
1011             fscal            = felec;
1012
1013             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1014
1015             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1016
1017             /* Update vectorial force */
1018             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1019             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1020             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1021
1022             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1023             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1024             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1025
1026             }
1027
1028             /**************************
1029              * CALCULATE INTERACTIONS *
1030              **************************/
1031
1032             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1033             {
1034
1035             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1036
1037             /* Compute parameters for interactions between i and j atoms */
1038             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1039
1040             /* EWALD ELECTROSTATICS */
1041
1042             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1043             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1044             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1045             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1046             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1047
1048             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1049             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1050             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1051
1052             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1053
1054             fscal            = felec;
1055
1056             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1057
1058             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1059
1060             /* Update vectorial force */
1061             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1062             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1063             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1064
1065             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1066             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1067             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1068
1069             }
1070
1071             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1072
1073             /* Inner loop uses 129 flops */
1074         }
1075
1076         /* End of innermost loop */
1077
1078         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1079                                               f+i_coord_offset,fshift+i_shift_offset);
1080
1081         /* Increment number of inner iterations */
1082         inneriter                  += j_index_end - j_index_start;
1083
1084         /* Outer loop uses 18 flops */
1085     }
1086
1087     /* Increment number of outer iterations */
1088     outeriter        += nri;
1089
1090     /* Update outer/inner flops */
1091
1092     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*129);
1093 }